[go: up one dir, main page]
More Web Proxy on the site http://driver.im/

TWI734330B - Support device, support method and recording medium - Google Patents

Support device, support method and recording medium Download PDF

Info

Publication number
TWI734330B
TWI734330B TW108148512A TW108148512A TWI734330B TW I734330 B TWI734330 B TW I734330B TW 108148512 A TW108148512 A TW 108148512A TW 108148512 A TW108148512 A TW 108148512A TW I734330 B TWI734330 B TW I734330B
Authority
TW
Taiwan
Prior art keywords
parameter
data
change point
input
learning
Prior art date
Application number
TW108148512A
Other languages
Chinese (zh)
Other versions
TW202030567A (en
Inventor
西剛史
Original Assignee
日商住友重機械工業股份有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 日商住友重機械工業股份有限公司 filed Critical 日商住友重機械工業股份有限公司
Publication of TW202030567A publication Critical patent/TW202030567A/en
Application granted granted Critical
Publication of TWI734330B publication Critical patent/TWI734330B/en

Links

Images

Classifications

    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
    • G05B23/00—Testing or monitoring of control systems or parts thereof
    • G05B23/02—Electric testing or monitoring
    • G05B23/0205—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults
    • G05B23/0218—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults characterised by the fault detection method dealing with either existing or incipient faults
    • G05B23/0243—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults characterised by the fault detection method dealing with either existing or incipient faults model based detection method, e.g. first-principles knowledge model
    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
    • G05B23/00—Testing or monitoring of control systems or parts thereof
    • G05B23/02—Electric testing or monitoring
    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
    • G05B23/00—Testing or monitoring of control systems or parts thereof
    • G05B23/02—Electric testing or monitoring
    • G05B23/0205—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults
    • G05B23/0218—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults characterised by the fault detection method dealing with either existing or incipient faults
    • G05B23/0221—Preprocessing measurements, e.g. data collection rate adjustment; Standardization of measurements; Time series or signal analysis, e.g. frequency analysis or wavelets; Trustworthiness of measurements; Indexes therefor; Measurements using easily measured parameters to estimate parameters difficult to measure; Virtual sensor creation; De-noising; Sensor fusion; Unconventional preprocessing inherently present in specific fault detection methods like PCA-based methods
    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
    • G05B23/00—Testing or monitoring of control systems or parts thereof
    • G05B23/02—Electric testing or monitoring
    • G05B23/0205—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults
    • G05B23/0259—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults characterized by the response to fault detection
    • G05B23/0267—Fault communication, e.g. human machine interface [HMI]
    • G05B23/0272—Presentation of monitored results, e.g. selection of status reports to be displayed; Filtering information to the user
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00—Machine learning

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Automation & Control Theory (AREA)
  • Theoretical Computer Science (AREA)
  • Software Systems (AREA)
  • Evolutionary Computation (AREA)
  • Data Mining & Analysis (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Medical Informatics (AREA)
  • Artificial Intelligence (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Mathematical Physics (AREA)
  • Human Computer Interaction (AREA)
  • Testing Or Calibration Of Command Recording Devices (AREA)
  • Testing And Monitoring For Control Systems (AREA)

Abstract

本發明的課題為輕易地獲取確定感測器的計測值的經時變化之參數。支援裝置(10)用以獲取確定感測器的計測值的經時變化之參數，其具備：資料獲取部(21)，獲取表示第1感測器的計測值的經時變化之第1學習用資料和表示第2感測器的計測值的經時變化且與第1學習用資料相關之第2學習用資料；輸入部(23)，輸入參數；參數獲取部(25)，依據與由輸入部輸入之參數對應之第1學習用資料的變化點位置及第2學習用資料的變化點位置而獲取預定值的參數。 The subject of the present invention is to easily obtain a parameter that determines the change over time of the measured value of the sensor. The support device (10) is used to obtain parameters that determine the changes over time of the measured value of the sensor, and includes: a data acquisition unit (21) that obtains the first learning that represents the change over time of the measured value of the first sensor Use data and second learning data representing the change over time of the measured value of the second sensor and related to the first learning data; input section (23), input parameters; parameter acquisition section (25), basis and reason The parameter input by the input unit corresponds to the change point position of the first learning material and the change point position of the second learning material to obtain a parameter of a predetermined value.

Description

支援裝置、支援方法及記錄媒體 Support device, support method and recording medium

本申請主張基於2019年1月31日申請之日本專利申請第2019-016314號的優先權。該日本申請的全部內容藉由參閱援用於本說明書中。 This application claims priority based on Japanese Patent Application No. 2019-016314 filed on January 31, 2019. The entire contents of this Japanese application are incorporated in this specification by reference.

本發明係有關一種用以獲取確定感測器的計測值的經時變化之參數之支援裝置、支援方法及支援程式。 The present invention relates to a supporting device, a supporting method and a supporting program for obtaining parameters that determine changes in the measured value of a sensor over time.

作為能夠檢測各種處理系統的狀態變化的監視裝置，例如已知如專利文獻1及2之監視裝置。在這樣的監視裝置中，藉由確定感測器的計測值的經時變化而判定是否發生處理的狀態變化，因此要求獲取用以確定該經時變化之參數。 As a monitoring device capable of detecting changes in the state of various processing systems, for example, monitoring devices such as Patent Documents 1 and 2 are known. In such a monitoring device, it is determined whether a process state change has occurred by determining the change over time in the measured value of the sensor, and therefore it is required to obtain a parameter for determining the change over time.

(先前技術文獻) (Prior technical literature) (專利文獻) (Patent Document)

專利文獻1：日本特開2012-141712號公報 Patent Document 1: JP 2012-141712 A

專利文獻2：日本特開平11-7317號公報 Patent Document 2: Japanese Patent Laid-Open No. 11-7317

然而，以往確定感測器的計測值的經時變化之參數係藉由使用過去的經驗值和簡易分析值而得。因此，在如沒有過去的經驗值之計測資料中，需要在試誤下探索參數，難以獲取參數或獲取參數之效率低。又，還可進行藉由使用簡易分析值而獲取參數，但由於僅藉由簡易計算而獲取參數，因此難以確定感測器的計測值的準確的經時變化。 However, in the past, the parameters that determine the change over time of the measured value of the sensor are obtained by using past experience values and simple analysis values. Therefore, in the measurement data without past experience values, it is necessary to explore the parameters under trial and error, and it is difficult to obtain the parameters or the efficiency of obtaining the parameters is low. In addition, it is also possible to obtain parameters by using simple analysis values. However, since the parameters are obtained only by simple calculations, it is difficult to determine the accurate changes over time of the measured values of the sensors.

在這點上，在專利文獻1及2中的任一監視裝置中，關於計算確定感測器的計測值的經時變化之參數皆未作出任何對策。 In this regard, in any of the monitoring devices in Patent Documents 1 and 2, no countermeasures have been taken regarding the calculation of parameters that determine the changes over time in the measured values of the sensors.

本發明的一樣態的示例性目的之一為，提供一種用以輕易地獲取確定感測器的計測值的經時變化之參數之支援裝置、支援方法及支援程式。 One of the exemplary purposes of the present invention is to provide a support device, a support method, and a support program for easily acquiring parameters that determine the changes in the measured value of the sensor over time.

為了解決上述問題，本發明的一樣態的支援裝置為用以獲取確定感測器的計測值的經時變化之參數之支援裝置，其具備：資料獲取部，獲取表示第1感測器的計測值的經時變化之第1學習用資料和表示第2感測器的計測值的經時變化且與第1學習用資料相關之第2學習用資料；輸入部，輸入參數；及參數獲取部，依據與由輸入部輸入之參數對應之第1學習用資料的變化點位置及第2學習用資料的變化點位置而獲取預定值的參數。 In order to solve the above-mentioned problems, the homogeneous support device of the present invention is a support device for acquiring a parameter that determines the change over time of the measured value of the sensor. The first learning data showing the change over time of the value and the second learning data showing the change over time of the measured value of the second sensor and related to the first learning data; input unit, input parameter; and parameter acquisition unit , Obtain a parameter of a predetermined value based on the position of the change point of the first learning material and the position of the change point of the second learning material corresponding to the parameter input by the input unit.

依上述樣態，獲取第1學習用資料和與該第1學習用資料相關之第2學習用資料，依據與由輸入部輸入之參數對應之第1學習用資料的變化點位置及第2學習用資料的變化點位置而獲取預定值的參數。依此，能夠輕易地獲取確定感測器的計測值的經時變化之參數。 本發明的另一樣態為支援方法。該方法為用以獲取確定感測器的計測值的經時變化之參數之支援方法，其包括如下步驟：藉由資料獲取部獲取表示第1感測器的計測值的經時變化之第1學習用資料和表示第2感測器的計測值的經時變化且與第1學習用資料相關之第2學習用資料；藉由輸入部輸入參數；及依據與由輸入部輸入之參數對應之第1學習用資料的變化點位置和第2學習用資料的變化點位置，藉由參數獲取部獲取預定值的參數。 本發明的又一樣態樣態為支援程式。該程式為為了獲取確定感測器的計測值的經時變化之參數而由電腦執行之支援程式，其使電腦執行如下步驟：獲取表示第1感測器的計測值的經時變化之第1學習用資料和表示第2感測器的計測值的經時變化且與第1學習用資料相關之第2學習用資料；及依據與由輸入部輸入之參數對應之第1學習用資料的變化點位置和第2學習用資料的變化點位置，藉由參數獲取部獲取預定值的參數。 再者，在方法、裝置、系統、電腦程式、資料構造、記錄媒體等之間彼此替換以上的構成要素的任意的組合或本發明的構成要素和表述者作為本發明的樣態亦有效。Obtain the first learning materials and the second learning materials related to the first learning materials according to the above-mentioned pattern, based on the change point positions of the first learning materials corresponding to the parameters input by the input unit and the second learning The position of the change point of the data is used to obtain the parameter of the predetermined value. According to this, it is possible to easily obtain the parameter that determines the change over time of the measured value of the sensor. Another aspect of the present invention is a support method. The method is a support method for obtaining parameters that determine the change over time of the measured value of the sensor, and it includes the following steps: obtaining a first sensor that represents the change over time of the measured value of the first sensor by a data acquisition unit The learning data and the second learning data representing the change over time of the measured value of the second sensor and related to the first learning data; the parameters are input by the input unit; and the basis corresponds to the parameters input by the input unit The position of the change point of the first learning material and the position of the change point of the second learning material are acquired by the parameter acquiring unit to obtain parameters of predetermined values. Another aspect of the present invention is a support program. This program is a support program executed by a computer in order to obtain parameters that determine the change over time of the measured value of the sensor, which causes the computer to perform the following steps: Obtain the first value representing the change over time of the measured value of the first sensor The learning data and the second learning data representing the change over time of the measured value of the second sensor and related to the first learning data; and the change based on the first learning data corresponding to the parameters input by the input unit The point position and the change point position of the second learning data are obtained by the parameter obtaining unit to obtain parameters of predetermined values. Furthermore, it is also effective to replace any combination of the above constituent elements or the constituent elements and expressions of the present invention between methods, devices, systems, computer programs, data structures, recording media, etc., as aspects of the present invention.

(發明之效果) 依本發明，能夠輕易地獲取確定感測器的計測值的經時變化之參數。 (Effects of the invention) According to the present invention, it is possible to easily obtain the parameters that determine the change over time of the measured value of the sensor.

以下，參閱圖式並藉由發明的實施形態對本發明進行說明，但以下的實施形態並不限定申請專利範圍之發明，又，在實施形態中說明之所有特徵的組合並非為發明的解決方案中必不可少的。對各圖式所示之相同或等同的構成要素、構件、處理，標註相同符號，並適當省略重複之說明。 圖1～圖6係用以說明本發明的實施形態之支援裝置及支援方法之圖。具體而言，圖1係表示本發明的一實施形態之支援裝置10的構成之圖，圖2及圖3係表示基於支援裝置10之支援方法的一例之流程圖，圖4及圖5係用以說明計測資料的變化點位置的計算方法的一例之圖，圖6係用以說明基於支援裝置10之支援方法之圖。 支援裝置10為支援確定感測器的計測值的經時變化之參數的獲取者，係藉此支援例如發電廠或化學工廠等處理系統中的計測資料的波形的特徵變化檢測者。為了計測資料的波形的特徵變化檢測，需要在演算法中設定參數。這樣的參數為確定表示感測器的計測值的經時變化之計測資料的變化點位置者。藉由設定最佳的參數，能夠確定計測資料中的準確的變化點位置，能夠檢測出計測資料的準確的波形特徵變化。藉此例如能夠準確地評估處理系統的運行狀況。 支援裝置10具備參數最佳化機構20、資料評估機構30及資料記憶部40。參數最佳化機構20獲取作為確定感測器的計測值的經時變化之預定值的參數的一例之最佳參數。資料評估機構30使用由參數最佳化機構20獲取之最佳參數而確定表示感測器的計測值的經時變化之計測資料的變化點位置，藉此進行計測資料的波形的特徵變化檢測，甚至能夠評估處理系統的運行狀況。資料記憶部40儲存各種計測資料，並且存儲由參數最佳化機構20獲取之最佳參數或由參數最佳化機構20及資料評估機構30計算或獲取之計測資料的變化點位置資料。 支援裝置10例如與設置於處理系統中之複數個感測器(未圖示)連接，構成為能夠獲取表示該感測器的計測值的經時變化之計測資料。又，支援裝置10與用以輸入資訊之操作部(未圖示)及用以輸出資訊之顯示部(未圖示)連接。藉此，依據由操作部輸入之資訊而進行演算，並且將其演算結果顯示於顯示部，使得作業人員能夠一邊藉由顯示部識別畫面一邊藉由操作部對支援裝置10輸入所需之資訊。支援裝置10為具備CPU及記憶體等之電腦裝置。記憶體中儲存有用以執行基於本實施形態之支援裝置10之支援方法的各動作之支援程式。再者，規定後述之本實施形態之支援方法之程式在電腦上運行而使CPU進行之處理分別與本實施形態的支援裝置10及支援方法中的對應之要素的功能及動作相同。 以下，對支援裝置10的各種功能方塊進行說明。 參數最佳化機構20具備學習用資料獲取部21、雜訊去除部22、參數輸入部23、資料顯示部24及最佳參數獲取部25。為了獲取最佳參數，學習用資料獲取部21獲取表示感測器的計測值的經時變化之計測資料作為學習用資料。在此，感測器例如為壓力感測器、溫度感測器或流量感測器。雜訊去除部22去除由學習用資料獲取部21獲取之計測資料的雜訊而僅獲取已去除雜訊之計測資料。參數輸入部23受理用以確定感測器的計測值的經時變化之參數的輸入。該參數的輸入例如藉由作業人員經由操作部輸入而進行。 資料顯示部24顯示由雜訊去除部22獲取之已去除雜訊之計測資料。又，資料顯示部24在顯示部中顯示用以敦促作業人員輸入參數之參數輸入欄、獲取最佳參數所需之執行按鈕(例如變化點計算執行按鈕及變化點提取執行按鈕)及學習用資料的變化點位置的一致率的判定結果等參數最佳化機構20獲取最佳參數所需之資訊(參閱圖6)。 最佳參數獲取部25具備計算計測資料的變化點位置之變化點位置計算部26和提取由變化點位置計算部26計算出之變化點位置之變化點位置提取部27。變化點位置的計算方法並不受限定，例如能夠使用k近鄰法(k-nearest neighbor algorithm)(參閱圖4)或奇異譜轉換(Singular Spectrum Transformation)(參閱圖5)等眾所周知的波形特徵變化檢測方法。 資料評估機構30具備評估用資料獲取部31、雜訊去除部32、最佳參數輸入部33、資料顯示部34及變化點位置獲取部35。為了檢測計測資料的波形的特徵變化，評估用資料獲取部31獲取表示感測器的計測值的經時變化之計測資料作為評估用資料。評估用資料與由參數最佳化機構20的學習用資料獲取部21獲取之學習用資料為基於相同的感測器的計測值之相同種類的計測資料。如此，使用從相同種類的計測資料獲得之最佳參數而進行計測資料的波形的特徵變化檢測，因此能夠確定計測資料中的準確的變化點位置。 雜訊去除部32去除由評估用資料獲取部31獲取之計測資料的雜訊而僅獲取已去除雜訊之計測資料。最佳參數輸入部33受理用以確定感測器的計測值的經時變化之最佳參數的輸入。最佳參數的輸入例如藉由從參數最佳化機構20或資料記憶部40接收由最佳參數獲取部25獲取之最佳參數而進行。或者，還能夠藉由作業人員經由操作部輸入由參數最佳化機構20獲取之最佳參數而進行最佳參數的輸入。 資料顯示部34顯示由雜訊去除部32獲取之已去除雜訊之計測資料。又，與參數最佳化機構20的資料顯示部24相同地，資料顯示部34在顯示部中顯示用以敦促作業人員輸入最佳參數之參數輸入欄、評估資料所需之執行按鈕(例如變化點計算執行按鈕及變化點提取執行按鈕)等資料評估機構30評估資料所需之資訊。 變化點位置獲取部35具備計算計測資料的變化點位置之變化點位置計算部36和提取由變化點位置計算部36計算出之變化點位置之變化點位置提取部37。變化點位置的計算方法並不受限定，與基於最佳參數獲取部25之計算方法相同，例如能夠使用k近鄰法(參閱圖4)或奇異譜轉換(參閱圖5)等眾所周知的波形特徵變化檢測方法。 資料記憶部40具備計測資料記憶部41、最佳參數記憶部42及變化點位置資料記憶部43。計測資料記憶部41儲存來自設置於處理系統中的各感測器之計測資料。儲存之計測資料包括用以處理參數最佳化機構20之學習用資料和用以處理資料評估機構30之評估用資料。最佳參數記憶部42儲存由最佳參數獲取部25獲取之最佳參數。變化點位置資料記憶部43儲存分別在變化點位置提取部27、37中提取之計測資料的變化點位置資料。儲存於資料記憶部40中之該等記憶資料例如與處理系統的運行時間或運行狀況等建立對應關係。 再者，關於上述之參數最佳化機構20、資料評估機構30及資料記憶部40的各種構成的具體動作，將在後述之支援方法中進行詳細敘述。 以下，作為本發明的一實施形態之支援方法，對使用支援裝置10之動作的一例進行說明。首先，參閱圖2對使用支援裝置10的參數最佳化機構20之動作的一例進行說明。 在圖2中，首先，由學習用資料獲取部21獲取表示第1感測器的計測值的經時變化之第1學習用資料和表示第2感測器的計測值的經時變化之第2學習用資料(S10)。第1學習用資料與在資料評估機構30中作為評估對象之計測資料為相同種類的計測資料，這樣的計測資料能夠從與由評估用資料獲取部31獲取之計測資料的感測器相同的感測器獲取。第2學習用資料為與第1學習用資料相關之計測資料。第2學習用資料為針對第1學習用資料之基準資料。典型而言，第2學習用資料為與第1學習用資料具有輸入或輸出(換言之原因或結果)的響應關係之計測資料。此時，第1學習資料和第2學習資料並不限於如一方相對於另一方藉由一個系統輸入輸出之直接輸入輸出之樣態，還包括如藉由複數個系統輸入輸出之間接輸入輸出之樣態(例如，第1學習資料輸入於第1系統，該第1系統的輸出成為第2系統的輸入，該第2系統的輸出為第2學習資料之樣態)。具有這樣的輸入或輸出的響應關係之計測資料彼此具有相同位置的變化點位置。例如，在第1學習用資料為來自檢測處理系統的既定位置的溫度之溫度感測器之計測資料之情況下，第2學習用資料可以為來自檢測成為該既定位置的溫度變化的原因之蒸汽的壓力或流量之感測器的計測資料。或者，第2學習用資料可以為相關係數與第1學習用資料具有一定值以上的關係之計測資料。 接著，藉由雜訊去除部22去除第1學習用資料及第2學習用資料的雜訊(S11)，藉由資料顯示部24將去除雜訊之第1學習用資料及第2學習用資料顯示於顯示部(S12)。然後，作業人員視覺辨認顯示於顯示部之第1學習用資料及第2學習用資料的各波形，並且經由操作部藉由參數輸入部23輸入確定感測器的計測值的經時變化之參數(S13)。此處的參數為作業人員適當確定之臨時參數。 之後，依據在步驟S13中輸入之參數，藉由變化點位置計算部26分別計算第1學習用資料及第2學習用資料的變化點位置(S14)。 作為變化點位置的計算方法的一例，可舉出圖4所示之k近鄰法。在圖4中，橫軸為時間，縱軸為感測器的計測值。k近鄰法為眾所周知的波形特徵變化檢測方法，將簡單地進行說明，k近鄰法中，以計算時刻t為界在未來側創建長度w的向量d。藉由滑動相同長度w的向量而在過去側準備n個向量qi，創建過去矩陣(1列為1向量)。當創建過去矩陣時，在未來側向量與過去矩陣之間的時間的距離短的情況下，類似度高，因此變化度小。為了避免該情況，如圖4所示，設置間隔距離g。藉由將過去側的各向量和未來側的向量代入到使用餘弦距離之下述式中而計算臨時變化度ztmp 。 【數式1】

Figure 02_image001
然後，如下述式將最小值設為變化度z。 z=minztmp 在k近鄰法中，在步驟S13中輸入之參數為相當於圖4的橫軸的時間的寬度之參數。具體而言，該參數與圖4的間隔距離g、時間寬度M及窗尺寸w對應。 作為變化點位置的計算方法的另一例，可舉出圖5所示之奇異譜轉換。在圖5中，橫軸為時間，縱軸為感測器的計測值。奇異譜轉換(Singular Spectrum Transform)中，在比變化度計算時刻t更靠過去側以任意的長度(窗尺寸)w切出時序資料而創建向量，並藉由以點數τ滑動該向量而創建n個向量。將這n個向量用作過去側(n×w)矩陣。藉由對該矩陣進行特異值分解而取出任意個過去的代表向量。另一方面，在未來側亦創建相同的矩陣，並進行特異值分解，由此取出1個未來的代表向量。藉由使用由複數個過去的代表向量構成之矩陣U和1個未來的代表向量β(t)藉由下述式而計算該時刻的變化度z(t)。 【數式2】
Figure 02_image003
【數式3】
Figure 02_image005
在奇異譜轉換中，在步驟S13中輸入之參數為相當於圖5的橫軸的時間的寬度之參數。具體而言，該參數與圖5的間隔距離g、時間寬度M及窗尺寸w對應。 作業人員藉由參數輸入部23輸入該等複數個參數，並藉由k近鄰法或奇異譜轉換等波形的特徵變化檢測方法而計算第1學習用資料及第2學習用資料的各變化點位置。之後，藉由變化點位置提取部27提取在步驟S14中計算出之變化點位置(S15)，針對第1學習用資料及第2學習用資料將各變化點位置的一致率的判定結果與該等變化點位置一併顯示於顯示部(S16)。 在此，圖6係步驟S16後的基於顯示部之顯示樣態的一例。在該例中，顯示區域50包括第1學習用資料顯示欄60、第2學習用資料顯示欄70、參數輸入欄80、變化點計算執行按鈕90、變化點提取執行按鈕92及一致率判定結果94。第1學習用資料顯示欄60中，在以第1感測器的計測值為縱軸及以時間為橫軸之座標軸中，示出了第1學習用資料的波形62和其變化點位置64。變化點位置64在週期性地重複大致類似的圖案之波形的各週期中以縱線顯示。同樣地，第2學習用資料顯示欄70中，在以第2感測器的計測值為縱軸及以時間為橫軸之座標軸中，示出了第2學習用資料的波形72和其變化點位置74。變化點位置74在週期性地重複大致類似的圖案之波形的各週期中以縱線顯示。如此，藉由以使表示時間之橫軸一致的方式配置顯示欄60、70，能夠視覺辨認各學習用資料的變化點位置64、74的一致率。 又，為了計算各學習用資料的變化點位置64、74，在參數輸入欄80中顯示在步驟S13中輸入之各參數82、84、86。藉此，能夠視覺辨認各參數與各學習用資料的變化點位置之間的對應關係。藉由作業人員經由操作部選擇變化點計算執行按鈕90及變化點提取執行按鈕92，能夠計算或提取各學習用資料的變化點位置。一致率判定結果94包括一致率的判定結果表示一致時之項目(“是”)96和一致率的判定結果表示不一致之項目(“否”)98，並依據一致率的判定結果而點亮任一項目的燈。藉此，作業人員能夠輕易地視覺辨認與在步驟S13中輸入之參數對應之第1學習用資料的變化點位置與第2學習用資料的變化點位置是否一致。 返回到圖2的流程圖，在步驟S17中藉由最佳參數獲取部25判定第1學習用資料的變化點位置與第2學習用資料的變化點位置是否一致，並在判定為不一致之情況下(S17中為“否”)，資料顯示部24敦促作業人員再次輸入參數，並反覆進行步驟S13至S16的一系列的步驟至判定結果判定為一致為止。 在判定結果判定為一致之情況下(S17中為“是”)，藉由最佳參數獲取部25獲取此時的預定值的參數作為最佳參數。獲取之最佳參數儲存於最佳參數記憶部42中，並在後述之資料評估機構中使用。 在步驟S17中的一致率的判定中，比較各變化點位置64、74，並能夠以是否在允許誤差內一致而進行判定。亦即，在變化點位置的一致率落在允許誤差內之情況下，將該參數作為最佳參數。再者，變化點位置的一致率的計算亦可以由最佳參數獲取部25自動進行。 接著，參閱圖3對使用支援裝置10的資料評估機構30之動作的一例進行說明。在圖3中，首先，為了檢測計測資料的波形的特徵變化，藉由評估用資料獲取部31獲取與第1學習用資料相同種類的計測資料作為評估用資料(S10)。評估用資料與第1學習用資料為相同的感測器的計測值且不同時間的計測資料。 接著，藉由雜訊去除部32去除評估用資料的雜訊(S21)，並藉由資料顯示部34將去除雜訊之評估用資料顯示於顯示部(S22)。又，最佳參數輸入部33從最佳參數記憶部42獲取針對與評估用資料相同種類的計測資料之最佳參數，並將該最佳參數輸入於資料評估機構30(S23)。再者，最佳參數的輸入還能夠由作業人員視覺辨認顯示於顯示部之評估用資料的波形之同時經由操作部再次輸入最佳參數。 之後，依據在步驟S23中輸入之最佳參數，藉由變化點位置計算部36而計算評估用資料的變化點位置(S24)。 在參數最佳化機構20的動作的一例中說明之內容對應於變化點位置的計算方法。 如此，作業人員依據最佳參數而藉由k近鄰法或奇異譜轉換等波形的特徵變化檢測方法而計算評估用資料的變化點位置，之後，藉由變化點位置提取部37提取在步驟S24中計算出之變化點位置(S25)。如此提取之變化點位置顯示於顯示部作為評估用資料的評估結果(S26)。 如上所述，本實施形態之支援裝置獲取第1學習用資料和與該第1學習用資料相關之第2學習用資料，依據與由輸入部輸入之參數對應之第1學習用資料的變化點位置及第2學習用資料的變化點位置而獲取預定值的參數。依此，能夠輕易地獲取確定感測器的計測值的經時變化之參數。因此，例如，即使在如沒有過去的經驗值之計測資料中，亦能夠高效地獲取參數，藉此能夠輕易地進行計測資料的波形的特徵變化檢測。又，與依據計測值的臨界值而進行波形的特徵變化檢測之方法相比，例如還能夠輕易地掌握導致在計測值的臨界值以下發生之故障之現象，因此例如能夠毫無遺漏地檢測處理系統的變化。此外，例如，在處理系統中，計測資料的變化還有可能為導致系統異常之一因素，因此藉由運用本實施形態之支援裝置、支援方法及支援程式，還能夠有助於系統的異常偵知。 本發明並不限定於上述實施形態，能夠運用各種變形。 在上述實施形態中，作為獲取最佳參數之一例，對使用一個第1學習用資料之例子進行了說明，但本發明並不限於此，亦可以藉由比較複數個第1學習用資料與第2學習用資料而獲取最佳參數。亦即，第1學習用資料的計測資料亦可以為多維。這樣的計測資料可以為複數個第1感測器各自的計測值的經時變化。此時，針對第1學習用資料及第2學習用資料之各變化點位置的一致率的判定例如可以藉由如下方式進行：提取分別對複數個第1學習用資料進行平均等並相加而得之變化點位置，並比較該變化點位置與第2學習用資料的變換點位置。藉由使用複數個第1學習用資料，能夠減少處理系統的雜訊的影響而提取更準確的變化點位置。 又，在上述實施形態中，對支援裝置10具備參數最佳化機構20及資料評估機構30之例子進行了說明，但本發明並不限於此，支援裝置亦可以為至少具備參數最佳化機構20者。又，支援裝置10的動作並不限於所有動作被電腦的演算處理自動化者，還包括至少一部分介入基於作業人員之手動作業者。又，在上述實施形態中，基於顯示部之顯示樣態僅為一例，例如圖6的學習用資料並不限於基於圖表之波形資料的顯示，亦可以為基於數值之表形式等。 藉由上述發明的實施形態說明之實施樣態能夠依據用途而適當組合或者加以變更或改良而使用，本發明並不限定於上述之實施形態的記載。從申請專利範圍的記載而言，這樣的組合或者加以變更或改良之形態亦可以包括在本發明的技術範圍內係顯而易見的。Hereinafter, the present invention will be described with reference to the drawings and embodiments of the invention. However, the following embodiments do not limit the invention within the scope of the patent application. Moreover, the combination of all the features described in the embodiments is not a solution to the invention. necessary. The same or equivalent constituent elements, members, and processes shown in the various drawings are labeled with the same symbols, and repeated descriptions are appropriately omitted. 1 to 6 are diagrams for explaining the supporting device and the supporting method of the embodiment of the present invention. Specifically, FIG. 1 is a diagram showing the structure of a supporting device 10 according to an embodiment of the present invention, FIGS. 2 and 3 are flowcharts showing an example of a supporting method based on the supporting device 10, and FIGS. 4 and 5 are used Taking a diagram for explaining an example of a calculation method of the change point position of the measurement data, FIG. 6 is a diagram for explaining the support method by the support device 10. The support device 10 supports the acquisition of parameters that determine the changes over time in the measured values of the sensors, and thereby supports the detectors of characteristic changes in waveforms of measurement data in processing systems such as power plants or chemical plants. In order to detect the characteristic change of the waveform of the measurement data, it is necessary to set parameters in the algorithm. Such parameters are those that determine the position of the change point of the measurement data indicating the change over time in the measurement value of the sensor. By setting the optimal parameters, the accurate change point position in the measurement data can be determined, and the accurate waveform characteristic change of the measurement data can be detected. In this way, for example, the operating status of the processing system can be accurately evaluated. The support device 10 includes a parameter optimization mechanism 20, a data evaluation mechanism 30, and a data storage unit 40. The parameter optimizing mechanism 20 acquires an optimal parameter as an example of a parameter that determines a predetermined value of a time-dependent change in the measured value of the sensor. The data evaluation mechanism 30 uses the optimal parameters acquired by the parameter optimization mechanism 20 to determine the change point position of the measurement data representing the change over time of the measurement value of the sensor, thereby performing feature change detection of the waveform of the measurement data, It can even evaluate the health of the processing system. The data storage unit 40 stores various measurement data, and stores the optimal parameters acquired by the parameter optimization mechanism 20 or the change point position data of the measurement data calculated or acquired by the parameter optimization mechanism 20 and the data evaluation mechanism 30. The support device 10 is connected to, for example, a plurality of sensors (not shown) installed in the processing system, and is configured to be able to acquire measurement data indicating changes over time in the measurement values of the sensors. In addition, the support device 10 is connected to an operation unit (not shown) for inputting information and a display unit (not shown) for outputting information. Thereby, the calculation is performed based on the information input by the operation part, and the calculation result is displayed on the display part, so that the operator can input required information to the support device 10 through the operation part while recognizing the screen through the display part. The support device 10 is a computer device equipped with a CPU, a memory, and the like. The memory stores a support program for executing each action of the support method based on the support device 10 of this embodiment. In addition, the program defining the support method of the present embodiment described later runs on a computer and the processing performed by the CPU is the same as the functions and operations of the corresponding elements in the support device 10 and the support method of the present embodiment, respectively. Hereinafter, various functional blocks of the support device 10 will be described. The parameter optimization mechanism 20 includes a learning data acquisition unit 21, a noise removal unit 22, a parameter input unit 23, a data display unit 24, and an optimal parameter acquisition unit 25. In order to acquire the optimal parameter, the learning material acquisition part 21 acquires the measurement data which shows the temporal change of the measured value of a sensor as a learning material. Here, the sensor is, for example, a pressure sensor, a temperature sensor, or a flow sensor. The noise removal unit 22 removes the noise of the measurement data acquired by the learning data acquisition unit 21 and acquires only the measurement data from which the noise has been removed. The parameter input unit 23 receives input of a parameter for determining the change over time in the measured value of the sensor. The input of this parameter is performed, for example, by the operator's input via the operation unit. The data display unit 24 displays the noise-removed measurement data acquired by the noise removal unit 22. In addition, the data display unit 24 displays on the display unit a parameter input field for urging the operator to input parameters, execution buttons (such as change point calculation execution button and change point extraction execution button) and learning data required to obtain the optimal parameter. The parameter optimizing mechanism 20, such as the determination result of the coincidence rate of the position of the change point, obtains the information required by the optimal parameter (see FIG. 6). The optimal parameter acquisition unit 25 includes a change point position calculation unit 26 that calculates the change point position of the measurement data, and a change point position extraction unit 27 that extracts the change point position calculated by the change point position calculation unit 26. The calculation method of the change point position is not limited. For example, the well-known waveform feature change detection such as k-nearest neighbor algorithm (refer to Figure 4) or Singular Spectrum Transformation (refer to Figure 5) can be used. method. The data evaluation mechanism 30 includes an evaluation data acquisition unit 31, a noise removal unit 32, an optimal parameter input unit 33, a data display unit 34, and a change point position acquisition unit 35. In order to detect the characteristic change of the waveform of the measurement data, the evaluation data acquisition unit 31 acquires the measurement data representing the change over time in the measurement value of the sensor as the evaluation data. The evaluation data and the learning data acquired by the learning data acquisition unit 21 of the parameter optimization mechanism 20 are the same type of measurement data based on the measurement value of the same sensor. In this way, the characteristic change detection of the waveform of the measurement data is performed using the optimal parameters obtained from the same type of measurement data, so that the accurate change point position in the measurement data can be determined. The noise removal unit 32 removes the noise of the measurement data acquired by the evaluation data acquisition unit 31 and acquires only the measurement data from which the noise has been removed. The optimal parameter input unit 33 receives input of the optimal parameter for determining the temporal change of the measured value of the sensor. The input of the optimal parameter is performed, for example, by receiving the optimal parameter acquired by the optimal parameter acquisition unit 25 from the parameter optimization mechanism 20 or the data storage unit 40. Alternatively, it is also possible to input the optimal parameter by the operator inputting the optimal parameter obtained by the parameter optimization mechanism 20 through the operating unit. The data display unit 34 displays the noise-removed measurement data acquired by the noise removal unit 32. Also, similar to the data display unit 24 of the parameter optimization mechanism 20, the data display unit 34 displays on the display unit a parameter input field for urging the operator to input the optimal parameter, and an execution button (for example, change Click the calculation execution button and the change point extraction execution button) and other information needed by the data evaluation agency 30 to evaluate the data. The change point position acquisition unit 35 includes a change point position calculation unit 36 that calculates the change point position of the measurement data, and a change point position extraction unit 37 that extracts the change point position calculated by the change point position calculation unit 36. The calculation method of the change point position is not limited, and is the same as the calculation method based on the optimal parameter acquisition unit 25. For example, well-known waveform feature changes such as the k-nearest neighbor method (refer to Fig. 4) or singular spectrum conversion (refer to Fig. 5) can be used. Detection method. The data storage unit 40 includes a measurement data storage unit 41, an optimal parameter storage unit 42, and a change point position data storage unit 43. The measurement data storage unit 41 stores measurement data from each sensor installed in the processing system. The stored measurement data includes learning data used by the processing parameter optimization mechanism 20 and evaluation data used by the processing data evaluation mechanism 30. The optimal parameter storage unit 42 stores the optimal parameters acquired by the optimal parameter acquisition unit 25. The change point position data storage unit 43 stores the change point position data of the measurement data extracted by the change point position extraction units 27 and 37, respectively. The memory data stored in the data memory unit 40 establish a corresponding relationship with the operating time or operating status of the processing system, for example. Furthermore, the specific operations of the various configurations of the parameter optimization mechanism 20, the data evaluation mechanism 30, and the data storage unit 40 described above will be described in detail in the support method described later. Hereinafter, as a support method according to an embodiment of the present invention, an example of the operation using the support device 10 will be described. First, referring to FIG. 2, an example of the operation of the parameter optimization mechanism 20 using the support device 10 will be described. In FIG. 2, first, the learning data acquisition unit 21 acquires the first learning data representing the change with time of the measured value of the first sensor and the first learning data representing the change with time of the measured value of the second sensor. 2 Learning materials (S10). The first learning material is the same type of measurement data as the measurement data to be evaluated in the data evaluation organization 30. Such measurement data can be obtained from the same sensor as the sensor of the measurement data acquired by the evaluation data acquisition unit 31 Detector acquisition. The second learning materials are measurement data related to the first learning materials. The second learning materials are the reference materials for the first learning materials. Typically, the second learning material is measurement data that has a response relationship of input or output (in other words, cause or effect) with the first learning material. At this time, the first learning material and the second learning material are not limited to the form of direct input and output of one system with respect to the other, but also include the form of input and output through multiple system inputs and outputs. The aspect (for example, the first learning material is input to the first system, the output of the first system becomes the input of the second system, and the output of the second system is the aspect of the second learning material). The measurement data having such input or output response relationship have the same position of the change point position. For example, when the first learning data is measurement data from a temperature sensor that detects the temperature at a predetermined position of the processing system, the second learning data may be steam from detecting the cause of the temperature change at the predetermined position. The measurement data of the sensor of the pressure or flow rate. Alternatively, the second learning material may be measurement data in which the correlation coefficient and the first learning material have a relationship of a certain value or more. Next, the noise removal unit 22 removes the noise of the first learning data and the second learning data (S11), and the data display unit 24 removes the noise of the first learning data and the second learning data It is displayed on the display part (S12). Then, the operator visually recognizes the waveforms of the first learning data and the second learning data displayed on the display unit, and inputs the parameters that determine the change over time of the measured value of the sensor through the parameter input unit 23 through the operation unit (S13). The parameters here are temporary parameters appropriately determined by the operator. After that, based on the parameters input in step S13, the change point position calculation unit 26 calculates the change point positions of the first learning data and the second learning data (S14). As an example of the method of calculating the position of the change point, the k-nearest neighbor method shown in FIG. 4 can be cited. In FIG. 4, the horizontal axis is time, and the vertical axis is the measured value of the sensor. The k-nearest neighbor method is a well-known waveform feature change detection method, which will be briefly explained. In the k-nearest neighbor method, a vector d of length w is created on the future side with the calculation time t as the boundary. By sliding vectors of the same length w, n vectors qi are prepared on the past side to create a past matrix (1 column is 1 vector). When the past matrix is created, if the time distance between the future vector and the past matrix is short, the similarity is high, and therefore the degree of change is small. In order to avoid this situation, as shown in Fig. 4, an interval distance g is set. The temporary change degree z tmp is calculated by substituting each vector on the past side and a vector on the future side into the following equation using the cosine distance. [Numerical formula 1]
Figure 02_image001
Then, the minimum value is set as the degree of change z as in the following equation. z=minz tmp In the k-nearest neighbor method, the parameter input in step S13 is a parameter corresponding to the time width on the horizontal axis of FIG. 4. Specifically, this parameter corresponds to the separation distance g, the time width M, and the window size w in FIG. 4. As another example of the calculation method of the change point position, the singular spectrum conversion shown in FIG. 5 can be cited. In FIG. 5, the horizontal axis is time, and the vertical axis is the measured value of the sensor. In Singular Spectrum Transform, a vector is created by cutting out time series data with an arbitrary length (window size) w on the side farther from the time t of the degree of change calculation, and then sliding the vector by the number of points τ. n vectors. These n vectors are used as the past (n×w) matrix. Any number of past representative vectors is extracted by performing specific value decomposition on the matrix. On the other hand, the same matrix is also created on the future side, and singular value decomposition is performed, thereby extracting a representative vector of the future. The degree of change z(t) at the time is calculated by the following formula by using a matrix U composed of a plurality of past representative vectors and a future representative vector β(t). 【Numeral 2】
Figure 02_image003
【Numeral 3】
Figure 02_image005
In the singular spectrum conversion, the parameter input in step S13 is a parameter corresponding to the time width on the horizontal axis of FIG. 5. Specifically, this parameter corresponds to the separation distance g, the time width M, and the window size w in FIG. 5. The operator inputs the plurality of parameters through the parameter input unit 23, and calculates the position of each change point of the first learning data and the second learning data by a waveform feature change detection method such as k-nearest neighbor method or singular spectrum conversion . After that, the change point position extraction unit 27 extracts the change point position calculated in step S14 (S15), and compares the determination result of the coincidence rate of each change point position with respect to the first learning data and the second learning data. The position of the isochanging point is displayed on the display unit (S16). Here, FIG. 6 is an example of the display aspect based on the display part after step S16. In this example, the display area 50 includes a first learning material display field 60, a second learning material display field 70, a parameter input field 80, a change point calculation execution button 90, a change point extraction execution button 92, and a coincidence rate determination result 94. In the first learning data display column 60, the waveform 62 of the first learning data and its change point 64 are shown in the coordinate axis with the measured value of the first sensor on the vertical axis and time on the horizontal axis. . The change point position 64 is displayed as a vertical line in each cycle in which the waveform of a substantially similar pattern is periodically repeated. Similarly, in the second learning data display column 70, the waveform 72 of the second learning data and its changes are shown on the coordinate axis with the measured value of the second sensor on the vertical axis and time on the horizontal axis. Point location 74. The change point position 74 is displayed as a vertical line in each cycle in which the waveform of a substantially similar pattern is periodically repeated. In this way, by arranging the display columns 60 and 70 such that the horizontal axis indicating time coincides, the coincidence rate of the change point positions 64 and 74 of each learning material can be visually recognized. In addition, in order to calculate the change point positions 64, 74 of each learning material, the parameters 82, 84, and 86 input in step S13 are displayed in the parameter input column 80. Thereby, the correspondence relationship between each parameter and the change point position of each learning material can be visually recognized. When the operator selects the change point calculation execution button 90 and the change point extraction execution button 92 via the operation unit, the position of the change point of each learning material can be calculated or extracted. The coincidence rate judgment result 94 includes the item when the judgment result of the coincidence rate indicates agreement ("Yes") 96 and the item when the judgment result of the agreement rate indicates disagreement ("No") 98, and any item is lit based on the judgment result of the agreement rate. A lamp for an item. Thereby, the operator can easily visually recognize whether the change point position of the first learning material corresponding to the parameter input in step S13 is consistent with the change point position of the second learning material. Returning to the flowchart of FIG. 2, in step S17, the optimal parameter acquisition unit 25 determines whether the change point position of the first learning data and the change point position of the second learning data are consistent, and if it is judged to be inconsistent Next (No in S17), the data display unit 24 urges the operator to input the parameters again, and repeats a series of steps from steps S13 to S16 until the judgment result is judged to match. When the result of the determination is determined to be the same (Yes in S17), the optimal parameter acquiring unit 25 acquires the parameter of the predetermined value at this time as the optimal parameter. The obtained optimal parameters are stored in the optimal parameter memory 42 and used in the data evaluation mechanism described later. In the determination of the coincidence rate in step S17, the respective change point positions 64 and 74 are compared, and it is possible to determine whether or not they coincide within an allowable error. That is, in the case where the coincidence rate of the position of the change point falls within the allowable error, this parameter is taken as the optimal parameter. Furthermore, the calculation of the coincidence rate of the position of the change point can also be automatically performed by the optimal parameter acquisition unit 25. Next, an example of the operation of the data evaluation mechanism 30 using the support device 10 will be described with reference to FIG. 3. In FIG. 3, first, in order to detect the characteristic change of the waveform of the measurement data, the evaluation data acquisition unit 31 acquires the same type of measurement data as the first learning data as the evaluation data (S10). The evaluation data and the first learning data are measured values of the same sensor and measured data at different times. Next, the noise removal part 32 removes the noise of the evaluation data (S21), and the data display part 34 displays the noise-removed evaluation data on the display part (S22). In addition, the optimal parameter input unit 33 acquires the optimal parameter for the measurement data of the same type as the evaluation data from the optimal parameter storage unit 42, and inputs the optimal parameter into the data evaluation agency 30 (S23). Furthermore, the input of the optimal parameter can be visually recognized by the operator while re-entering the optimal parameter via the operation unit while visually recognizing the waveform of the evaluation data displayed on the display unit. After that, based on the optimal parameter input in step S23, the change point position calculation unit 36 calculates the change point position of the evaluation data (S24). The content described in an example of the operation of the parameter optimization mechanism 20 corresponds to the calculation method of the position of the change point. In this way, the operator calculates the change point position of the evaluation data by the k-nearest neighbor method or singular spectrum conversion and other waveform feature change detection methods based on the optimal parameters, and then the change point position extraction unit 37 extracts the change point in step S24 The calculated change point position (S25). The position of the change point thus extracted is displayed on the display part as the evaluation result of the evaluation data (S26). As described above, the support device of this embodiment acquires the first learning data and the second learning data related to the first learning data, based on the change point of the first learning data corresponding to the parameters input by the input unit The position and the position of the change point of the second learning material obtain parameters of predetermined values. According to this, it is possible to easily obtain the parameter that determines the change over time of the measured value of the sensor. Therefore, for example, even in measurement data such as no past experience value, parameters can be obtained efficiently, whereby the characteristic change detection of the waveform of the measurement data can be easily performed. In addition, compared with the method of detecting the characteristic change of the waveform based on the critical value of the measured value, for example, it is also possible to easily grasp the phenomenon that causes a malfunction below the critical value of the measured value, so for example, it is possible to detect and process without any omission. System changes. In addition, for example, in a processing system, changes in measurement data may be one of the factors that cause system abnormalities. Therefore, the use of the support device, support method, and support program of this embodiment can also help the system to detect abnormalities. Know. The present invention is not limited to the above-mentioned embodiment, and various modifications can be applied. In the above embodiment, as an example of obtaining the best parameter, an example of using a first learning material is explained, but the present invention is not limited to this, and it is also possible to compare a plurality of first learning materials with the first learning material. 2 Learn to use materials to obtain the best parameters. That is, the measurement data of the first learning data may be multi-dimensional. Such measurement data may be changes over time in the measurement values of each of the plurality of first sensors. At this time, the determination of the coincidence rate of each change point position of the first learning data and the second learning data can be performed by, for example, extracting and averaging a plurality of first learning data and adding them. Obtain the change point position, and compare the change point position with the change point position of the second learning material. By using a plurality of first learning data, it is possible to reduce the influence of the noise of the processing system and extract more accurate change point positions. Furthermore, in the above-mentioned embodiment, the example in which the support device 10 includes the parameter optimization mechanism 20 and the data evaluation mechanism 30 has been described, but the present invention is not limited to this, and the support device may have at least a parameter optimization mechanism. 20 persons. In addition, the actions of the support device 10 are not limited to those whose all actions are automated by computer arithmetic processing, but also include at least a part of those who intervene in manual operations based on the operator. In addition, in the above-mentioned embodiment, the display mode based on the display unit is only an example. For example, the learning data of FIG. 6 is not limited to the display of waveform data based on graphs, and may be in the form of a table based on numerical values. The implementation modes explained by the above-mentioned embodiments of the invention can be appropriately combined or used with changes or improvements depending on the application, and the present invention is not limited to the description of the above-mentioned embodiments. From the description of the scope of patent application, it is obvious that such a combination or a modified or improved form can also be included in the technical scope of the present invention.

10:支援裝置 21:學習用資料獲取部 23:參數輸入部 25:最佳參數獲取部 24:資料顯示部 35:變化點位置獲取部10: Support device 21: Learning Materials Acquisition Department 23: Parameter input section 25: Best parameter acquisition department 24: Data display department 35: Change point position acquisition department

[圖1]係表示本發明的一實施形態之支援裝置10的構成之圖。 [圖2]係表示基於支援裝置10之支援方法的一例之流程圖。 [圖3]係表示基於支援裝置10之支援方法的一例之流程圖。 [圖4]係用以說明作為計測資料的變化點位置的計算方法的一例之k近鄰法之圖。 [圖5]係用以說明作為計測資料的變化點位置的計算方法的一例之奇異譜轉換之圖。 [圖6]係用以說明基於支援裝置10之支援方法之圖。Fig. 1 is a diagram showing the structure of a support device 10 according to an embodiment of the present invention. [FIG. 2] is a flowchart showing an example of a support method based on the support device 10. [FIG. 3] is a flowchart showing an example of a support method based on the support device 10. Fig. 4 is a diagram for explaining the k-nearest neighbor method as an example of the calculation method of the change point position of the measurement data. [Fig. 5] A diagram for explaining the singular spectrum conversion as an example of the calculation method of the change point position of the measurement data. [Fig. 6] is a diagram for explaining the support method based on the support device 10.

10:支援裝置 10: Support device

20:參數最佳化機構 20: Parameter optimization mechanism

21:學習用資料獲取部 21: Learning Materials Acquisition Department

22:雜訊去除部 22: Noise removal section

23:參數輸入部 23: Parameter input section

24:資料顯示部 24: Data display department

25:最佳參數獲取部 25: Best parameter acquisition department

26:變化點位置計算部 26: Change point position calculation department

27:變化點位置提取部 27: Change point location extraction part

30:資料評估機構 30: Data Evaluation Agency

31:評估用資料獲取部 31: Evaluation Data Acquisition Department

32:雜訊去除部 32: Noise removal section

33:最佳參數輸入部 33: Optimal parameter input section

34:資料顯示部 34: Data display department

35:變化點位置獲取部 35: Change point position acquisition department

36:變化點位置計算部 36: Change point position calculation department

37:變化點位置提取部 37: Change point location extraction part

40:資料記憶部 40: Data Memory Department

41:計測資料記憶部 41: Measurement data storage unit

42:最佳參數記憶部 42: Best parameter memory

43:變化點位置資料記憶部 43: Change point location data memory

Claims (11)

一種支援裝置，其用以獲取確定感測器的計測值的經時變化之參數，前述支援裝置具備：資料獲取部，獲取表示第1感測器的計測值的經時變化之第1學習用資料和表示第2感測器的計測值的經時變化且與前述第1學習用資料具有輸入或輸出的響應關係之第2學習用資料；輸入部，輸入參數；及參數獲取部，依據與藉由前述輸入部輸入之參數對應之前述第1學習用資料的變化點位置及前述第2學習用資料的變化點位置而獲取預定值的參數。 A support device for acquiring a parameter that determines a change over time in a measured value of a sensor. The support device includes: a data acquisition unit that acquires a first learning device that indicates the change over time in the measured value of the first sensor The data and the second learning data representing the change over time of the measured value of the second sensor and having an input or output response relationship with the aforementioned first learning data; the input unit, the input parameter; and the parameter acquisition unit, based on and The parameter of the predetermined value is obtained by the position of the change point of the first learning data and the position of the change point of the second learning data corresponding to the parameters input by the input unit. 如申請專利範圍第1項所述之支援裝置，其中前述第1學習用資料為來自處理系統的既定位置處的溫度感測器之計測資料，前述第2學習用資料為來自檢測成為該既定位置的溫度變化的原因之事像之感測器的計測資料。 The support device described in the first item of the scope of patent application, wherein the first learning data is measured data from a temperature sensor at a predetermined position of the processing system, and the second learning data is derived from detection of the predetermined position The cause of the temperature change is like the measurement data of the sensor. 如申請專利範圍第1或2項所述之支援裝置，其進一步具備：資料顯示部，顯示前述第1學習用資料、前述第2學習用資料及參數的輸入欄。 The support device described in item 1 or 2 of the scope of patent application further includes: a data display unit that displays the first learning data, the second learning data, and input fields for parameters. 如申請專利範圍第3項所述之支援裝置，其中前述資料顯示部在藉由前述輸入部輸入參數之後進一步顯示與由前述輸入部輸入之參數對應之前述第1學習用資料的變化點位置及前述第2學習用資料的變化點位置和各變化點位置的一致率的判定結果。 The support device according to the third item of the scope of patent application, wherein the data display unit further displays the change point position and the change point of the first learning data corresponding to the parameter input by the input unit after the parameter is input by the input unit The result of the determination of the coincidence rate between the position of the change point of the second learning material and the position of each change point. 如申請專利範圍第1或2項所述之支援裝置，其中前述資料獲取部包括獲取2個以上的前述第1學習用資料之步驟，前述參數獲取部依據前述2個以上的第1學習用資料的變化點位置和前述第2學習用資料的變化點位置而獲取前述預定值的參數，前述2個以上的第1學習用資料的變化點位置係依據與由前述輸入部輸入之參數對應之前述2個以上的第1學習用資料而得。 For the support device described in item 1 or 2 of the scope of patent application, the data acquisition unit includes a step of acquiring two or more of the first learning materials, and the parameter acquisition unit is based on the two or more first learning materials. The change point position of the second learning material and the change point position of the second learning material to obtain the parameter of the predetermined value. The change point position of the two or more first learning materials is based on the aforementioned corresponding to the parameter input by the input unit Obtained from 2 or more first learning materials. 如申請專利範圍第1或2項所述之支援裝置，其進一步具備：變化點位置獲取部，依據由前述參數獲取部獲取之預定值的參數而獲取表示前述第1感測器的計測值的經時變化之評估用資料的變化點位置。 The support device described in item 1 or 2 of the scope of patent application, further comprising: a change point position acquiring unit that acquires a value representing the measured value of the first sensor based on the parameter of the predetermined value acquired by the parameter acquiring unit The location of the change point of the data used for the assessment of changes over time. 一種支援方法，其用以獲取確定感測器的計測值的經時變化之參數，前述支援方法包括如下步驟：藉由資料獲取部獲取表示第1感測器的計測值的經時 變化之第1學習用資料和表示第2感測器的計測值的經時變化且與前述第1學習用資料具有輸入或輸出的響應關係之第2學習用資料；藉由輸入部輸入參數；及依據與由前述輸入部輸入之參數對應之前述第1學習用資料的變化點位置和前述第2學習用資料的變化點位置，藉由參數獲取部獲取預定值的參數。 A support method for obtaining a parameter that determines the change over time of a measured value of a sensor. The aforementioned support method includes the following steps: acquiring, by a data acquisition unit, the elapsed time representing the measured value of the first sensor The changed first learning data and the second learning data representing the change over time of the measured value of the second sensor and having an input or output response relationship with the first learning data; input parameters through the input unit; And based on the change point position of the first learning data and the change point position of the second learning data corresponding to the parameters input by the input unit, the parameter acquisition unit acquires a parameter of a predetermined value. 如申請專利範圍第7項所述之支援方法，其進一步包括如下步驟：將前述第1學習用資料、前述第2學習用資料及參數的輸入欄顯示於資料顯示部；及在藉由前述輸入部輸入參數之後，藉由前述資料顯示部顯示與由前述輸入部輸入之參數對應之前述第1學習用資料的變化點位置及前述第2學習用資料的變化點位置和各變化點位置的一致率的判定結果。 For example, the support method described in item 7 of the scope of patent application further includes the steps of: displaying the input fields of the first learning data, the second learning data, and parameters on the data display part; After inputting the parameters, the data display unit displays the change point positions of the first learning data and the change point positions of the second learning data corresponding to the parameters input by the input unit and the coincidence of the change point positions Rate of the judgment result. 如申請專利範圍第8項所述之支援方法，其再次進行如下步驟：在前述一致率的判定結果表示不一致的情況下，輸入前述參數；及獲取前述預定值的參數。 For the support method described in item 8 of the scope of patent application, the following steps are performed again: in the case where the determination result of the aforementioned coincidence rate indicates inconsistency, the aforementioned parameter is input; and the aforementioned parameter of the predetermined value is obtained. 如申請專利範圍第8項所述之支援方法，其中 在前述一致率的判定結果表示一致的情況下，獲取前述預定值的參數作為最佳參數。 The support method as described in item 8 of the scope of patent application, in which In the case where the determination result of the aforementioned coincidence rate indicates agreement, the parameter of the aforementioned predetermined value is acquired as the optimal parameter. 一種記錄媒體，係儲存有支援程式者，為了獲取確定感測器的計測值的經時變化之參數而由電腦執行該支援程式，前述支援程式用以使前述電腦執行如下步驟：獲取表示第1感測器的計測值的經時變化之第1學習用資料和表示第2感測器的計測值的經時變化且與前述第1學習用資料具有輸入或輸出的響應關係之第2學習用資料；及依據與由輸入部輸入之參數對應之前述第1學習用資料的變化點位置和前述第2學習用資料的變化點位置，藉由參數獲取部獲取預定值的參數。 A recording medium that stores a support program. The support program is executed by a computer in order to obtain parameters that determine changes in the measured value of a sensor over time. The first learning data representing the change over time of the measured value of the sensor and the second learning data showing the change over time of the measured value of the second sensor and having an input or output response relationship with the aforementioned first learning data Data; and based on the change point position of the first learning data and the change point position of the second learning data corresponding to the parameters input by the input unit, the parameter acquiring unit acquires a parameter of a predetermined value.
TW108148512A 2019-01-31 2019-12-31 Support device, support method and recording medium TWI734330B (en)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
JP2019-016314 2019-01-31
JP2019016314 2019-01-31

Publications (2)

Publication Number Publication Date
TW202030567A TW202030567A (en) 2020-08-16
TWI734330B true TWI734330B (en) 2021-07-21

Family

ID=71840565

Family Applications (1)

Application Number Title Priority Date Filing Date
TW108148512A TWI734330B (en) 2019-01-31 2019-12-31 Support device, support method and recording medium

Country Status (4)

Country Link
JP (1) JP7449248B2 (en)
KR (1) KR20210124214A (en)
TW (1) TWI734330B (en)
WO (1) WO2020158466A1 (en)

Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPH117317A (en) * 1997-06-17 1999-01-12 Toshiba Corp Device for monitoring abnormality of power generating plant
CN1514209A (en) * 2003-08-01 2004-07-21 重庆大学 Rotary machine failure intelligent diagnosis method and device
US20120022350A1 (en) * 2010-07-21 2012-01-26 Streamline Automation, Llc Sensor fusion and probabilistic parameter estimation method and apparatus
JP2012141712A (en) * 2010-12-28 2012-07-26 Toshiba Corp Process monitoring diagnostic system
US20170079596A1 (en) * 2009-04-22 2017-03-23 Streamline Automation, Llc Probabilistic parameter estimation using fused data apparatus and method of use thereof
US20170277559A1 (en) * 2016-03-25 2017-09-28 Daqri, Llc Classifying work processes
WO2018132786A1 (en) * 2017-01-13 2018-07-19 Ground Truth Consulting System and method for predicting well production
WO2018200541A1 (en) * 2017-04-24 2018-11-01 Carnegie Mellon University Virtual sensor system

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP5953779B2 (en) 2012-02-01 2016-07-20 富士ゼロックス株式会社 Failure prediction system, failure prediction device, and program
JP6803241B2 (en) * 2017-01-13 2020-12-23 アズビル株式会社 Time series data processing device and processing method
JP6791892B2 (en) * 2018-01-18 2020-11-25 ファナック株式会社 Anomaly detection parameter adjustment display device

Patent Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPH117317A (en) * 1997-06-17 1999-01-12 Toshiba Corp Device for monitoring abnormality of power generating plant
CN1514209A (en) * 2003-08-01 2004-07-21 重庆大学 Rotary machine failure intelligent diagnosis method and device
US20170079596A1 (en) * 2009-04-22 2017-03-23 Streamline Automation, Llc Probabilistic parameter estimation using fused data apparatus and method of use thereof
US20120022350A1 (en) * 2010-07-21 2012-01-26 Streamline Automation, Llc Sensor fusion and probabilistic parameter estimation method and apparatus
JP2012141712A (en) * 2010-12-28 2012-07-26 Toshiba Corp Process monitoring diagnostic system
US20170277559A1 (en) * 2016-03-25 2017-09-28 Daqri, Llc Classifying work processes
WO2018132786A1 (en) * 2017-01-13 2018-07-19 Ground Truth Consulting System and method for predicting well production
WO2018200541A1 (en) * 2017-04-24 2018-11-01 Carnegie Mellon University Virtual sensor system

Also Published As

Publication number Publication date
JPWO2020158466A1 (en) 2021-12-02
TW202030567A (en) 2020-08-16
JP7449248B2 (en) 2024-03-13
KR20210124214A (en) 2021-10-14
WO2020158466A1 (en) 2020-08-06

Similar Documents

Publication Publication Date Title
CN112955839B (en) Abnormality detection device, abnormality detection method, and program
US10228994B2 (en) Information processing system, information processing method, and program
US20150276557A1 (en) State monitoring system, state monitoring method and medium
US20160231738A1 (en) Information processing apparatus and analysis method
US20170249559A1 (en) Apparatus and method for ensembles of kernel regression models
Goulding et al. Fault detection in continuous processes using multivariate statistical methods
KR101978569B1 (en) Apparatus and Method for Predicting Plant Data
TW201911074A (en) Information processing device and information processing method
US10955818B2 (en) System and method for extracting principal time series data
CN108369416B (en) Abnormality diagnosis system
JP7127305B2 (en) Information processing device, information processing method, program
JP4952688B2 (en) Sleep determination device and sleep determination method
CN112463646B (en) Sensor abnormity detection method and device
US20230213924A1 (en) Abnormal irregularity cause identifying device, abnormal irregularity cause identifying method, and abnormal irregularity cause identifying program
US11004002B2 (en) Information processing system, change point detection method, and recording medium
JP5949135B2 (en) Abnormality diagnosis method and abnormality diagnosis device
Zamba et al. A multivariate change-point model for change in mean vector and/or covariance structure
TWI734330B (en) Support device, support method and recording medium
JP2009110262A (en) Filter device, filtering method for filter device and filtering program
JP2006017467A (en) Odor specifying device
JP2016012263A (en) Abnormality diagnostic apparatus
Hu et al. Direct projection to latent variable space for fault detection
US20230229136A1 (en) Abnormal irregularity cause identifying device, abnormal irregularity cause identifying method, and abnormal irregularity cause identifying program
Tian et al. Statistical process control for multistage processes with non-repeating cyclic profiles
US20230213927A1 (en) Abnormal irregularity cause identifying device, abnormal irregularity cause identifying method, and abnormal irregularity cause identifying program