Classifications

• • G—PHYSICS
  • G01—MEASURING; TESTING
  • G01G—WEIGHING
  • G01G23/00—Auxiliary devices for weighing apparatus
  • G01G23/18—Indicating devices, e.g. for remote indication; Recording devices; Scales, e.g. graduated
  • G01G23/36—Indicating the weight by electrical means, e.g. using photoelectric cells
  • G01G23/37—Indicating the weight by electrical means, e.g. using photoelectric cells involving digital counting
  • G01G23/3728—Indicating the weight by electrical means, e.g. using photoelectric cells involving digital counting with wireless means
• • G—PHYSICS
  • G01—MEASURING; TESTING
  • G01G—WEIGHING
  • G01G19/00—Weighing apparatus or methods adapted for special purposes not provided for in the preceding groups
  • G01G19/08—Weighing apparatus or methods adapted for special purposes not provided for in the preceding groups for incorporation in vehicles
  • G01G19/12—Weighing apparatus or methods adapted for special purposes not provided for in the preceding groups for incorporation in vehicles having electrical weight-sensitive devices
• • G—PHYSICS
  • G06—COMPUTING; CALCULATING OR COUNTING
  • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
  • G06N3/00—Computing arrangements based on biological models
  • G06N3/02—Neural networks
  • G06N3/04—Architecture, e.g. interconnection topology
  • G06N3/045—Combinations of networks

Definitions

• the invention relates to a method according to the introductory part of patent claim 1 for measuring loads directed to structures.
• Structures here mean stationary structures such as buildings, bridges, rails (railroads), quays, stationary load beds or the like, and moving structures, such as vehicles, hoisting gear and the like.
• the invention also relates to a method according to the introductory part of patent claim 3 for measuring the weight of a vehicle.
• Vehicle here means a bed or space to be moved on wheels, runners or similar structures, on which bed or space for instance gods and/or people can be transported, and which can be moved by its own power source or hauled, separately or in a combination, by a suitable hauling apparatus.
• the vehicle can be a car, for instance a truck, a railway vehicle, such as a engine or a railway car; a sleigh; a vehicle provided with one or several track chains, or even an airplane (being moved on wheels on the airfield), just to mention some examples.
• a load bed such as a vehicle or a container, supported for instance by legs or wheels
• detectors advantageously strain gauge detectors
• the load weight is defined on the basis of the signals obtained from these detectors.
• the said detectors are installed in the solid frame of the load bed, in the vicinity of the fastening points of the supports; they are used for measuring deformations and/or strain caused in the frame by the load, and the load weight is defined on the basis of these measurements.
• a drawback with the above described method and apparatus is that the installation of the detectors at the points of measurement in the vicinity of the fastening points of support is mainly carried out experimentally, and hence it may be difficult to find the most suitable points of measurement in various load bed constructions.
• the problem with the latter method is that the points of measurement in the load bed are defined by a mathematical process which requires modeling of the load bed. This delays the application of the method, particularly in structures which have not been modeled before.
• Another drawback is the fact that the calibration of the detectors attached to the points of measurement is carried out straightforwardly by defining the correction factors on the basis of the loads of other points of support affecting each point. This arrangement becomes problematic, particularly when the number of the points of measurement grows and the measured load bed or other structure becomes complicated.
• the object of the invention is to eliminate at least part of the above mentioned drawbacks and to simplify the method of measurement.
• a particular object of the invention is to introduce a new and efficient method for defining the load in various structures, so that the numbers of points of measurement and/or complexities of construction in the structure do not cause insurmountable difficulties when measuring the loads directed thereto.
• the method of the invention for measuring the load of a vehicle is characterized by the features stated in the novelty part of the patent claim 3.
• the detectors measuring deformation in structures are attached to the frame parts of the structures, most advantageously in the vicinity of the support points.
• the measurement signals obtained from the detectors are processed by means of a predetermined neural network, so that from the output level of the network, there are obtained the loads directed to desired points of the structures, and that the neural network is in advance trained with test loads to process the measurement signals of the said measuring detectors.
• - predetermined loads with known weights are arranged at the support points of the structures; - the measurement signals obtained from the measuring detectors are preprocessed and considered as input signals from the input units of the input layer of the predetermined neural network;
• the load directed to one or several points of support is defined by means of the neural network and the predetermined weights.
• An advantage of the method of the invention is that it can be applied to many types of structures. These structures can be stationary, such as buildings, bridges, load beds and loading ramps, or movable, such as vehicles, trucks, hoisting gear or the like.
• the measuring system of the invention is particularly suited to observing variation in loads and to controlling and/or defining the distribution of loads in the different parts of the structures.
• the method can be used for detecting load peaks and for alarming when a preset load limit is surpassed, as a total load value and or with respect to chosen structures.
• Yet another advantage of the invention is that the number of measuring points and/or desired measurement signals does not have to be limited. By applying the method of the invention, there can be processed a large number of measurement signals, on the condition that the measuring system is provided with a sufficient data processing capacity, particularly calculatory capacity.
• the most advantageous application of the invention relates to a method for measuring the load and/or load distribution of a vehicle.
• the detectors measuring the deformation of structures are attached to the load-bearing parts of the vehicle, particularly to the frame parts, most advantageously in the vicinity of the support points.
• the measurement signals obtained from the measuring detectors are processed by means of a predetermined neural network, so that from the output layer of the neural network there are obtained weight loads directed to desired spots of the vehicles, particularly to one or several points of support, and that the neural network is in advance trained with test loads to process the measurement signals from the said detectors.
• the measuring detectors can be attached to the frame beams of the vehicle, particularly a truck and/or a trailer, in the vicinity of the points of suspension or points of support of the wheels, to which support points the reaction forces are directed.
• the locations of the measuring points/detectors do not have to be set at the theoretically exact optimum points, but it suffices that the gauges are located in the correct area.
• the correct area is dependent on the structure, and it is generally the area surrounding the support points of the structure; in the case of a truck, for instance, it is the area surrounding the wheels and/or wheel bogies.
• Figure 1 is a schematic illustration of the beam structure
• Figure 2 illustrates one junction A of the beams in the beam structure of Figure 1 as well as connected measuring detectors and the processing system of the measurement signals in block diagram form;
• Figure 3 illustrates a three-layered neural network for calculating the loads
• Figure 4 illustrates a processing unit, i.e. a neuron, of the neural network
• Figures 5 and 6 illustrate in flowchart form the training process of the neural network, the said neural network being suited to the load measuring system;
• Figure 7 illustrates in flowchart form the measuring of the load carried out by means of the neural network
• Figure 8 is a schematic top-view illustration of a vehicle combination where in the frame of the tractor and the trailer, there are arranged detectors in order to measure the weight and/or weight distribution, and the obtained measurement signals are processed by a suitable data processing unit according to the method of the invention;
• Figure 9 is a top-view illustration of a detail of the vehicle combination of Figure 8, i.e djped the fastening of the wheel to the frame structure of the vehicle and the installation of the measuring detectors in connection with the frame part;
• Figure 10 is a side-view illustration of a detail of the vehicle combination of Figure 8 (cf. Figure 9);
• FIG 11 illustrates how the measuring method of the invention is applied to the weight measuring system of the vehicle of Figure 8.
• Figure 1 illustrates a lattice-like steel beam structure 1, which forms for instance the frame of a building or part of the frame.
• the horizontal beams 2 of the steel beam structure 1 form one floor level or the like.
• the horizontal beams 2 are supported against the foundation 4 and/or to the preceding floor level by vertical beams 3, 3'.
• junctions of the horizontal and vertical beams 2, 3, 3' constitute the locations or support points, through which the weight load, among others, is directed to the various parts of the structure and is shifted in the structures.
• the horizontal beams 2; 2 ⁇ , 2 ⁇ , 2 ⁇ are suitably interconnected at the vertical beams 3, 3'.
• the beams 2, 3, 3' are I-beams.
• the measuring detectors 6; 6 1 , 6 , 6 3 , 6 4 , 6 5 and ⁇ 6 are advantageously realized of two or more strain gauge elements arranged crosswise or at an angle with respect to each other.
• the locations of the points of measurement are defined for example by means of the known FEM (Finite Element Method), and the measuring detectors 6 are installed in an area which is suitable with respect to the theoretically calculated optimal point of measurement.
• the points of measurement are placed in the vicinity of the support points especially because on the basis of the difference of the shearing strengths affective on the different sides of the support points, it is possible to determine the support reactions, such as weight loads directed to the support points.
• the detectors 6 there are measured deformations, such as bending and/or strain and compression in structures, in this case in the horizontal beams 2.
• measuring detectors 6; 6 ⁇ , 6& In connection with the vertical beams 3, 3' of the steel beam structure 1 of Figure 1, in this embodiment in the web parts 5; 5 4 , there can also be provided measuring detectors 6; 6 ⁇ , 6&. In addition to the above described detectors 6; 6 , 6 2 , 6 ⁇ , 6 ⁇ , 6 ⁇ , 6 ⁇ , they can also be installed at each desired junction A; Al, A2, A3,... or alternatingly with respect to these. In the latter case, at the junction there is measured the compression directed to this support point , i.e. only vertical forces F, whereas in the former and more general case, all strains directed to the structures can be measured.
• the preprocessing unit 7 the measurement signal obtained from the measuring detectors 6 is processed, for instance amplified and stabilized, in order to be suitable for the input interfaces of the successive processing units.
• the measurement signals are fed to the neural network unit 8 for the processing proper of the measurement signals.
• the neural network unit 8 advantageously constitutes a data processing unit including one or several microprocessors .
• the measurement signals are processed in the neural network unit 8 by means of a recorded neural network program, so that from the output layer of the neural network there are obtained as results the loads directed to desired points in the structure 1, for instance to one or several support points A; Al, A2, A3, ....
• the neural network unit 8, i.e. the neural network can be considered to be composed of separate but interconnected calculatory or processing units.
• the neural network is trained in advance with test loads to process the measurement signals of the said measuring detectors.
• Figure 3 illustrates a three-layered neural network , which is a so-called perceptron network.
• Figure 4 illustrates a processing unit, i.e. neuron, of this type of neural network.
• the employed neural network is a network of three or more layers, comprising an input layer 12, one or several hidden layers 13 and an output layer 14.
• the analogous measurement signals x; X , X ⁇ , X2, X3 obtained from the measuring detectors 6 are processed in the preprocessing unit 7, where they are normalized to the area [0, 1] and digitized.
• the thus formed normatized input signals are fed to the neural network unit 8.
• the normalized input signals are fed into the input units 12 , 12 ⁇ , 12 ⁇ , 12 ⁇ of the input layer of the neural network.
• the output signals VQ ⁇ , VJO, V2 obtained from these input units are formed by multiplying the normalized input signals by the first weights wjfcl , and the obtained products are summed in a predetermined fashion in the processing unit 131, 13 ⁇ of the hidden layer.
• the output signals Vn*, Vj l obtained from the processing units 131, 13 ⁇ of the hidden layer 13 are multiplied by the second weights wjj2 , and the obtained products are suitably summed in the processing units 141, 14 ⁇ , 14-3 0 f the output layer 14 in order to create the desired output signals y; yo, yi, Y2 • Thereafter the output signals y are rescaled back to load or weight data (denormalization), which represents for instance loads directed to predetermined points A; Al, A2, A3,... in the structure 1.
• the operation of the neural network can be illustrated by observing the neuron of Figure 4.
• the element of each input vector x is multiplied in the neuron by a weight WJ corresponding to the input signal.
• the input signal can be either an analogous or a digital (binary) signal.
• the obtained weighted input signals XJWJ are summed up, and the term is subtracted from the sum.
• the non-linear function f is called the activation function.
• the non- linearity employed in a perceptron network is a so-called hard limiter non-linearity, which is defined as follows:
• the output error caused by the desired output signal of the neural network and the real output value can be minimized for instance by the Widrow-Hoff algorithm, which also is called the LMS (Least Mean Square) algorithm.
• the output error of the neural network can also be minimized by applying the gradient method, where the penalty function is the sum of the squares of the errors (delta rule).
• the activation function must be continuously derivating. The most generally used continuous activation function is defined as follows:
• the function 4 is called the Sigmoid function.
• the neural network is trained with several test loads and test measurements to process the measurement signals of the said detectors. This is advantageously carried out as follows (cf. Figures 5 and 6).
• the weights and WJJ2 of the neural network are formatted with small random coefficients, which are normalized within the range [0, 1] i.e.:
• V m for the output layer of the neural network:
• the calculated output signal V m proper and the desired known output signal [ are compared, and there is calculated their separation function i, which describes the error or the real loading situation and the loading situation of the moment as calculated by the neural network:
• the desired output signals j are obtained from known test loads directed to the support points, such as from load K of the support point A in Figure 2.
• the new input signals i.e. measurement signals can be selected by arranging for instance new test loads at the points of measurement. Thereafter the above described procedure is repeated, and new weights are defined again.
• the difference of the new and old weights remains below the threshold value, i.e. within the predetermined limits, the weights needed in the measurements proper are defined and they are recorded, whereafter, when the test loads are removed, the neural network is tuned for the measurements proper.
• the points of support of the structure 1 are subjected to various static and dynamic loads, depending on the application in question.
• the neural network is used for observing changes in the load and locating fluctuations therein. Measuring signals are collected from the detectors 6 at suitable predetermined intervals, and the length of these intervals can be adjusted by means of the program when necessary.
• the filling up of the input layer of the neural network with measurement data is observed, and when the processing units of the input layer have received the measurement data and the input layer is thus filled, the calculation is started.
• the collecting of measurement data is advantageously interrupted for the duration of the calculation.
• the calculation is realized as follows (cf. Figure 7).
• the measurement signals, i.e. the input signals xk are fed into the input layer of the neural network, whereafter the output signals of the network are calculated by using the formula
• the next measurement signals can be read into the input layer of the neural network, and the above described calculation procedure can be repeated.
• the load F directed to one or several support points A is defined on the basis of the measurement signals obtained from the detectors 6 connected to the points of measurement and by utilizing the neural network and the predetermined weights.
• FIG 8 is a top-view schematic illustration of a vehicle combination , to which the measuring system of the invention is now applied, and of the measuring arrangement in block diagram form.
• the frames 17, 18 of the tractor 15 and trailer 16 of this combination comprise two parallel frame beams 17a, 17b; 18a, 18b, suitably spaced and interconnected with suitable transversal supports (not illustrated in the drawings).
• the tractor 15 and the trailer 16 are provided with a number of wheel bogie structures 19; 19 1 , 19 2 , 19 , 19 4 , 19 5 , 19 6 , 19 7 , of which in this case four pairs 19; 19l, 19 2 , 193, 19 4 are provided in the tractor and three pairs 19; 19 ⁇ , I96, 19 ⁇ in the trailer.
• the wheels 19a, 19b and the wheel bogie structure 19 is in this case attached, by means of leaf springs 20a, 20b, to the two frame beams 17a, 17b; 18a, 18b on both sides of the axis 21, at suitable support and fastening points 22a, 23a; 22b, 23b, as is illustrated in Figures 9 and 10.
• the weights of the loads resting on the tractor 15 and the trailer 16 are defined from the measurement signals obtained from the detectors on the basis of the deformations in the frame 17, 18.
• the locations of the points of measurement are defined on the basis of the known element method FEM (Finite Element Method).
• FEM Finite Element Method
• the element method there are defined the shearing strains directed to the frame beams 17a, 17b; 18a, 18b.
• the application of the element method to measuring vehicle loads, and particularly to locating points of measurement, is explained in more detail in the international patent application PCT/FI94/00115.
• the element method can be used for defining the theoretical locations of the points of measurement in relation to the frame. It is pointed out that the detectors can be located at suitable points in the vicinity of the theoretically calculated points of measurement, when the measuring method of the present invention is applied.
• the points of measurement are located in the vicinity of the support points especially because on the basis of the difference of the shearing strengths affective on different sides of the support points, it is possible to define the support reactions, i.e. weight loads directed to the support points. This is particularly emphasized when there are used continuous loads, in which case the shearing strength curves and their values vary in the lengthwise direction of the load bed, in this case of the frame beams 17a, 17b; 18a, 18b.
• the measuring detectors are attached to the calculated points of measurement, as was stated above.
• the detectors are strain gauge detectors.
• At each point of measurement there is attached at least one detector advantageously comprising two or more strain gauge detectors.
• At each point of measurement there is thus formed a suitable detector element group, with detector elements on the same level but at an angle with respect to each other.
• a detector is also attached to the vertical side of the frame beam 17a, 17b; 18a, 18b, at least roughly on the location of the point of measurement.
• two detectors on both sides of the frame beam are advantageously attached.
• the vehicle frame is always subjected to sideways directed strain, when the wheels are located at different heights, or when the vehicle is placed on inclined ground or when the load is arranged asymmetrically. Therefore several detectors are provided at each point of measurement, so that these secondary effects can be eliminated. Moreover, the electric coupling of the detectors can be realized so that the influence of temperature in the obtained measuring results is eliminated.
• the strain gauge detectors are attached so that to the points of measurement 24a, 25 a, 26a, 27a; 24b, 25b, 26b, 27b of the frame beams 17a; 18a and 17b; 18b, there respectively belong the detector groups 28a, 29a; 30a, 31a; 32a, 33a and 34a, 35a and 28b, 29b; 30b, 31b; 32b, 33b; 34b, 35b, which are thus located on the opposite sides of the vertical part of the frame beams.
• the detector groups are connected to the measurement signal preprocessing unit 36, where the signals are suitably processed for the neural network unit 37.
• the signal preprocessing unit 36 comprises for instance a Wheatstone bridge unit, whereto part of the detectors is coupled so that the influences of the deformations in the detectors are amplified in the output poles of the bridge.
• the difference signal affecting over the bridge is further coupled, via the amplifier provided in the preprocessing unit 36, via a suitable signal processing unit and multiplexer further to the neural network unit 37.
• the preprocessing units 36; 361, 36 2 , 363, ... are used for processing the output signals from all detectors provided in connection with the wheel bogies 19, and these signals are fed into the neural network unit 37, as is illustrated in Figure 8.
• the neural network unit 37 is located in the cabin of the vehicle or the like, where it is connected to suitable display equipment, means for feeding instructions, possible printers and connections to external facilities, such as communication means in order to transport information for instance wirelessly to an external computer or other data collecting unit.
• the measurement signals are processed in the neural network unit 37, by means of the neural network program recorded in the memory thereof. This processing is advantageously realized in similar fashion as was illustrated above, in connection with Figures 5 and 6.
• Figure 11 illustrates in principle the operation of a vehicle measurement system in block diagram form.
• the wheel loads of a vehicle combination 15, 16 are calculated by means of the neural network unit 37.
• First the neural network is trained with test loads and test weighings, by using real wheel weights as feedback (cf. Figures 5 and 6.)
• the real wheel weights are measured by means of wheel scales 38 located under each wheel 19a ( Figures 9, 10 and 11) of each bogie 19.
• the measurement data is collected from the detectors, from the points of measurement at each bogie 19.
• the employed input signals of the neural network i.e. the input vector x are the measurement signals obtained from the detectors, from around each point of support 22a, 23a; 22b, 23b , which signals are preprocessed in the respective preprocessing units 36 and then averaged.
• the internal operation of the neural network is described above, in connection with Figures 2 - 7 and formulas 1 - 10.
• This embodiment also introduces a three-layered neural network, including an input layer, a hidden layer and an output layer, as is illustrated in Figure 11.
• Figure 11 shows the reference layer, whereto the weight values obtained from the wheel scales 38 are thus fed, and with which the signals obtained from the detectors and processed by the neural network are compared, and whereby the suitable weights are finally calculated for the signals in between the layers of the neural network (cf. Figure 6).

Landscapes

• Physics & Mathematics (AREA)
• Engineering & Computer Science (AREA)
• General Physics & Mathematics (AREA)
• Theoretical Computer Science (AREA)
• General Health & Medical Sciences (AREA)
• General Engineering & Computer Science (AREA)
• Biophysics (AREA)
• Computational Linguistics (AREA)
• Data Mining & Analysis (AREA)
• Evolutionary Computation (AREA)
• Artificial Intelligence (AREA)
• Molecular Biology (AREA)
• Computing Systems (AREA)
• Biomedical Technology (AREA)
• Life Sciences & Earth Sciences (AREA)
• Mathematical Physics (AREA)
• Software Systems (AREA)
• Health & Medical Sciences (AREA)
• Computer Networks & Wireless Communication (AREA)
• Force Measurement Appropriate To Specific Purposes (AREA)
• Length Measuring Devices With Unspecified Measuring Means (AREA)

Priority Applications (3)

Applications Claiming Priority (2)

Publications (1)

Family

ID=8540291

Family Applications (1)

Country Status (5)

Cited By (10)

Families Citing this family (2)

Citations (5)

• 1994
  • 1994-03-10 FI FI941153A patent/FI94677C/fi not_active IP Right Cessation
• 1995
  • 1995-03-10 EP EP95911349A patent/EP0749565B1/en not_active Expired - Lifetime
  • 1995-03-10 AU AU18953/95A patent/AU1895395A/en not_active Abandoned
  • 1995-03-10 WO PCT/FI1995/000133 patent/WO1995024616A1/en active IP Right Grant
  • 1995-03-10 DE DE69527135T patent/DE69527135D1/de not_active Expired - Lifetime

Patent Citations (5)

Also Published As

Similar Documents

Legal Events