CN115374572B - Process stability analysis system and method - Google Patents

Process stability analysis system and method Download PDF

Info

Publication number
CN115374572B
CN115374572B CN202211291735.5A CN202211291735A CN115374572B CN 115374572 B CN115374572 B CN 115374572B CN 202211291735 A CN202211291735 A CN 202211291735A CN 115374572 B CN115374572 B CN 115374572B
Authority
CN
China
Prior art keywords
correlation
characteristic
stability
analysis
module
Prior art date
Legal status (The legal status 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 status listed.)
Active
Application number
CN202211291735.5A
Other languages
Chinese (zh)
Other versions
CN115374572A (en
Inventor
王三明
臧道志
魏蔚
赵伟帆
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Nanjing Safety Worry Free Network Technology Co ltd
Original Assignee
Nanjing Safety Worry Free Network Technology Co ltd
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 Nanjing Safety Worry Free Network Technology Co ltd filed Critical Nanjing Safety Worry Free Network Technology Co ltd
Priority to CN202211291735.5A priority Critical patent/CN115374572B/en
Publication of CN115374572A publication Critical patent/CN115374572A/en
Application granted granted Critical
Publication of CN115374572B publication Critical patent/CN115374572B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/10—Geometric CAD
    • G06F30/17—Mechanical parametric or variational design
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
    • G06F17/10—Complex mathematical operations
    • G06F17/18—Complex mathematical operations for evaluating statistical data, e.g. average values, frequency distributions, probability functions, regression analysis
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
    • G06F2119/02—Reliability analysis or reliability optimisation; Failure analysis, e.g. worst case scenario performance, failure mode and effects analysis [FMEA]
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
    • Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
    • Y02P90/30—Computing systems specially adapted for manufacturing

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Geometry (AREA)
  • Data Mining & Analysis (AREA)
  • Pure & Applied Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Mathematical Optimization (AREA)
  • Mathematical Analysis (AREA)
  • Computational Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • Computer Hardware Design (AREA)
  • Mathematical Physics (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Operations Research (AREA)
  • Probability & Statistics with Applications (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Algebra (AREA)
  • Evolutionary Biology (AREA)
  • Databases & Information Systems (AREA)
  • Software Systems (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention discloses a process stability analysis system and a process stability analysis method, which comprise an input module, a time fitting module, a cleaning module, an association analysis model module, a threshold value determining module, a stability model module and an output module, wherein when a model is established, relative uncertainty possibly exists between characteristic parameters and association parameters under different working conditions in the process of establishing the model. And dynamically analyzing the relation between the characteristic parameters and the associated parameters according to the parameter data of different working conditions, and calculating the associated coefficient. The method solves the one-sidedness of the current single-parameter point position alarm or stability rate evaluation, avoids the subjectivity of single-device stability rate calculation, and comprehensively evaluates the stability of the production process from the aspect of the surface.

Description

Process stability analysis system and method
Technical Field
The invention relates to a system and a method for analyzing process stability, belonging to the technical field of production process quality.
Background
In the process of production, in order to ensure that the process can be safely and stably operated for a long time, people install various sensors at the production device through the construction of systems such as SCADA, SIS, DCS and the like, and the sensors are used for detecting data such as temperature, pressure, flow, liquid level and the like, so that the real-time perception of the actual operation condition of the production device and the actual process parameter condition is realized, and the actual production condition on site is known in time.
Whether the production process is stable or not is the core for guaranteeing the safety and stability of production and guaranteeing the product quality, and whether the stability calculation of the production process is reasonable or not directly determines the scientificity and fineness of production management. In practice, the method of production stability calculation varies according to the variation of many factors such as process, equipment, monitoring point location, environment, and physical laws. The setting process of the production stationarity calculating method is an optimization process comprehensively considering global factors. The use and analysis methods set under certain specific conditions are often difficult to apply directly to other environments.
The analysis method of the stability of the current production process can be mainly divided into the following categories:
1) Statistical analysis type: and outputting various statistical analysis reports or reports according to the acquired data and the management requirements of enterprises, and judging the stability of the production process according to whether the data such as the output, unit consumption, equipment operation and the like summarized by the reports meet expected standards.
2) History view class: the method comprises the steps of putting production data in a warehouse, calling data at a certain moment or a certain time period in the history according to needs, performing comparative analysis or trend analysis, and judging the stability of the production process according to the historical comparative analysis and the trend of production parameters.
3) Real-time alarm class: the method is to establish a threshold range for monitored data, and when the monitored process parameters do not meet the threshold rule, automatically send out alarm information to indicate possible abnormal states of the production process. Whether an alarm is generated or not is used for judging the stability of the production process.
4) Stability calculation class: and (4) counting the abnormal data alarm times and time of the production process parameters, and calculating the stability rate of the single process parameter. And judging the stability of the whole production device through the stability rate of a single parameter.
The traditional method for analyzing the stability of the production process can be summarized into several categories, such as a statistical analysis category, a history viewing category, a real-time alarm category, a stability rate calculation category and the like. These methods of use and analysis can only reflect the actual statistical results of production and the current situation of single-parameter point location, but cannot reflect the overall situation of the production process. The disadvantages of each type of method are summarized below:
1) Statistical analysis class: the method generally records actual production process and equipment operation data at that time according to actual management requirements of enterprises, and then realizes daily management requirements by using a basic statistical analysis method. Although the method is simple and intuitive, the result data only can reflect the production result condition and does not represent whether the production process is stable or not.
2) History view class: theoretically, the method can judge whether the current process is stable or not through comparison analysis and trend analysis by checking multi-point real-time and historical data. The method is only theoretically feasible, has numerous parameters influencing the stability of the process and large time span of comparison, and is difficult to obtain valuable and reliable information through manpower analysis.
3) Real-time alarm class: the method automatically judges whether a single parameter point location is normal or not in a threshold value mode. And summarizing multi-parameter information so as to deduce whether the process of the production device is normal and stable. On one hand, however, the problem that whether the threshold setting is reasonable exists, the enterprise parameters are usually set according to the requirements of safety and product quality, and the threshold range is relatively wide; on the other hand, the method can only feed back whether the current concerned parameter point position is in a normal range, but cannot feed back whether the overall production process is stable.
4) Stationary rate calculation class: the method calculates the stability of the current parameter point location by a stability rate calculation method according to the alarm condition of a single parameter point location, and simultaneously calculates the stability of the device by a device stability rate calculation method. However, the stability rate of a single parameter only can feed back the stability of the current parameter and cannot feed back the stability of the whole device and even the whole production process; the method for calculating the stability rate of the device only weights different parameters according to experience to obtain the stability rate of the device, and the method cannot ensure the rationality under different processes and working conditions.
Disclosure of Invention
The purpose of the invention is as follows: in order to overcome the defects in the prior art, the invention provides a process stability analysis system and a method, which are not limited by specific production devices, processes and working conditions, have greater universality and can be flexibly adjusted according to actual production parameter conditions and different working condition conditions. The invention does not model the internal principle of the production process, thereby having the characteristics of simplicity, sensitivity and intellectualization and having excellent universality for different processes and different working conditions.
The technical scheme is as follows: in order to achieve the purpose, the invention adopts the technical scheme that:
a process stability analysis method comprises the following steps:
step 1, determining process characteristics according to process historical data.
And 2, determining the relevant characteristics related to the process characteristics.
And 3, performing time fitting on the time characteristic relation between the process characteristic and the associated characteristic according to the process historical data.
And 4, eliminating data which occur at times other than the process characteristic occurrence time point in the associated characteristic to obtain the cleaned associated characteristic.
And 5, establishing a correlation analysis model according to the cleaned correlation characteristics and the process characteristics, calculating correlation coefficients of the cleaned correlation characteristics and the process characteristics through the correlation analysis model, and sequencing the correlation coefficients to obtain strongly correlated correlation characteristics and process characteristics.
And 6, determining a threshold value of the strongly correlated associated characteristics and a threshold value of the process characteristics according to the process historical data.
And 7, establishing a stability model according to the strongly correlated correlation characteristic and the process characteristic, and the threshold value of the strongly correlated correlation characteristic and the threshold value of the process characteristic.
And 8, calculating a regression coefficient in the stability model by regression analysis by using the process historical data to obtain the solved stability model.
And 9, collecting the process data to be analyzed, performing relevance analysis on the process data to be analyzed through a relevance analysis model, and obtaining the stability analysis evaluation of the production process through a stability model.
Preferably: and (5) associating the analysis models:
Figure 486627DEST_PATH_IMAGE001
wherein the content of the first and second substances,
Figure 824067DEST_PATH_IMAGE002
the coefficient of the correlation is represented by,
Figure 264276DEST_PATH_IMAGE003
the number of samples is represented as a function of,
Figure 163224DEST_PATH_IMAGE004
,
Figure 423304DEST_PATH_IMAGE005
a single associated characteristic parameter is represented which,
Figure 830014DEST_PATH_IMAGE006
the ith sample value representing a single characteristic parameter,
Figure 124730DEST_PATH_IMAGE007
represents the average of a single sample of the characteristic parameter,
Figure 427535DEST_PATH_IMAGE008
,
Figure 956604DEST_PATH_IMAGE009
a parameter representing a characteristic of the object is,
Figure 901426DEST_PATH_IMAGE010
the ith sample value representing the target characteristic parameter,
Figure 785068DEST_PATH_IMAGE011
represents the average of the target feature parameter samples.
Through the above calculation formula, the correlation between each parameter and the liquid level parameter can be calculated. The value interval of the correlation is
Figure 258775DEST_PATH_IMAGE012
The larger the value, the greater the correlation.
Preferably: the smoothness model in step 7:
variance smoothness:
Figure 759026DEST_PATH_IMAGE013
linear regression analysis:
Figure 976381DEST_PATH_IMAGE014
total smoothness:
Figure 747153DEST_PATH_IMAGE015
wherein the content of the first and second substances,
Figure 657340DEST_PATH_IMAGE016
the degree of flatness of the variance is expressed,
Figure 379309DEST_PATH_IMAGE017
representing the number of valid operating values of a single characteristic parameter,
Figure 400354DEST_PATH_IMAGE018
the m-th sample value representing a single characteristic parameter,
Figure 524168DEST_PATH_IMAGE019
representing a single characteristic parameter
Figure 103792DEST_PATH_IMAGE020
Is determined by the average value of (a) of (b),
Figure 47477DEST_PATH_IMAGE021
representing the parameters of the target feature to be fitted,
Figure 872214DEST_PATH_IMAGE022
the number of features is represented as such,
Figure 116113DEST_PATH_IMAGE023
the regression coefficients representing the individual characteristic parameters,
Figure 368103DEST_PATH_IMAGE024
representing a single characteristic parameter
Figure 566128DEST_PATH_IMAGE025
The number of sample values is one,
Figure 194556DEST_PATH_IMAGE026
represents the intercept of a linear regression analysis,
Figure 27383DEST_PATH_IMAGE027
the overall degree of smoothness is expressed as,
Figure 715853DEST_PATH_IMAGE028
the mth sample value representing the fitted target feature parameter,
Figure 404104DEST_PATH_IMAGE029
representing the fitted target characteristic parameter average value;
single characteristic parameter
Figure 570643DEST_PATH_IMAGE030
Average value of (a):
Figure 789135DEST_PATH_IMAGE031
fitted target feature parameter mean:
Figure 382927DEST_PATH_IMAGE032
。
preferably: the cost function of the regression analysis in step 8 is:
Figure 54080DEST_PATH_IMAGE033
wherein, in the process,
Figure 260196DEST_PATH_IMAGE034
the value of the cost is expressed,
Figure 333194DEST_PATH_IMAGE035
representing fitted target feature parameters
Figure 97888DEST_PATH_IMAGE036
The number of sample values is one,
Figure 990757DEST_PATH_IMAGE037
the weight is represented by a weight that is,
Figure 499099DEST_PATH_IMAGE038
representing a single characteristic parameter
Figure 190718DEST_PATH_IMAGE039
The number of sample values is one,
Figure 126313DEST_PATH_IMAGE040
the intercept is represented as a function of the distance between the points,
Figure 506479DEST_PATH_IMAGE041
representing parameters
Figure 84091DEST_PATH_IMAGE042
The L1 norm of (i.e., the sum of the absolute values of the elements in the vector), is also a function representing distance,
Figure 367567DEST_PATH_IMAGE043
representing the mean square error, lambda represents the weight parameter,
Figure 474063DEST_PATH_IMAGE044
represents the trained coefficients, comprises weight coefficients and intercept terms, is a vector with the length of n +1,
Figure 341525DEST_PATH_IMAGE045
is shown as
Figure 191669DEST_PATH_IMAGE046
A coefficient.
Preferably: the process characteristics in step 1 include production process, product yield, and product quality.
Preferably, the following components: the associated characteristics in the step 2 comprise process parameters of related devices, input raw material parameters and component parameters of semi-finished products.
A process stability analysis system comprises an input module, a time fitting module, a cleaning module, an association analysis model module, a threshold value determining module, a stability model module and an output module, wherein:
the input module is used for process data to be analyzed, process characteristics and associated characteristics.
And the time fitting module is used for performing time fitting on the time characteristic relation between the process characteristic and the associated characteristic according to the process data to be analyzed.
And the cleaning module is used for removing data generated by time except the process characteristic generation time point in the associated characteristics to obtain the cleaned associated characteristics.
And the correlation analysis model module is used for calculating correlation coefficients of the cleaned correlation characteristics and the process characteristics through a correlation analysis model, and sequencing the correlation coefficients to obtain strongly correlated correlation characteristics and process characteristics.
The threshold determination module is used for determining a threshold of the strongly correlated associated feature and a threshold of the process feature according to the process historical data.
And the stability model module is used for inputting the strongly correlated correlation characteristics and the process characteristics, and the thresholds of the strongly correlated correlation characteristics and the process characteristics into the solved stability model to obtain the stability analysis evaluation of the production process.
And the output module is used for outputting the stability analysis evaluation of the production process.
Preferably, the following components: the correlation analysis model module is provided with the following correlation analysis models:
Figure 828187DEST_PATH_IMAGE047
wherein the content of the first and second substances,
Figure 863840DEST_PATH_IMAGE048
the coefficient of the correlation is represented by,
Figure 218597DEST_PATH_IMAGE049
the number of samples is represented as a function of,
Figure 606853DEST_PATH_IMAGE050
,
Figure 363457DEST_PATH_IMAGE051
a single associated characteristic parameter is represented which,
Figure 811756DEST_PATH_IMAGE052
the ith sample value representing a single characteristic parameter,
Figure 889695DEST_PATH_IMAGE053
represents the average of a single sample of the characteristic parameter,
Figure 81642DEST_PATH_IMAGE054
,
Figure 692752DEST_PATH_IMAGE055
a parameter representing a characteristic of the object is,
Figure 577532DEST_PATH_IMAGE056
the ith sample value representing the target characteristic parameter,
Figure 641303DEST_PATH_IMAGE057
represents the average of the target characteristic parameter samples.
Preferably: the stability model module is provided with the following stability models:
variance smoothness:
Figure 135476DEST_PATH_IMAGE058
linear regression analysis:
Figure 601092DEST_PATH_IMAGE059
total smoothness:
Figure 391194DEST_PATH_IMAGE060
wherein the content of the first and second substances,
Figure 942261DEST_PATH_IMAGE061
the degree of flatness of the variance is expressed,
Figure 7169DEST_PATH_IMAGE062
representing the number of valid operating values of a single characteristic parameter,
Figure 828756DEST_PATH_IMAGE063
the m-th sample value representing a single characteristic parameter,
Figure 789759DEST_PATH_IMAGE064
representing a single characteristic parameter
Figure 828122DEST_PATH_IMAGE065
Is determined by the average value of (a),
Figure 165563DEST_PATH_IMAGE066
representing the parameters of the target feature to be fitted,
Figure 340192DEST_PATH_IMAGE067
the number of features is indicated as such,
Figure 242070DEST_PATH_IMAGE068
the regression coefficients representing the individual characteristic parameters,
Figure 767729DEST_PATH_IMAGE069
representing a single characteristic parameter
Figure 908860DEST_PATH_IMAGE070
The number of sample values is one,
Figure 937996DEST_PATH_IMAGE071
representing linear regression analysisThe intercept of (a) is calculated,
Figure 506381DEST_PATH_IMAGE072
the overall degree of smoothness is expressed as,
Figure 755222DEST_PATH_IMAGE073
the mth sample value representing the fitted target feature parameter,
Figure 434465DEST_PATH_IMAGE074
representing the fitted target feature parameter mean.
Preferably: the correlation analysis model module is provided with the following regression analysis cost function:
Figure 849266DEST_PATH_IMAGE075
wherein the content of the first and second substances,
Figure 588552DEST_PATH_IMAGE076
the value of the cost is expressed,
Figure 823224DEST_PATH_IMAGE077
second representing fitted target characteristic parameters
Figure 70272DEST_PATH_IMAGE078
The number of sample values is one,
Figure 74000DEST_PATH_IMAGE079
the weight is represented by a weight that is,
Figure 984187DEST_PATH_IMAGE080
representing a single characteristic parameter
Figure 706156DEST_PATH_IMAGE081
The number of sample values is one,
Figure 992781DEST_PATH_IMAGE082
the intercept is represented as a function of the distance between the points,
Figure 618059DEST_PATH_IMAGE083
representing parameters
Figure 433569DEST_PATH_IMAGE084
The norm of L1 of (a),
Figure 908412DEST_PATH_IMAGE085
representing the mean square error, lambda represents the weight parameter,
Figure 467570DEST_PATH_IMAGE086
the coefficients that are trained are represented by the coefficients,
Figure 711469DEST_PATH_IMAGE087
denotes the first
Figure 745151DEST_PATH_IMAGE081
A coefficient.
Compared with the prior art, the invention has the following beneficial effects:
1) Feature parameter selection on demand
When the model is built, the characteristic parameters can be flexibly selected according to the actual process condition and the management requirement, and can be a specific parameter or a mixed parameter formed by fusing a plurality of parameters. In the process, the characteristic parameters may even be changed as required.
2) Strongly associated parameter dynamic update
When the model is established, the characteristic parameters and the associated parameters under different working conditions are considered to have relative uncertainty. The method is different from the traditional method in fixed parameters and fixed weight, the relation between the characteristic parameters and the correlation parameters is dynamically analyzed through an analysis model according to parameter data of different working conditions, the correlation coefficient is calculated, the larger the correlation coefficient is, the stronger the correlation is, and the strong correlation parameters are further determined. Therefore, the method plays a role in dynamic ordering of relevance of the associated parameters and provides guarantee for subsequent further processing.
Drawings
FIG. 1 is a flow chart of the present invention.
Detailed Description
The present invention is further illustrated by the following description in conjunction with the accompanying drawings and the specific embodiments, it is to be understood that these examples are given solely for the purpose of illustration and are not intended as a definition of the limits of the invention, since various equivalent modifications will occur to those skilled in the art upon reading the present invention and fall within the limits of the appended claims.
A process stability analysis method comprises the following steps:
step 1, determining process characteristics according to process historical data.
The process characteristics may be production process, product yield, product quality, etc., and the determination of the process characteristics is specified by the relevant process personnel.
The determination of process characteristics can have a significant impact on the results and these parameters are highly correlated with the process attributes, importance and sensitivity of the monitored quantities themselves. The method can adapt to different scenes, and the characteristic parameters can be flexibly adjusted according to actual conditions. The general selection rules for these parameters are described below.
(1) Based on production device target determination
Different production devices have unique work tasks, and characteristic parameters are determined according to the principle that the task is executed most safely, most efficiently and most energy-saving and the influence on the production task result is the maximum. Example (c): the ammonia still is the operation equipment that makes the aqueous ammonia in the preceding production volatilize the release through the heat transfer of heat carrier, and the main focus is the efficiency and the energy consumption of evaporating ammonia. The change in the level of the ammonia still (efficiency, energy consumption) can be taken as a characteristic value.
(2) Determination from mixing parameters
Similar to the stability calculation logic, a plurality of parameters related to the process are combined by different weights through different influence factors, and the assignment rule of the combination parameters can be assigned by a process person.
And 2, determining the relevant characteristics related to the process characteristics.
The related characteristics are the process parameters of the current device, or the process parameters of the related device, even the composition parameters (quality data) of the related input raw materials and semi-finished products of the production device.
The correlation analysis model is used for automatically performing correlation analysis on the process parameters of the analysis object (corresponding device) in the determination of the correlation characteristics. The analyst can also bring the parameters (or specific parameters) and indexes (quality indexes and the like) of the relevant devices into the associated parameter analysis pool to participate in analysis and calculation.
And 3, performing time fitting on the time characteristic relation between the process characteristic and the associated characteristic according to the process historical data.
And 4, eliminating data generated at times except the process characteristic generation time point in the correlation characteristics to obtain the cleaned correlation characteristics.
And 5, establishing a correlation analysis model according to the cleaned correlation characteristics and the process characteristics, calculating correlation coefficients of the cleaned correlation characteristics and the process characteristics through the correlation analysis model, and sequencing the correlation coefficients to obtain strongly correlated correlation characteristics and process characteristics.
And (3) determining a correlation analysis model:
(1) historical data capture
a. Butt-jointing with related SCADA, SIS and DCS systems, and acquiring production process data in real time (if necessary, butt-jointing with a quality system and an energy system to acquire quality data and energy consumption data);
b. capturing historical data according to the determined characteristic parameters and the determined associated parameters;
c. and removing invalid data, and preprocessing the historical data (removing null values) to ensure the validity of each piece of data.
(2) Data cleansing
The purpose of data cleaning is to train the service analysis model and arrange more desirable data for the service analysis model. The analysis model training is to find out the correlation between the characteristic parameters and the correlation parameters, and certain specific characteristics must exist between the parameters correlated with the core parameters. Therefore, data that is "almost unchanged" is unlikely to affect a characteristic parameter that changes greatly. When data is washed, the associated parameters of the data which are almost unchanged need to be washed.
(3) Time stamp processing
And (4) relevance analysis, wherein relevance is analyzed when the characteristic parameters and the relevance parameters occur simultaneously. But data acquisition is usually performed by polling, so that there may be a certain time difference between the characteristic parameter and the associated parameter. At this time, time fitting needs to be performed according to the time characteristic relationship between the characteristic parameters and the associated parameters, and the numerical values of the characteristic parameters and the numerical values of the associated parameters are analyzed at the same time point.
Figure 176133DEST_PATH_IMAGE088
(4) Data re-cleansing
Since the actual correlation between the feature parameter and the correlation parameter needs to be analyzed, data of the correlation parameter actually existing at the occurrence time point of the feature parameter needs to be analyzed. At this time, data occurring at times other than the characteristic parameter occurrence time point in the associated parameters needs to be eliminated.
(5) Strong correlation parameter determination
And analyzing the relevance between the characteristic parameters and the relevance parameters by utilizing the spearman correlation coefficient and the significance test method thereof.
Suppose there are two vectors
Figure 804560DEST_PATH_IMAGE089
And
Figure 902966DEST_PATH_IMAGE090
all length being
Figure 325857DEST_PATH_IMAGE091
The correlation analysis model is as follows:
Figure 11179DEST_PATH_IMAGE092
wherein the content of the first and second substances,
Figure 177718DEST_PATH_IMAGE093
the coefficient of the correlation is represented by,
Figure 130630DEST_PATH_IMAGE094
the number of samples is represented as a function of,
Figure 724423DEST_PATH_IMAGE095
,
Figure 894111DEST_PATH_IMAGE096
a single associated characteristic parameter is represented which,
Figure 598762DEST_PATH_IMAGE097
the ith sample value representing a single characteristic parameter,
Figure 671760DEST_PATH_IMAGE098
represents the average of a single sample of the characteristic parameter,
Figure 436453DEST_PATH_IMAGE099
,
Figure 329323DEST_PATH_IMAGE100
a parameter representing a characteristic of the object is,
Figure 604709DEST_PATH_IMAGE101
the ith sample value representing the target characteristic parameter,
Figure 266634DEST_PATH_IMAGE102
represents the average of the target characteristic parameter samples.
Through the calculation formula, the correlation between each correlation parameter and the characteristic parameter can be calculated, and the value interval of the correlation is
Figure 202229DEST_PATH_IMAGE103
The larger the value, the greater the correlation.
And sequencing the correlation to obtain the parameters and the correlation coefficients with strong correlation.
And 6, determining the threshold value of the strongly correlated associated characteristic and the threshold value of the process characteristic according to the process historical data.
And determining the process parameters (indexes) strongly related to the characteristic parameters by using a data mining method according to the long-term historical data of the process parameters. And outputting the strongly relevant process parameters and the threshold range of the strongly relevant parameters in the process stable state.
And analyzing the data of the device production process, and finding out the data difference between the normal working state and other states of the device. For example, when the data of the ammonia still is analyzed, the sudden drop of the liquid level from 1200 to 0 is caused by the wiring and pulling of the instrument, the rising of the liquid level is caused by the raw material addition of the device, and the gradual drop of the liquid level from 600 is caused by the normal ammonia distillation behavior. According to the characteristic, effective operation values of all parameters in the normal ammonia distillation process are found out, and then the minimum value and the maximum value of the effective operation values, namely the interval of the relevant parameters, are calculated.
And 7, establishing a stability model according to the strongly correlated correlation characteristic and the process characteristic, and the threshold value of the strongly correlated correlation characteristic and the threshold value of the process characteristic.
The stability model is as follows:
variance smoothness:
Figure 582395DEST_PATH_IMAGE104
linear regression analysis:
Figure 160007DEST_PATH_IMAGE105
total smoothness:
Figure 180834DEST_PATH_IMAGE106
wherein the content of the first and second substances,
Figure 287330DEST_PATH_IMAGE107
the degree of flatness of the variance is expressed,
Figure 154792DEST_PATH_IMAGE108
representing the number of valid operating values of a single characteristic parameter,
Figure 4936DEST_PATH_IMAGE109
the m-th sample value representing a single characteristic parameter,
Figure 907033DEST_PATH_IMAGE110
representing a single characteristic parameter
Figure 685895DEST_PATH_IMAGE111
Is determined by the average value of (a) of (b),
Figure 306233DEST_PATH_IMAGE112
representing the parameters of the target feature to be fitted,
Figure 694489DEST_PATH_IMAGE113
the number of features is indicated as such,
Figure 185513DEST_PATH_IMAGE114
the regression coefficients representing the individual characteristic parameters,
Figure 899391DEST_PATH_IMAGE115
representing a single characteristic parameter
Figure 974401DEST_PATH_IMAGE116
The number of sample values is one,
Figure 166348DEST_PATH_IMAGE117
represents the intercept of a linear regression analysis,
Figure 43037DEST_PATH_IMAGE118
the overall degree of smoothness is expressed as,
Figure 927816DEST_PATH_IMAGE119
the mth sample value representing the fitted target feature parameter,
Figure 991587DEST_PATH_IMAGE120
representing the fitted target characteristic parameter mean.
Single characteristic parameter
Figure 488690DEST_PATH_IMAGE121
Average value of (a):
Figure 219886DEST_PATH_IMAGE122
fitted target feature parameter mean:
Figure 9987DEST_PATH_IMAGE123
the stability is the stability of each index and control parameter in the running process of the device, and is a quantitative index of the operation stability of the device. The total smoothness is the smoothness of the regression results for each parameter. The variance is used to calculate the smoothness, also called variance smoothness. The closer the index data distribution is to the average value, the smaller the variance is; the further the data distribution is from the mean, the greater the variance. The smaller the variance smoothness, the better, indicating a higher level of operation. And calculating the process stability of the device according to the collected characteristic parameters and the process production data of the associated parameters and the associated parameters by combining the associated parameters and coefficients analyzed in the early stage.
Such as: when the temperature is selected as the index, the formula is shown in
Figure 561054DEST_PATH_IMAGE124
The temperature is stable, and the pressure is selected as an index in the formula
Figure 853059DEST_PATH_IMAGE124
When the total stability is considered, the index is calculated by regression analysis
Figure 438761DEST_PATH_IMAGE125
And each index
Figure 665343DEST_PATH_IMAGE126
The regression coefficient between them is calculated according to the formula (2), and the total stability can be obtained.
And establishing a stability calculation model by utilizing the strongly correlated correlation characteristics and the process characteristics, and obtaining a threshold range of the stability according to the threshold of the strongly correlated correlation characteristics and the threshold of the process characteristics. And calculating the stability of the current production process according to the value of the latest strongly-relevant parameter acquired in real time. When the smoothness exceeds the threshold range, an alarm is automatically generated and pushed to related personnel.
And 8, calculating a regression coefficient in the stability model by regression analysis by using the process historical data to obtain the solved stability model.
The cost function of the regression analysis is:
Figure 703706DEST_PATH_IMAGE127
wherein the content of the first and second substances,
Figure 775567DEST_PATH_IMAGE128
the value of the cost is expressed,
Figure 717241DEST_PATH_IMAGE129
second representing fitted target characteristic parameters
Figure 849145DEST_PATH_IMAGE130
The number of sample values is one,
Figure 640383DEST_PATH_IMAGE131
the weight is represented by a weight that is,
Figure 515935DEST_PATH_IMAGE132
representing a single characteristic parameter
Figure 810650DEST_PATH_IMAGE133
The number of sample values is one,
Figure 877570DEST_PATH_IMAGE134
the intercept is represented by the sum of the intercept values,
Figure 890525DEST_PATH_IMAGE135
representing parameters
Figure 569768DEST_PATH_IMAGE136
The L1 norm of (i.e., the sum of the absolute values of the elements in the vector), is also a function representing distance,
Figure 718990DEST_PATH_IMAGE137
representing the mean square error, lambda represents the weight parameter,
Figure 723855DEST_PATH_IMAGE138
represents the trained coefficients, comprises weight coefficients and intercept terms, is a vector with the length of n +1,
Figure 459992DEST_PATH_IMAGE139
denotes the first
Figure 677347DEST_PATH_IMAGE140
A coefficient of the number of the elements.
And according to a stability calculation formula, automatically calculating the stability in real time according to the characteristic parameters and the real-time acquisition values of the associated parameters, and providing information such as real-time values, threshold values, difference values of the real-time values and the threshold values of the associated parameters.
And 9, collecting the process data to be analyzed, performing relevance analysis on the process data to be analyzed through a relevance analysis model, and obtaining the stability analysis evaluation of the production process through a stability model.
And automatically optimizing the stability correlation analysis model and the stability calculation model according to the latest acquired process parameter data. When the working condition changes, the stability correlation analysis model and the stability calculation model are automatically optimized according to the latest working condition, so that the model can be self-learned, self-grown and self-adaptive. And starting different smoothness models according to the characteristics of different working conditions.
When the working condition characteristics change, the correlation analysis model and the stability calculation model can periodically, automatically and/or manually and actively perform correlation analysis recalculation according to the collected related point location data, redefine correlation parameters, correlation coefficients and thresholds after recalculation, and perform characteristic marking on different working conditions. And automatically selecting the model corresponding to the working condition according to the characteristics of different working conditions at the later stage.
A process stability analysis system comprises an input module, a time fitting module, a cleaning module, an association analysis model module, a threshold value determining module, a stability model module and an output module, wherein:
the input module is used for process data to be analyzed, process characteristics and associated characteristics.
And the time fitting module is used for performing time fitting on the time characteristic relation between the process characteristic and the associated characteristic according to the process data to be analyzed.
And the cleaning module is used for removing data which occur at a time other than the process characteristic occurrence time point in the associated characteristic to obtain the cleaned associated characteristic.
The correlation analysis model module is provided with a correlation analysis model. The correlation analysis model module is provided with a cost function of regression analysis. And the correlation analysis module is used for calculating correlation coefficients of the cleaned correlation characteristics and the process characteristics and sequencing the correlation coefficients to obtain strongly correlated correlation characteristics and process characteristics.
The threshold determination module is used for determining a threshold of the strongly correlated associated feature and a threshold of the process feature according to the process historical data.
And the stability model module is provided with a stability model and is used for inputting the strongly correlated correlation characteristics and the process characteristics, and the thresholds of the strongly correlated correlation characteristics and the process characteristics into the solved stability model to obtain the stability analysis evaluation of the production process.
And the output module is used for outputting the stability analysis evaluation of the production process.
The method automatically performs relevance analysis on the process characteristic parameters and the relevant parameters thereof, and finally obtains the stability analysis evaluation of the production process through stability rate calculation.
(1) The method solves the one-sidedness of the current single-parameter point position alarm or stability rate evaluation, avoids the subjectivity of single-device stability rate calculation, and comprehensively evaluates the stability of the production process from the aspect of the surface;
(2) the self-learning, self-growing and self-adapting mechanism under different working conditions is provided, and the flexibility and the accuracy under the variable working conditions are realized;
(3) the method is weakly related to the logic of the process bottom layer, can be universally used for different production processes (devices), and has better universality.
The above description is only of the preferred embodiments of the present invention, and it should be noted that: it will be apparent to those skilled in the art that various modifications and adaptations can be made without departing from the principles of the invention and these are intended to be within the scope of the invention.

Claims (4)

1. A process stability analysis method is characterized by comprising the following steps:
step 1, determining process characteristics according to process historical data;
step 2, determining relevant characteristics related to the process characteristics;
step 3, performing time fitting on the time characteristic relation between the process characteristic and the associated characteristic according to the process historical data;
step 4, eliminating data generated in the time except the time point of occurrence of the process characteristic in the correlation characteristic to obtain the cleaned correlation characteristic;
step 5, establishing a correlation analysis model according to the cleaned correlation characteristics and the process characteristics, calculating correlation coefficients of the cleaned correlation characteristics and the process characteristics through the correlation analysis model, and sequencing the correlation coefficients to obtain strongly correlated correlation characteristics and process characteristics;
and (3) correlation analysis model:
Figure 738777DEST_PATH_IMAGE001
wherein the content of the first and second substances,
Figure 107441DEST_PATH_IMAGE002
the correlation coefficient is represented by a correlation coefficient,
Figure 532607DEST_PATH_IMAGE003
the number of samples is represented as a function of,
Figure 303117DEST_PATH_IMAGE004
,
Figure 257166DEST_PATH_IMAGE005
a single associated characteristic parameter is represented which,
Figure 706602DEST_PATH_IMAGE006
the ith sample value representing a single characteristic parameter,
Figure 873141DEST_PATH_IMAGE007
represents the average of a single sample of the characteristic parameter,
Figure 232578DEST_PATH_IMAGE008
,
Figure 888687DEST_PATH_IMAGE009
a parameter representing a characteristic of the object is,
Figure 966365DEST_PATH_IMAGE010
the ith sample value representing the target characteristic parameter,
Figure 936595DEST_PATH_IMAGE011
an average value representing a sample of the target characteristic parameter;
step 6, determining a threshold value of strongly related associated characteristics and a threshold value of process characteristics according to process historical data;
step 7, establishing a stability model according to the strongly correlated correlation characteristics and the process characteristics, and the thresholds of the strongly correlated correlation characteristics and the process characteristics;
variance smoothness:
Figure 806331DEST_PATH_IMAGE012
linear regression analysis:
Figure 243128DEST_PATH_IMAGE013
total smoothness:
Figure 932736DEST_PATH_IMAGE014
wherein, the first and the second end of the pipe are connected with each other,
Figure 730094DEST_PATH_IMAGE015
the degree of flatness of the variance is expressed,
Figure 392020DEST_PATH_IMAGE016
representing the number of valid operating values of a single characteristic parameter,
Figure 999719DEST_PATH_IMAGE017
the m-th sample value representing a single characteristic parameter,
Figure 176622DEST_PATH_IMAGE018
representing a single characteristic parameter
Figure 895179DEST_PATH_IMAGE019
Is determined by the average value of (a) of (b),
Figure 739507DEST_PATH_IMAGE020
representing the parameters of the target feature to be fitted,
Figure 518108DEST_PATH_IMAGE021
the number of features is indicated as such,
Figure 182307DEST_PATH_IMAGE022
the regression coefficients representing the individual characteristic parameters,
Figure 704555DEST_PATH_IMAGE023
representing a single characteristic parameter
Figure 341073DEST_PATH_IMAGE024
The number of sample values is one,
Figure 415208DEST_PATH_IMAGE025
represents the intercept of a linear regression analysis,
Figure 442070DEST_PATH_IMAGE026
the overall degree of smoothness is expressed as,
Figure 892643DEST_PATH_IMAGE027
the m-th sample value representing the fitted target feature parameter,
Figure 55771DEST_PATH_IMAGE028
representing the fitted target characteristic parameter average value;
step 8, calculating a regression coefficient in the stability model by regression analysis by utilizing the process historical data to obtain a well-solved stability model;
the cost function of the regression analysis is:
Figure 504070DEST_PATH_IMAGE029
wherein the content of the first and second substances,
Figure 142862DEST_PATH_IMAGE030
the value of the cost is expressed,
Figure 6912DEST_PATH_IMAGE031
representing fitted target feature parameters
Figure 149181DEST_PATH_IMAGE032
The number of sample values is one,
Figure 971643DEST_PATH_IMAGE033
the weight is represented by a weight that is,
Figure 832152DEST_PATH_IMAGE034
representing a single characteristic parameter
Figure 827790DEST_PATH_IMAGE035
The number of sample values is one,
Figure 699931DEST_PATH_IMAGE036
represents the intercept of the cost function of the regression analysis,
Figure 552349DEST_PATH_IMAGE037
representing parameters
Figure 775520DEST_PATH_IMAGE038
The norm of L1 of (a) is,
Figure 371587DEST_PATH_IMAGE039
denotes a mean square error, lambda denotes a weight parameter,
Figure 363813DEST_PATH_IMAGE040
the coefficients that are trained are represented by the coefficients,
Figure 387133DEST_PATH_IMAGE041
is shown as
Figure 159917DEST_PATH_IMAGE042
A coefficient;
and 9, collecting the process data to be analyzed, performing relevance analysis on the process data to be analyzed through a relevance analysis model, and obtaining the stability analysis evaluation of the production process through a stability model.
2. The process smoothness analysis method of claim 1, wherein: the process characteristics in step 1 include production process, product yield, and product quality.
3. The process smoothness analysis method of claim 2, wherein: the correlation characteristics in the step 2 comprise process parameters of related devices, input raw material parameters and component parameters of semi-finished products.
4. An analysis system based on the process stability analysis method of claim 1, characterized in that: the device comprises an input module, a time fitting module, a cleaning module, an association analysis model module, a threshold value determining module, a stability model module and an output module, wherein:
the input module is used for analyzing process data, process characteristics and associated characteristics;
the time fitting module is used for performing time fitting on the time characteristic relation between the process characteristic and the associated characteristic according to the process data to be analyzed;
the cleaning module is used for removing data generated in the time except the process characteristic generation time point in the correlation characteristics to obtain cleaned correlation characteristics;
the correlation analysis model module is used for calculating correlation coefficients of the washed correlation characteristics and the washed process characteristics through a correlation analysis model, and sequencing the correlation coefficients to obtain strongly correlated correlation characteristics and process characteristics;
the threshold determination module is used for determining a threshold of a strongly correlated associated characteristic and a threshold of a process characteristic according to process historical data;
the stability model module is used for inputting the strongly correlated correlation characteristics and the process characteristics, and the thresholds of the strongly correlated correlation characteristics and the process characteristics into the solved stability model to obtain the stability analysis evaluation of the production process;
and the output module is used for outputting the stability analysis evaluation of the production process.
CN202211291735.5A 2022-10-21 2022-10-21 Process stability analysis system and method Active CN115374572B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202211291735.5A CN115374572B (en) 2022-10-21 2022-10-21 Process stability analysis system and method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202211291735.5A CN115374572B (en) 2022-10-21 2022-10-21 Process stability analysis system and method

Publications (2)

Publication Number Publication Date
CN115374572A CN115374572A (en) 2022-11-22
CN115374572B true CN115374572B (en) 2023-02-03

Family

ID=84074047

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202211291735.5A Active CN115374572B (en) 2022-10-21 2022-10-21 Process stability analysis system and method

Country Status (1)

Country Link
CN (1) CN115374572B (en)

Family Cites Families (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109991909A (en) * 2017-12-29 2019-07-09 北京桑德环境工程有限公司 The wisdom progress control method of scattering sewage treatment factory
CN111125082B (en) * 2019-12-26 2023-09-22 北京工业大学 Cement process parameter data analysis method based on association rule mining
CN113919559A (en) * 2021-09-28 2022-01-11 南京南瑞继保电气有限公司 Ultra-short-term prediction method and device for equipment parameters of comprehensive energy system
CN114416707A (en) * 2021-12-10 2022-04-29 浙江蓝卓工业互联网信息技术有限公司 Method and device for automated feature engineering of industrial time series data

Also Published As

Publication number Publication date
CN115374572A (en) 2022-11-22

Similar Documents

Publication Publication Date Title
CN113255795B (en) Equipment state monitoring method based on multi-index cluster analysis
CN110320892B (en) Sewage treatment equipment fault diagnosis system and method based on L asso regression
CN101470426B (en) Fault detection method and system
CN117235649B (en) Industrial equipment state intelligent monitoring system and method based on big data
KR20170123332A (en) Data cleansing system and method for inferring feed composition
Franco et al. Economic design of Shewhart control charts for monitoring autocorrelated data with skip sampling strategies
JP4046309B2 (en) Plant monitoring device
CN110599020B (en) Method and system for evaluating running state of generator set and readable storage medium
CN101403915A (en) Product quality prediction technique for recombination assembly line
CN115640860B (en) Electromechanical equipment remote maintenance method and system for industrial cloud service
CN112415947A (en) CNC machine tool data acquisition and management method and system based on DTU equipment
CN117394337A (en) Power grid load early warning method and system thereof
CN108305002A (en) A kind of decreasing failure rate height recognition methods based on monitoring Weibull form parameter
CN117556366B (en) Data abnormality detection system and method based on data screening
CN115374572B (en) Process stability analysis system and method
CN116991146B (en) Control method and system of ultrasonic cleaning device
CN113934536A (en) Data acquisition method facing edge calculation
CN116308293B (en) Intelligent agricultural equipment management system and method based on digital platform
CN115034094B (en) Prediction method and system for operation state of metal processing machine tool
CN110517731A (en) Genetic test quality monitoring data processing method and system
RU2470352C1 (en) Statistical process control method (versions)
CN114819743B (en) Energy consumption diagnosis and analysis method for chemical enterprises
CN116720091A (en) Motor fault analysis method
CN117455460A (en) Equipment predictive maintenance equipment and method based on embedded intelligent board card
Li et al. Data-Driven Anomaly Detection and Early Warning Issues

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant