CN119091343A

CN119091343A - A method and system for predicting and warning defective products of manually assembled motors

Info

Publication number: CN119091343A
Application number: CN202410979885.8A
Authority: CN
Inventors: 张星; 张涛; 张存利
Original assignee: China Pilot Motor Shenzhen Co ltd
Current assignee: China Pilot Motor Shenzhen Co ltd
Priority date: 2024-07-22
Filing date: 2024-07-22
Publication date: 2024-12-06

Abstract

A defective product prediction early warning method and system for a manual assembly motor relate to the field of industrial detection, and the method comprises the steps of acquiring a real-time video stream in the motor assembly process and determining key action characteristics of a target person according to the real-time video stream; the method comprises the steps of comparing key action features with a preset standard action template to obtain action difference, establishing an applicability model of an assembly tool according to tool types, performance parameters and use state information of the assembly tool used by a target person, inputting the key action features into the applicability model to obtain tool application scores of the target person, acquiring real-time pressure data and real-time temperature data of an assembly environment in a motor assembly process, determining motor assembly scores of the target person, and sending early warning information to a supervision client when the motor assembly scores are lower than a preset qualified threshold. By implementing the method, the application degree of assembly personnel to the assembly tool can be identified, and the defective products of the motor can be predicted more comprehensively.

Description

Defective product prediction early warning method and system for manual assembly motor

Technical Field

The application relates to the field of industrial detection, in particular to a defective product prediction and early warning method and system for a manually assembled motor.

Background

In the manufacturing industry, product quality control is a key link for guaranteeing product reliability, safety and customer satisfaction. With the increase of market competition and the diversification of consumer demands, enterprises are increasingly focusing on improving the quality management efficiency in the production process. Especially in the manual assembly field of motors, the production of defective products often leads to increased cost and reduced production efficiency due to complex assembly process and susceptibility to human factors. Therefore, development of an efficient and accurate defective product prediction and early warning system becomes an important means for improving production quality.

In the related art, defective product prediction techniques mainly rely on automated visual inspection systems and machine learning methods. These systems capture images during assembly by high resolution cameras and use image recognition techniques to detect defects on the surface of the product. In addition, some advanced detection systems incorporate sensor data, such as temperature, pressure, etc., for multi-dimensional monitoring by data analysis software.

However, manual assembly of the motor is greatly affected by human factors, the technical proficiency of different assembly personnel is different, and the application degree of different assembly tools is also different, so that the quality of the product of the manual assembly of the motor may be uneven. Defective product prediction technology in the related art is difficult to effectively identify the application degree of assembly staff to an assembly tool, and the prediction of defective products is not comprehensive enough.

Disclosure of Invention

The application provides a defective product prediction early warning method and system for a manually assembled motor, which are used for recognizing the application degree of assembly staff on an assembly tool and predicting defective products of the motor more comprehensively.

The application provides a defective product prediction early warning method of a manual assembly motor, which is applied to a prediction early warning system and comprises the steps of acquiring a real-time video stream in the motor assembly process and determining key action characteristics of a target person according to the real-time video stream; the key action features comprise hand action tracks, force, speed and frequency of a target person, action difference degrees are obtained by comparing the key action features with preset standard action templates, an applicability model of an assembly tool is built according to tool types, performance parameters and use state information of the assembly tool used by the target person, the key action features are input into the applicability model to obtain tool application scores of the target person, real-time pressure data and real-time temperature data of an assembly environment in a motor assembly process are obtained, motor assembly scores of the target person are determined according to the action difference degrees, the real-time pressure data, the real-time temperature data and the tool application scores, and early warning information is sent to a supervision client when the motor assembly scores are lower than preset qualified thresholds.

In the above embodiment, the predictive early warning system analyzes the key motion characteristics of the target person by acquiring the real-time video stream, and compares the key motion characteristics with the standard motion template to obtain the motion difference degree. Meanwhile, an applicability model is built according to various attributes of an assembly tool used by a target person, and key action features are input into the model to obtain tool application scores. And then combining the real-time pressure and temperature data of the assembly environment, and integrating action difference degree, environment data and tool application scores to determine motor assembly scores of target personnel. And sending out early warning information when the score is lower than a preset threshold value. The method not only considers the standardization of personnel actions, but also considers the use suitability of tools and the influence of assembly environment, evaluates the assembly quality in a multi-dimensional manner, more comprehensively predicts the risk of defective products of the motor, can early warn in time, and reduces the generation of defective products.

In combination with some embodiments of the first aspect, in some embodiments, an applicability model of the assembly tool is built according to a tool type, performance parameters and use state information of the assembly tool used by a target person, and specifically comprises the steps of obtaining the tool type of the assembly tool used by the target person, wherein the tool type comprises an electric tool and a manual tool, obtaining performance parameters of the assembly tool, wherein the performance parameters comprise rated power, rotating speed, torque, material hardness and dimension specifications, obtaining use state information of the assembly tool, wherein the use state information comprises actual use duration, maintenance record and fault frequency of the tool, dividing the complexity value of the assembly tool according to the tool type, determining the application range of the assembly tool according to the performance parameters, determining the reliability value of the assembly tool according to the use state information, and integrating the complexity value, the application range and the reliability value of the assembly tool to build a three-dimensional applicability evaluation model of the assembly tool as the applicability model.

In the above embodiment, the prediction early warning system acquires information such as the tool type, the performance parameter, the use state and the like, then quantifies the attribute of the tool from three dimensions of complexity, application range and reliability, and finally synthesizes the three dimensions to establish a three-dimensional applicability evaluation model. The characteristics and actual use conditions of the tool are systematically considered, and the influence of the tool on the assembly quality is more accurately evaluated. Through finer tool applicability modeling, tool application conditions of personnel can be scored more accurately, and comprehensiveness and accuracy of defective product prediction are improved.

In combination with some embodiments of the first aspect, in some embodiments, acquiring real-time pressure data and real-time temperature data of an assembly environment in a motor assembly process, and determining a motor assembly score of a target person according to action difference degree, the real-time pressure data, the real-time temperature data and tool application scores, wherein the method specifically comprises the steps of detecting the real-time pressure data of the assembly environment in real time through a pressure sensor, and comparing the real-time pressure data with a preset standard pressure range to obtain pressure difference degree; the method comprises the steps of detecting real-time temperature data of an assembly environment in real time through a temperature sensor, comparing the real-time temperature data with a preset standard temperature range to obtain a temperature difference degree, and inputting the action difference degree, the pressure difference degree, the temperature difference degree and a tool application score into a preset motor quality evaluation model to obtain motor assembly scores of target personnel.

In the above embodiment, the predictive early warning system detects the pressure and temperature data in real time through the sensor, and compares the pressure and temperature data with a preset standard range to obtain the difference degree of the pressure and the temperature. And inputting the action difference degree, the pressure difference degree, the temperature difference degree and the tool application score into a preset motor quality evaluation model to obtain a final motor assembly score. Environmental data is fully utilized, so that evaluation is more comprehensive and objective, and meanwhile, the quality evaluation model is used for quantifying the influence weight of each index on the assembly quality, so that scoring is more scientific and reasonable. The defective product prediction early warning method is further perfected by refining the determination process of the assembly scores, and the accuracy and reliability of prediction are improved.

In combination with some embodiments of the first aspect, in some embodiments, before the step of acquiring a real-time video stream in the motor assembly process and determining key action features of the target person according to the real-time video stream, the method further includes acquiring historical assembly data of the target person, wherein the historical assembly data includes assembly time lengths, assembly qualification rates and bad reasons of a plurality of products, acquiring physiological feature parameters of the target person, wherein the physiological feature parameters include gender, age, eyesight and response speed, inputting the historical assembly data and the physiological feature parameters into a preset personal skill model to obtain the assembly skill level of the target person, determining the target person with the assembly skill level lower than a preset level threshold as an important monitoring object, and improving video acquisition frequency and action recognition accuracy of the important monitoring object.

In the above embodiment, the predictive early warning system obtains the historical assembly data and the physiological characteristic parameters of the target person, and inputs the historical assembly data and the physiological characteristic parameters into the preset personal skill model to obtain the assembly skill level assessment result. For people with lower skill level, the people are determined to be important monitoring objects, and the video acquisition frequency and the motion recognition accuracy of the people are improved. The method for adjusting the monitoring strategy according to the personnel skill difference can prevent defective products more pertinently. Through discernment high risk personnel in advance, strengthen the control dynamics to its assembly process, reduce the defective products rate from the source. Meanwhile, limited monitoring resources can be more intensively put into key personnel, so that the monitoring efficiency is improved and the cost is saved while the early warning effect is ensured.

In combination with some embodiments of the first aspect, in some embodiments, acquiring physiological characteristic parameters of the target person specifically includes acquiring facial feature images and action feature images of the target person through a camera, evaluating vision level, concentration and fatigue of the target person according to the facial feature images and the action feature images, determining hand flexibility of the target person according to the action feature images, and integrating the vision level, concentration, fatigue and hand flexibility of the target person to obtain the physiological characteristic parameters.

In the above embodiment, the predictive early warning system obtains the facial and action feature images of the person through the camera, and obtains the indexes such as eyesight, concentration, fatigue, hand flexibility and the like through analysis and evaluation, and then integrates the indexes into the final physiological feature parameters. The image information is fully utilized, and the physiological states of the personnel are evaluated at multiple angles through the technical means of face recognition, expression analysis, motion capture and the like, so that the quantification of the physiological characteristics is finer and finer. And meanwhile, an additional physiological sensor is avoided, and the complexity of the system is reduced. Physical quality factors are more comprehensively included in personal skill assessment, objective conditions of staff are considered, and abundant background information is provided for subsequent assembly process monitoring, so that the assembly quality is more humanized and personalized, and the assembly quality is better guided to be improved.

In combination with some embodiments of the first aspect, in some embodiments, after the step of sending the early warning information to the supervision client when the motor assembly score is lower than the preset qualification threshold, the method further includes pushing an assembly action video clip of the target person in the early warning information, determining a feedback opinion returned by the supervision client after receiving the early warning information, generating an abnormal motor report for recording an abnormal reason, a responsible person and lost man-hour if the feedback opinion is that the motor needs to be reworked or scrapped, performing statistical analysis on assembly quality of a plurality of production links within a preset time period according to an accumulated record of the abnormal motor report to obtain a quality improvement suggestion, and generating a work adjustment scheme including a station layout and staff skill requirements.

In the embodiment, the predictive early warning system firstly pushes a problem video segment in early warning information to help a supervisor to quickly locate the problem, then generates an abnormal motor report according to feedback comments of the supervisor if reworking or scrapping is needed, records personnel, time, loss and the like in detail, performs statistical analysis on the whole production process according to the abnormal report in a period of time to obtain a quality improvement suggestion, and finally forms an optimized work adjustment scheme. The passive early warning response is extended to active quality improvement, and the quality management concept of continuous improvement with the main prevention is reflected. Through data accumulation and flow optimization, the quality problem is solved from discovery to formation of a closed loop, so that the information circulation is smoother, and the system is more mature and perfect.

In combination with some embodiments of the first aspect, in some embodiments, according to the accumulated records of the abnormal motor reports, the assembly quality of the plurality of production links is statistically analyzed within a preset time period to obtain quality improvement suggestions, and the quality improvement suggestions specifically include performing associated mining on the abnormal motor reports, the production task list, the process parameters and the equipment state to construct a digital twin model of the whole assembly production process, performing simulation deduction on the assembly process of the plurality of production links through the digital twin model within the preset time period to determine quality risks and corresponding influence degrees of the plurality of production links, and determining optimized production process parameters and process control schemes according to the quality risks to obtain the quality improvement suggestions.

In the embodiment, the prediction and early warning system firstly collects various data to construct a digital twin model of assembly production, then utilizes the model to simulate and deduce the assembly process within a period of time to identify the quality risk and influence of each link, and finally optimizes the production process parameters and the process control scheme to form an improvement suggestion. The quality analysis method by means of the digital twin technology can fully excavate and correlate various data in the production process, simulate the production condition through virtual simulation and predict the hidden quality trouble. The analysis period is shortened, the analysis cost is reduced, and meanwhile, the effects of different optimization schemes can be evaluated through the model, so that the optimal quality improvement path is found out. The digital means and quality management are fully combined, so that the manufacturing process is more transparent and controllable, and the quality optimization is more scientific and efficient.

In a second aspect, embodiments of the present application provide a predictive early warning system comprising one or more processors and a memory coupled to the one or more processors, the memory for storing computer program code comprising computer instructions that the one or more processors invoke to cause the predictive early warning system to perform a method as described in the first aspect and any of the possible implementations of the first aspect.

In a third aspect, embodiments of the present application provide a computer program product comprising instructions that, when run on a predictive early warning system, cause the predictive early warning system to perform a method as described in the first aspect and any possible implementation of the first aspect.

In a fourth aspect, embodiments of the present application provide a computer readable storage medium comprising instructions that, when executed on a predictive early warning system, cause the predictive early warning system to perform a method as described in the first aspect and any possible implementation of the first aspect.

It will be appreciated that the predictive early warning system provided in the second aspect, the computer program product provided in the third aspect and the computer storage medium provided in the fourth aspect are each configured to perform the method provided by the embodiment of the present application. Therefore, the advantages achieved by the method can be referred to as the advantages of the corresponding method, and will not be described herein.

One or more technical solutions provided in the embodiments of the present application at least have the following technical effects or advantages:

1. By adopting the method of analyzing key action characteristics of target personnel according to the real-time video stream and comparing the key action characteristics with the standard template and combining the applicability evaluation of the assembly tool and the real-time monitoring data of the assembly environment, the motor assembly quality score is comprehensively determined, so that the quality risk of the motor assembly process can be comprehensively evaluated from multiple dimensions such as personnel operation, tool use, environmental conditions and the like, the limitation that the assembly quality problem caused by the fact that the product surface defect detection is only relied on and the artificial factors is difficult to identify in the prior art is effectively solved, and further, the prediction and the early warning of the more comprehensive, accurate and forward defective products are realized. The method considers not only the standardization of personnel operation, but also the appropriateness of tool selection and the influence of environmental parameters, and performs finer and finer excavation on the causes of defective products through multi-source data fusion and intelligent analysis, so that the hidden quality hazards are eliminated in a sprouting state, and the quality management concept mainly comprising prevention is fully exerted.

2. Because the personal skill assessment is carried out according to the personnel history assembly data and the physiological characteristic parameters, and the key monitoring method is given to the personnel with lower skill level, the high risk personnel with weak assembly skill can be identified in advance, the supervision and guidance on the operation process of the personnel can be enhanced in a targeted manner, the personnel skill difference neglected in the prior art is effectively solved, and further the fine management of point-surface combination according to the difference of the personnel is realized. By means of the personal skill model, the method quantifies factors such as historical performance, physical quality and the like of staff into skill scores, and then the staff is classified and monitored according to the scores, so that quality management resources are saved, and early warning accuracy is improved.

3. The method adopts the closed loop processing flow after the early warning information is sent, and the method is fed back to the generation of an abnormal report from a supervisory person and then to the whole quality analysis and process optimization, so that the method can quickly respond after the problem is found, the correction measures are formulated, the teaching and training is absorbed from the correction measures, the production system and the process are continuously perfected, the defects that the early warning processing flows in the form, the problem is difficult to form, the closed loop is solved, and the quality improvement measures cannot fall to the ground are effectively overcome, and the whole quality management target is further realized. The method tightly combines early warning information feedback and production practice, so that the problem response is more efficient. Meanwhile, the transverse and longitudinal circulation of information between personnel and departments is smoother, so that the cooperative efficiency is improved, and the quality management system is improved.

Drawings

FIG. 1 is a schematic flow chart of a defective product prediction and early warning method for a manually assembled motor in an embodiment of the application;

FIG. 2 is another flow chart of a defective product prediction and early warning method for a manually assembled motor according to an embodiment of the present application;

Fig. 3 is a schematic structural diagram of an entity device of a predictive early warning system according to an embodiment of the application.

Detailed Description

The terminology used in the following embodiments of the application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. As used in the specification of the present application, the singular forms "a", "an", "the" and "the" are intended to include the plural forms as well, unless the context clearly indicates to the contrary. It should also be understood that the term "and/or" as used in this disclosure is intended to encompass any or all possible combinations of one or more of the listed items.

The terms "first," "second," and the like, are used below for descriptive purposes only and are not to be construed as implying or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defining "a first" or "a second" may explicitly or implicitly include one or more such feature, and in the description of embodiments of the application, unless otherwise indicated, the meaning of "a plurality" is two or more.

In order to facilitate understanding, an application scenario of the embodiment of the present application is described below.

A final assembly shop of a large motor manufacturing company mainly produces a plurality of series of brushless motors, speed reducing motors and stepping motors. Because market demands are changeable, the product is updated frequently, so that various motor models are produced by a mixed line, and a large number of manual assembly processes are adopted. However, with the improvement of the quality requirements of users and the aggravation of competition in the same industry, the problem of defective products in the assembly link is increasingly prominent. Statistics show that reworking and scrapping losses caused by assembly quality problems are up to millions of yuan each month, and delivery delays also occur.

In the related art, supervision and management of assembly quality can be realized by adopting a mode of combining post-sampling detection and manual inspection. The method comprises the steps of detecting the product in a destructive or nondestructive mode to judge whether the assembly quality is qualified, and meanwhile, arranging a special person to carry out regular or irregular inspection on an assembly site, directly observing the operation standardization of assembly staff, and carrying out examination and guidance on the assembly staff. A scenario of a defective product prediction early warning method using a manually assembled motor in the related art is described below.

The technical engineer of the technical center of the motor company is the king and is responsible for optimizing the assembly process of the stepping motor. He knows that the qualification rate of the assembly line after one time is 92% in the last year, and thousands of defective products are returned due to the quality problems such as case opening, case damage, functional failure and the like. After investigation, the king finds that the problems of irregular operation, wrong tools and the like commonly exist in the assembly, such as interface damage caused by too strong force when connecting wire harnesses or screw tooth sliding caused by wrong use of a screwdriver.

By adopting the defective product prediction early warning method of the manual assembly motor, the spanning from passive detection to active prevention is realized through intelligent sensing, data fusion analysis and risk early warning feedback closed loop in the assembly process, the generation of defective products can be reduced to the maximum extent, the quality cost saving is realized, the cause of quality problems can be accurately positioned, the assembly process is continuously optimized and perfected, and the double improvement of the product quality and the production efficiency is realized.

The following describes a scenario in which the defective product prediction and early warning method of the manually assembled motor of the present application is used.

After the scheme of the application is introduced by motor companies, the king is responsible for specific implementation. The stepping motor production line is provided with a set of visual perception equipment, and high-definition videos in the assembly process can be collected in real time. By means of a hand motion recognition algorithm, the system can accurately analyze key features such as hand track, force, speed and frequency of an assembler, and the key features are compared with a pre-recorded standard motion template to recognize the completion degree of standard motion. When the deviation is too large, the system can automatically give an early warning. Meanwhile, the system is also connected with a digital management system of the assembly tool, so that the applicability of the tool for staff to get can be accurately judged. In addition, an environmental sensor is added to monitor the temperature and humidity, the cleanliness and the like of the assembly site on line. The system comprehensively analyzes the influence of assembly actions, tool applicability and environmental factors on assembly quality, obtains quality risk scores and gives early warning in time. After half a year application, the one-time offline qualification rate of the assembly line of the stepping motor is improved to 98%, the quality problems of product unpacking, damage and the like are greatly reduced, and the repair cost is reduced by more than 20%.

Therefore, by adopting the defective product prediction and early warning method for the manually assembled motor, disclosed by the embodiment of the application, the problem of hidden quality hazards caused by extensive management in the assembly process can be effectively solved while defective product prediction and early warning are realized, so that the whole flow and whole element management and control of the product quality are realized, the assembly production is promoted to develop towards the intelligent and refined directions, and the method has important significance for improving the market competitiveness of manufacturing enterprises.

It should be noted that, the predictive early warning system described herein refers to a computer with data collection and processing functions, and its specific implementation is similar to a server, and for convenience of description, the system will be abbreviated in some cases.

For ease of understanding, the method provided in this embodiment is described in the following in conjunction with the above scenario. Referring to fig. 1, a flow chart of a defective product prediction and early warning method for a manually assembled motor according to an embodiment of the application is shown.

S101, acquiring a real-time video stream in the motor assembly process, and determining key action characteristics of a target person according to the real-time video stream.

The predictive early warning system is used for acquiring video image information of an assembly process in real time by arranging high-definition cameras at key positions of a manual assembly production line. The system utilizes computer vision and image processing technology to analyze the video stream frame by frame, and accurately positions and identifies the hand area of each assembly staff through a target detection and tracking algorithm. On the basis, the system comprehensively utilizes algorithms such as hand skeleton key point extraction, track tracking, action segmentation and the like, dynamically extracts key action characteristic parameters of the hands of assembly staff, and mainly comprises hand movement tracks, speeds and accelerations, finger opening and closing angles and frequencies, wrist gestures and rotation angles and the like. Meanwhile, the information such as the category, the pose and the like of the assembly object can be identified.

For example, when the brushless motor is assembled, the system can obtain parameters such as hand movement track, speed and the like when the stator is installed through hand action analysis of assembly staff, and judge whether the stator is taken according to a standard posture or not. The action features extracted by the system form time series data, and a basis is provided for subsequent assembly quality evaluation.

S102, comparing the key action features with a preset standard action template to obtain action difference degree.

After the key characteristics of the hand actions of the assembly staff are obtained, the predictive early warning system compares the key characteristics with the standard action templates which are pre-recorded into the system for analysis, and the normalization and standardization level of the action execution is obtained. Specifically, the standard action template is a standard action video which is demonstrated and recorded by a process expert and a senior staff, and the system extracts action characteristic parameters of the standard action template as the golden position for judging the assembly action of the staff. The system adopts a time sequence similarity algorithm, such as a Dynamic Time Warping (DTW) algorithm, and the like, to align and match an actual action sequence of staff with a standard template, calculates action characteristic deviation values of each time step, and comprehensively considers weights of the deviation values to obtain the comprehensive difference degree of the whole action process. The degree of difference can be quantified in terms of percentage, and the larger the degree of difference is, the larger the degree of deviation between the actual operation of staff and a standard action template is, namely, the less standard the assembly action is, and the higher the quality risk is. By comparing the employee actions with the standard templates in real time, the system can accurately judge the execution condition of each assembly action, discover the nonstandard actions in time and provide data support for deviation correction.

For example, when the system finds that a staff is installing a planetary gear of a gear motor, when the difference degree between the hand movement track and a standard template is up to 30%, the system judges that the staff is likely to have the problems of inaccurate installation position, overlarge action range and the like, and further triggers early warning to prompt a manager to intervene in guidance.

S103, establishing an applicability model of the assembly tool according to the tool type, the performance parameters and the use state information of the assembly tool used by the target personnel.

In the motor assembly process, the selection and the use condition of the assembly tool have direct influence on the product quality. Therefore, the predictive early warning system comprehensively considers factors such as the type, performance parameters, use states and the like of the assembly tool, and establishes a tool applicability evaluation model. Firstly, the system needs to acquire and input static attribute parameters such as model specifications, precision grades, measuring range, validity period and the like of different types of assembly tools (such as screwdrivers, socket wrenches, jacks and the like). Secondly, the system needs to collect the use state data of the assembly tool in real time, and mainly comprises tool receiving records, use frequency, accumulated use duration, periodic spot check calibration records and the like. After the above tool attribute and state data are obtained, the system utilizes a machine learning algorithm (such as a Support Vector Machine (SVM), a random forest, etc.), and establishes a tool applicability evaluation model, i.e. a correlation model of each parameter of the tool and the assembly quality. The model can output a tool suitability score to quantitatively evaluate the rationality of tool selection and use.

For example, by analyzing the use history data of a certain socket wrench, it is found that the accumulated use time period is longer than 1000 hours, and the socket wrench is used for high-strength tightening operation for a long time, and the accuracy level is also lower, so that the applicability score is lower, and the screw thread is likely to be damaged and the torque is likely to be missed by continued use. When the system judges that the applicability score of the tool used by the assembly staff is lower than a preset threshold based on the model, the early warning prompts that the tool needs to be replaced or maintained.

The input data of the applicability model includes tool types (e.g., power tools and hand tools), performance parameters (e.g., power rating, rotational speed, torque, material hardness, and dimensional specifications), and usage status information (e.g., actual usage time of the tool, maintenance records, and failure frequency) of the assembly tools used by the target personnel. Through analysis of these data, the model can output the complexity value, application range and reliability value of the assembly tool. Finally, combining the numerical values of the three dimensions, the model can give a three-dimensional applicability evaluation result of the assembly tool. The model is trained by adopting a machine learning algorithm such as a Support Vector Machine (SVM), a random forest and the like, and the training standard is that the model can accurately evaluate the rationality of tool selection and use.

S104, inputting the key action features into the applicability model to obtain tool application scores of target personnel.

After the assembly tool applicability evaluation model is established, the predictive early warning system inputs key action characteristic parameters of assembly staff into the model, and the tool application condition of the staff is automatically evaluated. In particular, the system focuses on hand motion characteristics associated with tool use, such as wrist application force magnitude and direction, finger grip position and angle, and the like. The system performs matching analysis on the action characteristic data and attribute parameters such as model specification, precision grade and the like of the tool, and judges whether the staff uses the tool by adopting a correct operation method according to the requirements of an operation instruction. Meanwhile, the system can evaluate the proficiency of staff using tools and judge the proficiency by the characteristics of consistency, stability and the like of hand actions. After comprehensively considering the suitability of tool selection and the normalization of the use method, the model outputs the tool operation score of the employee, and quantitatively reflects the rationality and normalization of the tool use. The scoring result may be a percentage or may be represented by a hierarchy (e.g., A, B, C, D, etc.). When the tool application score of staff is low, the system can early warn to prompt that the staff may have the problems of wrong tool selection, improper operation method and the like, and management staff is required to provide necessary guidance and training.

S105, acquiring real-time pressure data and real-time temperature data of an assembly environment in the motor assembly process, and determining motor assembly scores of target personnel according to action difference degrees, the real-time pressure data, the real-time temperature data and tool application scores.

Fluctuations in the environmental conditions of the assembly operation can have an impact on the quality of the product. Therefore, the predictive early warning system can also collect pressure and temperature data of the assembly workshop environment in real time to be used as an auxiliary reference for judging the assembly quality. The system can continuously monitor and record the pressure and temperature change conditions of the assembly environment by deploying the wireless pressure sensor and the temperature sensor on the assembly site and establishing communication connection with the data acquisition terminal. After the environmental parameter data are obtained, the system performs association analysis on indexes such as action difference degree of assembly staff and tool application scores, and a comprehensive quality evaluation model of the assembly process is constructed. The model adopts a weighted scoring method, different weight coefficients are given to each index, and finally the comprehensive quality score of staff assembly operation, namely the motor assembly score, is obtained.

For example, the system assigns a 40% weight to the action variance, the tool applies a 30% weight score, and the ambient pressure and temperature data each account for 15% weight. When a worker assembles the rotor of the stepping motor, the action difference degree is 10%, the tool application score is B, meanwhile, the assembly time pressure is 101kPa, and the temperature is 25 ℃ and is within a reasonable range. After weighted calculation, the motor assembly score of the staff is 85 points, which indicates that the assembly quality is good and the possibility of qualified products is high. By collecting multi-source heterogeneous data in the assembly process and performing comprehensive judgment, the system can monitor the assembly quality condition more comprehensively and dynamically, and reduces the evaluation blind area.

And S106, when the motor assembly score is lower than a preset qualification threshold, sending early warning information to the supervision client.

According to the motor assembly scores of assembly staff, the prediction early warning system can judge the quality risk level of the assembly process in real time and determine whether to send early warning information. Specifically, the system presets a qualification threshold (such as 80 points) of the motor assembly score, and when the score of staff is lower than the threshold, the system considers that the motor has a failure risk and needs early warning prompt and timely intervention. The early warning information can be automatically pushed to mobile terminal APP or PC terminal software of workshop supervisory personnel, and assembly quality abnormality is prompted in modes of popup window reminding, voice broadcasting, signal lamp flashing and the like. Meanwhile, the system can push early warning information and the video clips in the assembly process together, so that supervision staff can visually see that problems occur, and the system can quickly respond. Of course, for general pre-warnings with scores slightly below the threshold, the system will only alert, while for severe pre-warnings with scores well below the threshold, the system will require immediate stopping of the assembly operation, full inspection of the motor, and retraining of the assembly staff.

For example, when a worker assembles an encoder of a servo motor, the motor assembly score is only 65, the system can immediately send early warning information to a workshop owner, and video clips with the problems of collision, insufficient fastening moment and the like when the worker assembles the encoder are attached, so that the motor is required to be rechecked and criticized education is carried out on the worker. Through timely early warning feedback, the system can eliminate the quality hidden trouble in the sprouting state, avoids the defective products to flow into the next procedure to the maximum extent, and truly realizes the forward movement of quality management.

In the above embodiment, by combining visual perception, sensing detection and algorithm analysis, the personnel actions, tool applicability, environmental impact and the like in the assembly process are evaluated in real time, and the assembly quality risk is comprehensively perceived. In practical application, the application scene and effect of the method can be further expanded by introducing functional modules such as personnel physiological state detection, production line production plan identification, digital twin simulation and the like.

For example, the influence of the fatigue state of staff on the assembly quality is judged through a staff fatigue evaluation model, the quality standards and the assembly requirements of different products are distinguished through assembly line production plan management, and then, the assembly difficulty, the quality risk and the like are prejudged in advance in the product design and process planning stages by utilizing a digital twin technology.

The following supplements the scenario of the present embodiment.

In view of the good application effect of the stepper motor assembly line, the motor company decides to popularize the system in all motor assembly lines. But further optimization is required in consideration of the characteristic difference of different production lines. For example, in brushless motor assembly lines, it is important to pay attention to the fatigue state of the assembly staff, as this is related to key quality characteristics such as coil winding tension. Thus, the king team adds an employee fatigue assessment model to the system. The fatigue degree of the staff is judged by combining the factors of the age, the work age, the working time of the work and the like of the staff, and the fatigue degree is related to the assembly action performance of the staff so as to more comprehensively evaluate the risk of the assembly process. In addition, in the assembly line of the gear motor, the production states of different motor models need to be monitored in a distinguishing mode.

The team further develops a production line production plan management module, can automatically identify the current motor model of the production line, and call corresponding quality standards and process parameters, so that the accurate evaluation of the assembly process of different motor models is realized. Customized intelligent reports are developed for different production lines, quality trends and risk distribution in the assembly process are presented in a visual mode, a manager is helped to better get insight into quality conditions, and production organization is optimized.

In the production preparation stage of a new servo motor project, the system can also discover design and process defects in advance through process simulation and simulate an assembly process, and the quality risk is prevented from an earlier stage. The continuous optimization and the expansion application of the system lay a solid foundation for the company to realize the quality control of the full value chain, and strongly support the high-quality development of the enterprise.

In combination with the above scenario, a further more specific flow of the method provided in this embodiment will be described below. Fig. 2 is a schematic flow chart of a defective product prediction and early warning method for a manually assembled motor according to an embodiment of the application.

S201, acquiring a real-time video stream in the motor assembly process, and determining key action characteristics of a target person according to the real-time video stream.

Referring to step S101, the predictive early warning system determines key motion characteristics of the target person.

In some embodiments, a predictive early warning system acquires historical assembly data of a target person, wherein the historical assembly data comprises assembly time, assembly qualification rate and bad reasons of a plurality of products, physiological characteristic parameters of the target person are acquired, the physiological characteristic parameters comprise gender, age, eyesight and reaction speed, the historical assembly data and the physiological characteristic parameters are input into a preset personal skill model to obtain the assembly skill level of the target person, the target person with the assembly skill level lower than a preset level threshold is determined to be an important monitoring object, and video acquisition frequency and action recognition accuracy of the important monitoring object are improved.

The input data of the personal skill model includes historical assembly data (e.g., assembly duration, assembly yield, and adverse causes of the plurality of products) and physiological characteristic parameters (e.g., gender, age, vision, and reaction rate) of the target person. Through analysis of these data, the model can output the assembly skill level of the target person. The model will mark people with assembly skill levels below a preset threshold as important monitoring objects. The model fully considers the historical performance and personal characteristics of the staff, and can judge the actual operation capability of the staff objectively and dynamically through training of a machine learning algorithm. The criteria for training are such that the model can accurately identify high risk personnel with weak assembly skills.

Specifically, the predictive early warning system may further collect and analyze historical work data and personal physiological characteristics of the assembly personnel to more fully evaluate the assembly skill level and quality risk level thereof. Firstly, the system can extract the historical assembly data of staff from production management systems such as MES, ERP and the like, and mainly comprises the number of products involved in assembly, assembly time of each product, assembly once qualification rate, main reasons of defective products and the like. Through statistical analysis of historical data, the system can preliminarily judge the assembly efficiency and quality stability of the staff.

For example, if the data shows that a staff member needs 30 minutes to assemble a product on average and the once-through assembly percent of pass is 98%, but the product is scrapped due to misoperation occasionally, the staff member can be judged to be more skilled, but the detail control needs to be enhanced. Second, the system may also collect key physiological characteristic parameters of the employee, such as gender, age, vision, reaction rate, etc., as factors for assessing the skill of their assembly. The eyesight can be obtained through annual physical examination reports of staff, and the response speed can be tested by the system design man-machine interaction game. By comparing the assembly data differences of staff under different physiological conditions, the influence rule of human factors on the assembly quality can be studied in depth. After the two types of data are acquired, the system inputs the two types of data into a personal skill assessment model constructed based on a machine learning algorithm to obtain the comprehensive skill level score of the staff. The model fully considers the historical performance and personal characteristics of the staff, and can objectively and dynamically judge the actual operation capability of the staff. When the model judges that the assembly skill level of a certain staff is lower than a preset threshold, the system marks the staff as a key monitoring object, and the video acquisition frequency and the motion recognition precision of the staff assembly process are improved later, so that the assembly quality risk of the staff can be early warned more rapidly and accurately. At the same time, the system will push the skill assessment report of the employee to the workshop manager, suggesting that it enhance skill training and job guidance for the employee.

In some embodiments, the predictive early warning system acquires facial feature images and action feature images of the target person through the camera, evaluates the vision level, concentration and fatigue of the target person according to the facial feature images and the action feature images, determines the hand flexibility of the target person according to the action feature images, and integrates the vision level, concentration, fatigue and hand flexibility of the target person to obtain physiological feature parameters.

Specifically, the predictive early warning system also utilizes advanced computer vision technology to evaluate the physiological state of the staff in real time through image analysis of the face and action characteristics of the staff so as to obtain the human factor data more dynamically and objectively. Specifically, the system collects facial close-up images and whole-body action images of staff in real time through high-definition cameras on an assembly line, and performs feature extraction and analysis by using algorithms such as face recognition and human body posture estimation. First, for facial feature images, the system highlights the eye area of the employee, and evaluates his vision level by means of the pupil diameter, blink frequency, eye movement trajectory, etc. For example, if an employee's pupil diameter is detected in a dilated state for a long period of time and blink frequency is less than 15 times per minute, it may be indicated that his vision is impaired or that a tired state is apparent. Meanwhile, the system can analyze the concentration degree of the facial expressions of the staff, and judges whether the staff is fully concentrated or not through indexes such as eyebrow positions, mouth angle radian, facial muscle tension and the like. For example, when an employee's eyebrow is locked, and gaze is stagnant, the system may determine that his attention is not focused. The overall fatigue state of the face of the employee can be estimated from indexes such as the size of the dark eye circles, the degree of slackening of the eye bags, and the darkness of the skin. The hand flexibility of staff is mainly extracted through action feature images, and indexes such as the bending and stretching amplitude of fingers, the rotation angle of wrists and the consistency of hand actions are mainly analyzed by the system. For example, when the stiff and curved fingers of the employee are detected, the employee can be determined to have poor flexibility. By comprehensively analyzing the evaluation results of the face and the action characteristics of the staff, the system can dynamically judge the physiological state of the staff, and when the staff is found to have poor eyesight, poor attention, poor hand flexibility or higher fatigue degree, corresponding early warning prompts are generated and the manager is recommended to pay attention. Meanwhile, the system also carries out correlation analysis on the physiological state evaluation result of the staff and the historical working data thereof, and is used for dynamically updating the staff skill level model so as to enable the staff skill level model to be closer to the real-time state of the staff. The human-computer collaborative multidimensional employee state perception is adopted, so that the predictive early warning system can more finely control the quality risk of the human factor number, and the dynamic optimization of the assembly process management is realized.

S202, comparing the key action features with a preset standard action template to obtain action difference degree.

Referring to step S102, the predictive early warning system determines the degree of action difference.

S203, establishing an applicability model of the assembly tool according to the tool type, the performance parameters and the use state information of the assembly tool used by the target personnel.

Referring to step S103, the predictive early warning system establishes an applicability model of the assembly tool.

In some embodiments, the predictive early warning system obtains tool types of assembly tools used by target personnel, the tool types comprise electric tools and manual tools, performance parameters of the assembly tools are obtained, the performance parameters comprise rated power, rotating speed, torque, material hardness and dimension specifications, service state information of the assembly tools is obtained, the service state information comprises actual service time length, maintenance record and fault frequency of the tools, the complexity level value of the assembly tools is divided according to the tool types, the application range of the assembly tools is determined according to the performance parameters, the reliability level value of the assembly tools is determined according to the service state information, and the three-dimensional applicability assessment model of the assembly tools is built by integrating the complexity level value, the application range and the reliability level value of the assembly tools to serve as an applicability model.

Specifically, the predictive early warning system further analyzes various attribute parameters of tools used by assembly staff in depth, and builds a tool applicability evaluation model for judging whether the staff can operate the tools in the hands skillfully and correctly so as to further predict assembly quality risks. Firstly, the system can automatically acquire the type information of the handheld tool through RFID tags worn by staff or visual identification technology, and the type information is mainly divided into two major types of electric tools (such as an electric drill, an electric wrench and the like) and manual tools (such as a screwdriver, a socket wrench and the like). In general, power tools are considered complex tools because of their high complexity of operation and high risk levels, while hand tools are relatively simple. Meanwhile, the system can collect key performance parameters of various assembly tools, such as rated power, idle rotation speed, maximum torque and the like of the electric tool, and material hardness, size specification and the like of the manual tool. These parameters determine the difficulty and accuracy requirements of the assembly task that the tool is capable of, and corresponding requirements are placed on the skill level of the staff.

For example, an electric wrench with higher torque is more suitable for assembling large-scale heavy components, and has high requirements on force control and stability of staff, while a small-sized straight screwdriver is more suitable for assembling precise electronic components, and has higher requirements on hand-eye coordination capability of staff. In addition, the system dynamically tracks the use state data of each tool, including accumulated use time, maintenance records, failure times and the like, which reflect the reliability of the tool in the assembly task to a certain extent.

When the failure frequency of a tool is high, it may cause an interruption in the assembly task or quality abnormality. Based on the tool data, the system constructs a three-dimensional tool applicability assessment model. The model starts from three dimensions of tool complexity, tool application range and tool reliability, the weight of each index is determined by using a hierarchical analysis method, and finally the applicability comprehensive score of each tool is calculated.

When an employee holds a certain tool to perform assembly operation, the system can compare the action characteristics of the tool with the applicability scores of the tool in real time, and if the action characteristics are not matched with the applicability scores of the tool, quality early warning can be generated and risks can be prompted. For example, when a person engages in the assembly of a small screw with a conventional high torque electric wrench, the system immediately warns the person that there may be an improper force control problem. For another example, when an employee is assigned to a new assembly station, in the face of some unused tools, the system may prompt the employee to operate carefully, avoiding quality accidents due to tool discomfort. By constructing the tool applicability evaluation model, the prediction and early warning system can consider the matching degree of staff and tools as an important influence factor of assembly quality, and accordingly, staff skill training and tool management optimization can be developed in a targeted manner, and finally, the deep fusion of the staff, tools and processes is realized, so that the overall quality level of assembly operation is improved.

S204, inputting the key action features into the applicability model to obtain tool application scores of target personnel.

Referring to step S104, the predictive early warning system determines a tool deployment score.

S205, acquiring real-time pressure data and real-time temperature data of an assembly environment in the motor assembly process, and determining motor assembly scores of target personnel according to action difference degrees, the real-time pressure data, the real-time temperature data and tool application scores.

Referring to step S105, the predictive early warning system determines a motor assembly score.

In some embodiments, the predictive early warning system detects real-time pressure data of the assembly environment in real time through a pressure sensor and compares the real-time pressure data with a preset standard pressure range to obtain a pressure difference degree, detects real-time temperature data of the assembly environment in real time through a temperature sensor and compares the real-time temperature data with the preset standard temperature range to obtain a temperature difference degree, and inputs the action difference degree, the pressure difference degree, the temperature difference degree and a tool application score into a preset motor quality assessment model to obtain a motor assembly score of a target person.

The input data of the motor quality assessment model comprises action variability of assembly staff, tool application scores, real-time pressure and temperature data of an assembly environment. The model adopts a weighted scoring method, gives different weight coefficients to each index, and finally outputs the comprehensive quality score of employee motor assembly. The higher the score, the greater the probability that the employee assembled a good in the current situation. The model adopts a decision tree algorithm, and through the training of a large amount of historical assembly data, the training standard is that the model can accurately predict the influence of each factor combination on the motor quality.

Specifically, the predictive early warning system also senses pressure and temperature parameters of the assembly environment in real time, and takes the pressure and temperature parameters into consideration as key environmental factors affecting the assembly quality of the motor. Firstly, the system collects the air pressure data of the local environment in real time through high-precision pressure sensors arranged around the assembly station, and compares the air pressure data with a preset standard pressure range (such as 1 atmosphere plus or minus 0.1 atmosphere), so as to calculate the pressure difference degree. When the ambient pressure is too high or too low, it may affect the normal performance of certain assembly processes, such as glue curing, vacuum adsorption, etc. For example, in assembling highly sealed electronic devices, if the ambient pressure fluctuates too much, internal components may become wetted and fail. Meanwhile, the system can monitor the temperature distribution condition of the assembly room in real time through a distributed temperature sensor network, and particularly, the system is used for a plurality of local areas sensitive to temperature. When the actual temperature exceeds a preset standard range (such as 25+/-5 ℃), the system calculates the temperature difference degree and pre-warns the quality risk possibly caused.

For example, during assembly of a servo motor, if the ambient temperature of the stator winding area is too high, the winding insulation material may age or break down, resulting in reduced motor performance. Therefore, monitoring the assembly environment temperature in real time is critical to quality control of certain precision assembly processes. After the pressure difference and the temperature difference are obtained, the system inputs the pressure difference and the temperature difference, the action difference of staff, the tool application score and other factors into a pre-trained motor quality evaluation model. The model adopts a decision tree algorithm, and can accurately predict the influence of various element combinations on the motor quality through training of a large amount of historical assembly data. The output result of the model is the motor assembly score of the staff, and the higher the score is, the greater the probability that the staff assembles a qualified product under the current environmental condition is. Once the score is below a preset threshold, the system will send a quality warning to related staff and management personnel to prompt the assembly environment to have risk of pressure or temperature abnormality, and suggest taking measures to adjust, such as turning on a dehumidification device, adjusting the temperature of an air conditioner, etc. Meanwhile, the system can dynamically adjust the quality detection frequency and detection items of the affected working procedure so as to discover and isolate the problem products in time and avoid unqualified products from flowing into the next working procedure. By bringing the assembly environment into the motor quality evaluation category, the prediction and early warning system can monitor various factors influencing the assembly quality more comprehensively and dynamically, so that high-quality products can be stably output under complex and changeable workshop environments.

S206, when the motor assembly score is lower than a preset qualification threshold, sending early warning information to the supervision client.

Referring to step S106, the predictive early warning system sends early warning information to the supervision client when the motor assembly score is lower than a preset qualification threshold.

S207, pushing assembly action video clips of target personnel in the early warning information, and determining feedback comments returned by the supervision client after receiving the early warning information.

When the predictive early warning system judges that the motor assembly score of a certain assembly staff is lower than a preset threshold value, the early warning message pushing is automatically triggered. Meanwhile, the system can also intelligently extract the key video clips in the process of assembling the current motor by the staff and send the key video clips and the early warning message to the client equipment of the workshop supervisory personnel. These video clips are typically 10-30 seconds long, and focus on showing details of the quality problems that may be easily caused by irregular actions, improper use of tools, etc. that exist in the staff assembly operations. Through the video picture, the supervision personnel can intuitively see the triggering reason of the early warning, quickly judge the quality risk level of the motor, judge whether the staff operation needs correction or not, and the like. Meanwhile, the system also supports a supervisory person to check real-time video monitoring of the assembly station at any time through the client, and can control fast forward, playback, pause and the like of the video. After receiving the early warning information, the supervisory personnel need to go to an assembly site for checking, and feedback treatment comments on the client side, and the method mainly comprises the options of reworking the motor, scrapping the motor, temporarily placing the motor, qualified motor quality, continuing assembly and the like. The system can automatically collect and record feedback comments of the supervisory personnel for subsequent responsibility tracing and performance assessment.

For example, when the system finds that a worker has incorrect assembly angles of magnetic steel for a plurality of times when assembling the rotor of the stepping motor, the system sends early warning information to the production main pipe and the quality main pipe, and attaches video clips of hand action details when the worker assembles the magnetic steel. After the two hosts receiving the early warning look up the video, the two hosts go to the assembly line to check at present, finally judge that the rotor needs to be reworked, and criticize and educate the staff.

And S208, if the feedback opinion is that the motor needs to be reworked or scrapped, generating an abnormal motor report for recording the reasons, responsible persons and lost man-hours of the abnormality.

When the supervisory personnel selects 'motor reworking required' or 'motor scrapping required' from feedback comments of the early warning information, the prediction early warning system automatically generates an abnormal motor report. The report is in the form of a file and is used for recording the contents of reasons for causing abnormal motor quality, information of responsible persons such as assembly staff, lost working hours caused by reworking or scrapping and the like. The abnormal reasons mainly refer to the problems of the system which are prompted in the early warning information, such as irregular assembly actions, improper use of tools, interference of an assembly environment and the like, and meanwhile, other problems found by a supervisor during on-site inspection can be supplemented. The responsible person information includes the names, job numbers, belonging teams, etc. of the relevant assembly staff, the names and the job titles of the supervisory staff such as workshop supervisors, quality supervisors, etc. The lost working hours mainly refer to the time cost of reworking or scrapping treatment, and the factors such as the number of reworking or scrapping motors, the assembly time of single-piece motors, the reworking or scrapping treatment time and the like need to be comprehensively considered. The system automatically estimates the lost man-hour according to a preset empirical formula and converts the lost man-hour into economic cost.

For example, according to the abnormal motor report generated by the system, it can be seen that the photoelectric encoder of a certain servo motor needs to be reworked due to improper operation of assembly staff, and the reworking needs about 50 minutes through preliminary accounting, which is equivalent to 0.83 man-hour loss, and the economic loss is about 150 yuan. The processing process and the result of each quality early warning event can be recorded in detail through the generation of the abnormal motor report, so that the follow-up responsibility tracing is facilitated, and data reference can be provided for management works such as performance assessment and cost accounting.

S209, according to the accumulated records of the abnormal motor reports, carrying out statistical analysis on the assembly quality of a plurality of production links within a preset time period to obtain quality improvement suggestions.

The predictive early warning system continuously tracks and records each abnormal motor event together, and periodically performs statistical analysis on the data to form an assembly quality improvement report. Reports are usually formed once every week or every month, and quality early warning occurrence conditions of all production links in the time period are mainly analyzed, wherein the quality early warning occurrence conditions comprise the number and frequency of early warning events, related assembly procedures, assembly staff, main reasons, loss degree and the like of early warning. After comparing and analyzing the data of different production links, the system utilizes an expert experience knowledge base to output a targeted suggestion for improving the assembly quality by combining the weight distribution of the early warning reasons. These suggestions may relate to various aspects of optimizing assembly process flows, improving assembly fixtures, enhancing staff skill training, improving shop management regimes, and the like.

For example, the monthly quality improvement report shows that the stator winding process quality of the brushless motor assembly line is high in early warning times, mainly because the assembly staff has insufficient skills, long operation time, and physical fatigue resulting in movement deformation. In order to solve the problem, the system suggests optimizing an assembly procedure of a winding procedure, reducing repetitive labor of workers, improving a winding tool, improving automation degree, increasing personnel skill training and frequency of on-duty, improving working hour quota, rewarding and punishment system and the like. Through the system excavation of the causes of quality problems of all links, the prediction early warning system can output quality improvement measures with strong operability, and the power-assisted assembly workshop realizes the refinement and the flow of quality management.

In some embodiments, the predictive early warning system can perform associated mining on abnormal motor reports, production task lists, process parameters and equipment states to construct a digital twin model of the whole assembly production process, simulate and deduct assembly processes of a plurality of production links through the digital twin model in a preset time period to determine quality risks and corresponding influence degrees of the plurality of production links, and determine optimized production process parameters and process control schemes according to the quality risks to obtain quality improvement suggestions.

The digital twin model constructs the digital mapping of the assembly workshop by virtual simulation, 3D modeling and other technologies, and comprises the elements of personnel, equipment, materials, environment and the like. The input data of the model includes abnormal motor reports, production task orders, process parameter settings, and equipment status data. The quality risk level and the influence degree of the whole production links can be output through the simulation deduction of the model, various optimization schemes are simulated and evaluated, and finally operable quality improvement suggestions are output. The training of the model follows rules in the expert experience knowledge base, the criteria being that it enables accurate prognosis of quality risk through simulation and provides an effective countermeasure.

Specifically, the predictive early warning system can realize panoramic quality management and predictive optimization of the production process by constructing a digital twin model of the whole assembly production process. Firstly, the system adopts a big data mining technology to carry out association analysis on abnormal motor reports, production task lists, process parameter settings and equipment state data in various production management systems, and key factors affecting production quality and interaction rules thereof are mined. On the basis, the system utilizes the technologies of virtual simulation, 3D modeling and the like to construct a digital twin model of an assembly workshop, and the digital twin model comprises digital mapping of various elements such as personnel, equipment, materials, environment and the like. Through the model, all links of assembly production can be synchronized and restored to a high degree in real time in a virtual space. Based on the digital twin model, the system can carry out deduction simulation on the workshop production process within a preset time period (such as 1 week in the future). By setting different production tasks and technological parameter combination schemes, the system can comprehensively simulate the production process under each scheme and evaluate the quality risk level of each production link and the influence degree of the whole production. For example, when the production task amount of a certain batch of products is large and the process difficulty is high, the system simulation discovers that the links of assembly, debugging, detection and the like all face large quality risks, wherein the risks of increasing the reject ratio caused by insufficient skill of staff in the assembly link are most prominent, and delay delivery of the production task may be caused. By summarizing quality risk assessment results of all links, the system can identify bottleneck short plates which restrict production quality improvement. On the basis, the system can automatically search the coping schemes of quality risks, such as optimizing production task sequencing, adjusting process parameters, increasing quality detection points and the like, verify the improvement effect of each scheme through a simulation model, and finally screen out the optimal process control strategy combination to form an operable quality improvement suggestion for managers to refer to decisions.

For example, aiming at the skill risk of the assembly link, the system recommends measures such as carrying out staff skill training in advance, reasonably allocating task saturation, increasing first-part detection frequency and the like, and through simulation verification, the staff reject ratio can be reduced by 2 percentage points, and the probability of on-time delivery of the production task can be improved to more than 95%. Through simulation deduction and optimization decision of digital twin driving, the predictive early warning system can convert passive post quality management into active pre risk prediction and process optimization, continuous improvement of production organization modes is realized while the product quality is ensured, and finally double improvement of quality and efficiency is brought. In the future, the digital twin technology is expected to become an important means for quality management of assembly production in the intelligent manufacturing era.

S210, generating a work adjustment scheme comprising station layout and employee skill requirements.

On the basis of generating the assembly quality improvement suggestion, the predictive early warning system can be further designed to form an employee work adjustment scheme. The scheme is mainly based on motor assembly scores and abnormal records of staff, and the current station of the staff is reevaluated and adjusted to match the actual operation skill level of the staff. Meanwhile, the scheme can also provide targeted training requirements for the skill shortboards of staff, and promote the improvement of comprehensive capacity of the staff. When the adjustment scheme is formulated, the system mainly considers the following factors of learning ability and skill improvement potential of staff, productivity balance and beat balance requirements of an assembly line, rationalization suggestion and adjustment willingness of staff and the like. Based on the post adjustment rules in the expert knowledge base, the system utilizes a Hungary algorithm, a genetic algorithm and other combined optimization methods to perform optimal matching between staff and stations, so as to form a set of feasible work adjustment scheme. For example, according to the adjustment scheme generated by the system, staff A who originally manually assemble parts frequently has quality abnormality due to insufficient action flexibility, and recommends that the staff A transfers to a product packaging station, and staff B obviously improves the assembly quality and efficiency after training for several rounds, recommends that the staff A is an assembly team leader and is responsible for teaching new staff. Meanwhile, the adjustment scheme generated by the system also provides decision reference for management staff, so that the management staff gets rid of experience and random personnel management, and really performs teaching and sentry matching.

In the embodiment of the application, as the multidimensional quality prediction model integrating historical assembly data, employee physiological characteristics, assembly tool attributes and environmental parameters is introduced, and the digital twin technology is used for carrying out simulation deduction and risk assessment on the whole production process, a set of intelligent quality management system for assembly production integrating data acquisition, risk early warning and optimization decision is constructed, the problems of delayed preventive measures, incomplete analysis of influencing factors, lack of quantitative verification of an improvement scheme and the like in the traditional quality management mode are effectively solved, the transition from passive detection to active prevention is realized, the quality management of assembly is advanced to each link of the production process, the quality problem is prevented, the quality level and the operation efficiency of assembly production are obviously improved, and powerful support is provided for lean production and agile management in the intelligent manufacturing era.

The following describes a prediction early warning system in the embodiment of the present application from the perspective of hardware processing, please refer to fig. 3, which is a schematic diagram of a physical device structure of the prediction early warning system in the embodiment of the present application.

It should be noted that, the structure of the predictive early warning system shown in fig. 3 is only an example, and should not bring any limitation to the functions and the application scope of the embodiment of the present invention.

As shown in fig. 3, the predictive early warning system includes a central processing unit (Central Processing Unit, CPU) 301 that can perform various appropriate actions and processes, such as performing the methods described in the above embodiments, according to a program stored in a Read-Only Memory (ROM) 302 or a program loaded from a storage section 308 into a random access Memory (Random Access Memory, RAM) 303. In the RAM 303, various programs and data required for the system operation are also stored. The CPU 301, ROM 302, and RAM 303 are connected to each other through a bus 304. An Input/Output (I/O) interface 305 is also connected to bus 304.

Connected to the I/O interface 305 are an input section 306 including an audio input device, a push button switch, and the like, an output section 307 including a Liquid crystal display (Liquid CRYSTAL DISPLAY, LCD) and an audio output device, an indicator lamp, and the like, a storage section 308 including a hard disk, and the like, and a communication section 309 including a network interface card such as a LAN (Local Area Network) card, a modem, and the like. The communication section 309 performs communication processing via a network such as the internet. The drive 310 is also connected to the I/O interface 305 as needed. A removable medium 311 such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, or the like is installed on the drive 310 as needed, so that a computer program read therefrom is installed into the storage section 308 as needed.

In particular, according to embodiments of the present invention, the processes described above with reference to flowcharts may be implemented as computer software programs. For example, embodiments of the present invention include a computer program product comprising a computer program embodied on a computer readable medium, the computer program comprising a computer program for performing the method shown in the flowchart. In such an embodiment, the computer program may be downloaded and installed from a network via the communication portion 309, and/or installed from the removable medium 311. When the computer program is executed by a Central Processing Unit (CPU) 301, various functions defined in the present invention are performed.

Specific examples of a computer-readable storage medium include, but are not limited to, an electrical connection having one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-Only Memory (ROM), an erasable programmable read-Only Memory (Erasable Programmable Read Only Memory, EPROM), a flash Memory, an optical fiber, a portable compact disc read-Only Memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In the context of this document, a computer readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device.

The flowcharts and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to various embodiments of the present invention. Where each block in the flowchart or block diagrams may represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that, in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures.

Specifically, the prediction and early warning system of the embodiment includes a processor and a memory, and when the computer program is executed by the processor, the method for predicting and early warning defective products of the manually assembled motor provided by the embodiment is implemented.

In another aspect, the present invention also provides a computer readable storage medium, which may be included in the predictive early warning system described in the above embodiment, or may exist alone without being assembled into the predictive early warning system. The storage medium carries one or more computer programs which, when executed by a processor of the predictive early warning system, cause the predictive early warning system to implement the defective product predictive early warning method for a manually assembled motor provided in the above embodiment.

While the application has been described in detail with reference to the foregoing embodiments, it will be understood by those skilled in the art that the foregoing embodiments may be modified or equivalents may be substituted for some of the features thereof, and that the modifications or substitutions do not depart from the spirit of the embodiments.

As used in the above embodiments, the term "when..is interpreted as meaning" if..or "after..or" in response to determining..or "in response to detecting..is" depending on the context. Similarly, the phrase "when determining..or" if (a stated condition or event) is detected "may be interpreted to mean" if determined.+ -. "or" in response to determining.+ -. "or" when (a stated condition or event) is detected "or" in response to (a stated condition or event) "depending on the context.

Those of ordinary skill in the art will appreciate that implementing all or part of the above-described method embodiments may be accomplished by a computer program to instruct related hardware, the program may be stored in a computer readable storage medium, and the program may include the above-described method embodiments when executed. The storage medium includes a ROM or a random access memory RAM, a magnetic disk or an optical disk, and other various media capable of storing program codes.

Claims

1. A method for predicting and warning defective products of manually assembled motors, characterized in that it is applied to a prediction and warning system, and the method comprises:

Acquire a real-time video stream during the motor assembly process, and determine key motion features of the target person based on the real-time video stream; the key motion features include the hand motion trajectory, strength, speed and frequency of the target person;

Comparing the key action features with a preset standard action template to obtain the action difference;

Establishing a suitability model of the assembly tool according to the tool type, performance parameters and usage status information of the assembly tool used by the target personnel;

Inputting the key action features into the applicability model to obtain the tool application score of the target person;

Acquire real-time pressure data and real-time temperature data of the assembly environment during the motor assembly process, and determine the motor assembly score of the target person according to the action difference, the real-time pressure data, the real-time temperature data and the tool application score;

When the motor assembly score is lower than a preset qualified threshold, a warning message is sent to the supervision client.

2. The method according to claim 1 is characterized in that the establishment of the applicability model of the assembly tool according to the tool type, performance parameters and usage status information of the assembly tool used by the target personnel specifically comprises:

Obtaining the tool type of the assembly tool used by the target person; the tool type includes power tools and hand tools;

Obtaining performance parameters of the assembly tool; the performance parameters include rated power, rotation speed, torque, material hardness and size specifications;

Acquiring usage status information of the assembly tool; the usage status information includes actual usage time, maintenance records and failure frequency of the tool;

Classifying the complexity values of the assembly tools according to the tool types, determining the applicable scope of the assembly tools according to the performance parameters, and determining the reliability values of the assembly tools according to the usage status information;

The complexity value, application scope and reliability value of the assembly tool are comprehensively considered to establish a three-dimensional applicability evaluation model of the assembly tool as a applicability model.

3. The method according to claim 1, characterized in that the step of acquiring real-time pressure data and real-time temperature data of the assembly environment during the motor assembly process, and determining the motor assembly score of the target personnel according to the action difference, the real-time pressure data, the real-time temperature data and the tool application score, specifically comprises:

The real-time pressure data of the assembly environment is detected by a pressure sensor in real time, and the real-time pressure data is compared with a preset standard pressure range to obtain a pressure difference;

The real-time temperature data of the assembly environment is detected by a temperature sensor in real time, and the real-time temperature data is compared with a preset standard temperature range to obtain a temperature difference;

The action difference, the pressure difference, the temperature difference and the tool application score are input into a preset motor quality assessment model to obtain the motor assembly score of the target person.

4. The method according to claim 1, characterized in that before the step of acquiring a real-time video stream during the motor assembly process and determining key action features of the target person according to the real-time video stream, the method further comprises:

Obtaining historical assembly data of a target person; the historical assembly data includes assembly time, assembly qualification rate and defect reasons of multiple products;

Acquiring physiological characteristic parameters of the target person; the physiological characteristic parameters include gender, age, vision and reaction speed;

Inputting the historical assembly data and the physiological characteristic parameters into a preset personal skill model to obtain the assembly skill level of the target person;

The target personnel whose assembly skill level is lower than the preset level threshold are determined as key monitoring objects, and the video acquisition frequency and action recognition accuracy of the key monitoring objects are improved.

5. The method according to claim 4, characterized in that the step of obtaining the physiological characteristic parameters of the target person specifically comprises:

Acquire a facial feature image and a motion feature image of a target person through a camera, and evaluate the target person's vision level, concentration, and fatigue level based on the facial feature image and the motion feature image;

Determining the hand dexterity of the target person according to the action feature image;

The vision level, concentration, fatigue level and hand dexterity of the target person are integrated to obtain physiological characteristic parameters.

6. The method according to claim 1, characterized in that after the step of sending a warning message to a supervision client when the motor assembly score is lower than a preset qualified threshold, the method further comprises:

Pushing the assembly action video clip of the target person in the warning information, and determining the feedback returned by the supervision client after receiving the warning information;

If the feedback is that the motor needs to be reworked or scrapped, an abnormal motor report is generated that records the cause of the abnormality, the person responsible, and the lost working time;

Based on the accumulated records of the abnormal motor reports, statistical analysis is performed on the assembly quality of multiple production links within a preset time period to obtain quality improvement suggestions;

Generate work adjustment plans including workstation layout and employee skill requirements.

7. The method according to claim 6 is characterized in that, according to the accumulated records of the abnormal motor reports, statistical analysis is performed on the assembly quality of multiple production links within a preset time period to obtain quality improvement suggestions, which specifically includes:

Conduct correlation mining on abnormal motor reports, production task orders, process parameters and equipment status to build a digital twin model of the entire assembly production process;

Within a preset time period, the assembly process of multiple production links is simulated and deduced through the digital twin model to determine the quality risks of the multiple production links and the corresponding impact levels;

Based on the quality risks, the optimized production process parameters and process control solutions are determined to obtain quality improvement suggestions.

8. A prediction and warning system, characterized in that the prediction and warning system comprises: one or more processors and a memory; the memory is coupled to the one or more processors, the memory is used to store computer program code, the computer program code comprises computer instructions, and the one or more processors call the computer instructions so that the prediction and warning system executes the method as described in any one of claims 1-7.

9. A computer-readable storage medium, comprising instructions, characterized in that when the instructions are executed on a prediction and early warning system, the prediction and early warning system executes the method according to any one of claims 1 to 7.

10. A computer program product, characterized in that when the computer program product is run on a prediction and early warning system, the prediction and early warning system is enabled to execute the method according to any one of claims 1 to 7.