CN114862064B - Systematic failure prediction method and system for hydraulic forging press - Google Patents

Systematic failure prediction method and system for hydraulic forging press Download PDF

Info

Publication number
CN114862064B
CN114862064B CN202210780611.7A CN202210780611A CN114862064B CN 114862064 B CN114862064 B CN 114862064B CN 202210780611 A CN202210780611 A CN 202210780611A CN 114862064 B CN114862064 B CN 114862064B
Authority
CN
China
Prior art keywords
failure
early warning
time
preset
variable
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
CN202210780611.7A
Other languages
Chinese (zh)
Other versions
CN114862064A (en
Inventor
计鑫
赵华
张胜
潘高峰
王鑫
赵欢
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.)
Tianjin Tianduan Press Group Co ltd
Original Assignee
Tianjin Tianduan Press Group 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 Tianjin Tianduan Press Group Co ltd filed Critical Tianjin Tianduan Press Group Co ltd
Priority to CN202210780611.7A priority Critical patent/CN114862064B/en
Publication of CN114862064A publication Critical patent/CN114862064A/en
Application granted granted Critical
Publication of CN114862064B publication Critical patent/CN114862064B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
    • 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
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G08—SIGNALLING
    • G08B—SIGNALLING OR CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS
    • G08B21/00—Alarms responsive to a single specified undesired or abnormal condition and not otherwise provided for
    • G08B21/18—Status alarms
    • G08B21/182—Level alarms, e.g. alarms responsive to variables exceeding a threshold
    • 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
    • Y02P90/80—Management or planning

Abstract

The invention discloses a systematic failure prediction method and a systematic failure prediction system for a forging hydraulic press, which belong to the technical field of predictive maintenance of the forging hydraulic press and comprise the following steps: 1. collecting state data of a hydraulic forging press; 2. setting a preset early warning rule and a preset failure rule of the predictive maintenance system; 3. after triggering of a preset early warning rule, receiving state data of each forging hydraulic press, and taking the state data as an interpretation variable of a regression analysis training sample; 4. recording the time difference between alarm failure and early warning, taking the estimated failure time as the actual failure time value of a regression analysis training sample, and taking the estimated failure time of the regression analysis as an interpreted variable; 5. correcting the explained variable according to the weight; 6. triggering a preset early warning rule and giving preventive maintenance measures after a fault occurs; 7. if preventive maintenance measures are taken, the early warning is reset, and if the preventive maintenance measures are not taken, the preset failure rule triggers the system to fail.

Description

Systematic failure prediction method and system for hydraulic forging press
Technical Field
The invention belongs to the technical field of predictive maintenance of hydraulic forging presses, and particularly relates to a systematic failure prediction method and a systematic failure prediction system for a hydraulic forging press.
Background
The part produced by the hydraulic forging press can be widely applied to the fields of aerospace, nuclear power, automobiles, ocean engineering and the like, and plays an important role in guaranteeing national economy and national defense safety. The economic loss caused by equipment downtime is huge, and equipment health guarantee is important and urgent. Therefore, the method and the system for predicting the systematic failure of the hydraulic forging press are designed and developed, and the method and the system are of great significance for judging the systematic failure time of equipment and taking corresponding preventive measures before failure.
Disclosure of Invention
The invention provides a systematic failure prediction method and a systematic failure prediction system for a forging hydraulic press, aiming at solving the technical problems in the prior art. A certain hydraulic forging press can obtain an estimated failure time value according to the state data and the regression function of the hydraulic forging press and provide preventive maintenance measures for users.
It is a first object of the present invention to provide a systematic failure prediction system for a forging press, comprising:
step one, collecting state data of a hydraulic forging press;
step two, setting a preset early warning rule and a preset failure rule of the predictive maintenance system;
step three, after triggering of a preset early warning rule, receiving state data of each forging hydraulic press, and taking the state data as an interpretation variable x of a regression analysis training sample;
recording the time difference between alarm failure and early warning, and defining the time difference as estimated failure time; taking the estimated failure time as the failure time actual value y of a regression analysis training sample, taking the estimated failure time of the regression analysis as an interpreted variable z, wherein the interpreted variable z of the training set has an expression as follows:
Figure 551817DEST_PATH_IMAGE001
；
wherein:
Figure 84429DEST_PATH_IMAGE002
，
Figure 17750DEST_PATH_IMAGE003
，……
Figure 952208DEST_PATH_IMAGE004
for the explanation variables, n is the number of the explanation variables, and n is more than or equal to 2;
Figure 476730DEST_PATH_IMAGE005
in order to be a term of the offset,
Figure 965481DEST_PATH_IMAGE006
，
Figure 109017DEST_PATH_IMAGE007
，……
Figure 897982DEST_PATH_IMAGE008
for each explanatory variable coefficient;
derived from the interpreted variable z-expression of the training set
Figure 62247DEST_PATH_IMAGE009
The expression of (c) is:
Figure 569451DEST_PATH_IMAGE010
；
wherein the content of the first and second substances,
Figure 375733DEST_PATH_IMAGE011
to train the interpreted variable matrix of the set,
Figure 753625DEST_PATH_IMAGE012
in order to interpret the matrix of coefficients of the variables,
Figure 495316DEST_PATH_IMAGE013
to explain the variable matrix;
Figure 958659DEST_PATH_IMAGE014
a transpose matrix for interpreting the variable coefficients; i is the number of training set samples;
actual value matrix y of training set (i) The expression of (c) is:
Figure 568631DEST_PATH_IMAGE015
；
wherein, 400 (i) Is an error matrix;
the loss function defined after averaging the loss functions of m samples is:
Figure 332188DEST_PATH_IMAGE016
；
according to the small batch gradient descent method, the t-th iteration machine selects a subset D containing K samples t Calculating the gradient of the loss function of each sample of the subset and averaging, and then updating the parameters:
Figure 838256DEST_PATH_IMAGE017
；
wherein: theta is a fitting function parameter result, and alpha is a learning factor;
fifthly, correcting the explained variable z according to the weight of the value calculated by theta;
step six, triggering a preset early warning rule after a certain fault of a certain forging hydraulic machine occurs, calculating the estimated failure time of corresponding early warning according to a regression function, and giving preventive maintenance measures;
and seventhly, if preventive maintenance measures are taken, early warning is reset, and if the preventive maintenance measures are not taken, a preset failure rule triggers the system to fail.
Preferably, the first step is specifically: the controller of each hydraulic forging press acquires the following data of equipment operation through a sensor: the oil cleanliness, the working condition environmental pollution degree, the oil temperature and the environmental temperature; the controller of each forging hydraulic press counts the following operating data: the equipment investment time, the equipment actual working time, the maximum tonnage use times in the current time period and the total tonnage use times of the equipment.
Preferably, the preset early warning rules comprise early warning that the oil cleanliness reaches ISO15, early warning that the oil temperature reaches 45 ℃ and strain early warning.
Preferably, the preset failure rule comprises an alarm for the oil cleanliness reaching ISO19, an alarm for the oil temperature reaching 60 ℃ and a strain alarm.
Preferably, the estimated time to failure comprises: the oil cleanliness estimated failure time, the oil temperature estimated failure time and the strain estimated failure time.
A second object of the present invention is to provide a method of predicting systematic failure of a forging hydraulic press, comprising:
a parameter acquisition module: collecting state data of a hydraulic forging press;
setting a module: setting a preset early warning rule and a preset failure rule of the predictive maintenance system;
a sample data generation module: when the preset early warning rule is triggered, the predictive maintenance system receives the state data of each forging hydraulic press, the state data is used as an interpretation variable x of a regression analysis training sample,
a data fitting module: recording the time difference between alarm failure and early warning, and defining the time difference as estimated failure time; taking the estimated failure time as the failure time actual value y of a regression analysis training sample, taking the estimated failure time of the regression analysis as an interpreted variable z, wherein the interpreted variable z of the training set has an expression as follows:
Figure 320053DEST_PATH_IMAGE001
；
wherein:
Figure 202558DEST_PATH_IMAGE002
，
Figure 695987DEST_PATH_IMAGE003
，……
Figure 638536DEST_PATH_IMAGE004
for the explanation variables, n is the number of the explanation variables, and n is more than or equal to 2;
Figure 76470DEST_PATH_IMAGE005
in order to be a bias term, the bias term,
Figure 28246DEST_PATH_IMAGE006
，
Figure 235236DEST_PATH_IMAGE018
，……
Figure 614265DEST_PATH_IMAGE008
for each explanatory variable coefficient;
derived from the interpreted variable z-expression of the training set
Figure 539496DEST_PATH_IMAGE009
The expression of (c) is:
Figure 415003DEST_PATH_IMAGE019
；
wherein the content of the first and second substances,
Figure 7658DEST_PATH_IMAGE020
to train the interpreted variable matrix of the set,
Figure 292009DEST_PATH_IMAGE012
in order to explain the matrix of coefficients of the variables,
Figure 704536DEST_PATH_IMAGE021
to explain the variable matrix;
Figure 732535DEST_PATH_IMAGE014
a transpose matrix for interpreting the variable coefficients; i is the number of training set samples;
actual value matrix y of training set (i) The expression of (a) is:
Figure 179697DEST_PATH_IMAGE015
；
wherein, 400 (i) Is an error matrix;
the average loss function over m samples defines the loss function as:
Figure 634949DEST_PATH_IMAGE022
；
according to the small batch gradient descent method, the t-th iteration machine selects a subset D containing K samples t Calculating the gradient of the loss function of each sample in the subset and averaging, and then updating the parameters:
Figure 675717DEST_PATH_IMAGE023
；
wherein: theta is a fitting function parameter result, and alpha is a learning factor;
a correction module: correcting the explained variable z according to the weight of the value theta;
the early warning module: when a certain fault of a certain hydraulic forging press occurs, triggering a preset early warning rule, calculating the estimated failure time of corresponding early warning according to a regression function, and giving preventive maintenance measures;
a processing module: if preventive maintenance measures are taken, the early warning is reset, and if the preventive maintenance measures are not taken, the system fails after preset failure rules are triggered.
Preferably, the parameter obtaining process is as follows: the controller of each hydraulic forging press collects the following data of the equipment operation through a sensor: the oil cleanliness, the working condition environmental pollution degree, the oil temperature and the environmental temperature; the controller of each forging hydraulic press counts the following operating data: the equipment investment time, the equipment actual working time, the maximum tonnage use times in the current time period and the total tonnage use times of the equipment.
Preferably, the preset early warning rules comprise early warning that the oil cleanliness reaches ISO15, early warning that the oil temperature reaches 45 ℃ and strain early warning.
Preferably, the preset failure rules comprise an alarm that the oil cleanliness reaches ISO19, an alarm that the oil temperature reaches 60 ℃ and a strain alarm.
Preferably, the estimated time to failure comprises: the oil cleanliness pre-estimation failure time, the oil temperature pre-estimation failure time and the strain pre-estimation failure time.
The invention has the advantages and positive effects that:
1. the method adopts the multiple linear regression algorithm to solve the prediction method of the systematic failure time of the forging hydraulic press, gives temporary measures and preventive maintenance measures, and reduces the loss caused by equipment shutdown failure.
2. According to the method, the early warning rules of each failure mode of the hydraulic forging press are preset, the time for triggering the early warning rules is used as the starting time of the estimated failure time, the final failure time of the equipment is used as the ending time of the estimated failure time, on one hand, the failure prediction is emphasized, and on the other hand, the collected data are convenient to predict by adopting a regression algorithm.
3. The regression function parameters of the invention can be continuously learned and iterated according to the accumulation of data, so that the systematic failure prediction of the forging hydraulic press is more accurate.
Drawings
FIG. 1 is a diagram of a predictive maintenance procedure for oil cleanliness for a 7000 ton free forging hydraulic press.
Detailed Description
In order to further understand the contents, features and effects of the present invention, the following embodiments are illustrated and described in detail with reference to the accompanying drawings:
the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. Based on the technical solutions in the present invention, all other embodiments obtained by a person of ordinary skill in the art without making creative efforts belong to the protection scope of the present invention.
Please refer to fig. 1.
A method of predicting systematic failure of a forging press, comprising:
step one, acquiring state data of a hydraulic forging press: the controller of each hydraulic forging press acquires equipment operation data through a sensor, and the method comprises the following steps: the cleanliness of oil, the degree of working condition environmental pollution, the temperature of the oil, the environmental temperature and the like; the controller of each hydraulic forging press counts the operation data, and comprises the following steps: the equipment investment time, the equipment actual working time, the maximum tonnage use times in the current time period and the maximum tonnage use times of the equipment.
Step two, presetting early warning rules for a predictive maintenance system of a forging hydraulic press, comprising the following steps: the oil cleanliness reaches ISO15 early warning, the oil temperature reaches 45 ℃, and strain early warning and the like are carried out.
The system presets failure rules, which comprise: the oil cleanliness reaches ISO19 alarm, the oil temperature reaches 60 ℃ alarm, strain alarm and the like. The system records the time difference between alarm failure and early warning as the estimated failure time, and comprises the following steps: and the oil cleanliness pre-estimated failure time, the oil temperature pre-estimated failure time, the strain pre-estimated failure time and the like. And the estimated failure time is used as the actual failure time value y of the regression analysis training sample.
Step three, after the preset early warning rule is triggered, the predictive maintenance system receives the data of each forging hydraulic press as an explanatory variable x of a regression analysis training sample,
and step four, taking the estimated failure time of the regression analysis as an explained variable z. The interpreted variable z of the training set is expressed as:
Figure 507407DEST_PATH_IMAGE024
；
wherein:
Figure 543496DEST_PATH_IMAGE002
，
Figure 435228DEST_PATH_IMAGE003
，……
Figure 556768DEST_PATH_IMAGE004
for the explanation variables, n is the number of the explanation variables, and n is more than or equal to 2;
Figure 192149DEST_PATH_IMAGE005
in order to be a term of the offset,
Figure 489269DEST_PATH_IMAGE025
，
Figure 286324DEST_PATH_IMAGE018
，……
Figure 160739DEST_PATH_IMAGE008
coefficients for each of the explanatory variables;
derived from the z-expressions of the interpreted variables of the training set
Figure 68652DEST_PATH_IMAGE009
The expression of (c) is:
Figure 344913DEST_PATH_IMAGE019
；
wherein, the first and the second end of the pipe are connected with each other,
Figure 312869DEST_PATH_IMAGE026
to train the interpreted variable matrix of the set,
Figure 674580DEST_PATH_IMAGE012
in order to explain the matrix of coefficients of the variables,
Figure 527129DEST_PATH_IMAGE021
to explain the variable matrix;
Figure 657896DEST_PATH_IMAGE014
a transpose matrix for the interpretation of the variable coefficients; i is the number of training set samples;
actual value matrix y of training set (i) The expression of (a) is:
Figure 62333DEST_PATH_IMAGE015
；
wherein, 400 (i) Is an error matrix;
Ɛ (i) are independent and have the same distribution and follow a normal distribution with a mean value of 0, i.e.：
Ɛ (i) ～N(0,σ)。
To make the predicted values (output values of the fitting function) and the true values (output values of the true value function) closest, it is necessary to minimize the loss function. The loss function defined after averaging the loss functions of m samples is:
Figure 380182DEST_PATH_IMAGE027
；
according to the small batch gradient descent method, a sub-set D containing K samples is selected by the t-th iteration machine t The gradient of the loss function of each sample of the subset is calculated and averaged, and then the parameter is updated.
Figure 161056DEST_PATH_IMAGE028
；
Wherein: theta is the fitting function parameter result, and alpha is the learning factor.
And step five, correcting the fitting function z according to the actual experience and the weight of the calculated value of theta. Such as: the estimated failure time of the oil cleanliness is a function of the equipment use time, the actual equipment working time and the working condition environmental pollution degree; the predicted failure time of the oil temperature is a function of the maximum tonnage use times and the environmental temperature in the current time period; the strain prediction failure time is a function of the maximum tonnage use times of the equipment and the actual working time of the equipment.
And step six, when a certain forging hydraulic machine applies the established predictive maintenance system, triggering a preset early warning rule after a certain fault occurs, calculating the estimated failure time of the corresponding early warning by the system according to a regression function, and giving preventive maintenance measures.
And seventhly, if the user takes preventive maintenance measures, early warning reset is carried out, and if the user does not take preventive maintenance measures, the preset failure rule triggers the system to fail.
The following is an example of oil cleanliness failure prediction:
according to the requirements of critical elements such as a hydraulic valve used by a hydraulic forging press and the like on the cleanliness of ISO oil, the main valve of the hydraulic forging press needs to reach ISO20/18/15, and the pilot control valve needs to reach ISO18/16/13. Therefore, ISO oil cleanliness ratings need to be at least ISO18 before the system can reach the desired operating conditions. According to historical data and regression algorithm of enough multiple equipment, after the oil pollution detector of the predictive maintenance system of the forging hydraulic press detects that the oil cleanliness estimated failure time z exceeds the early warning value ISO15, the inverse x of the time of the equipment in use is estimated 1 Reciprocal x of actual working time of equipment 2 Degree of environmental pollution x 3 As a function of (c).
Figure 146330DEST_PATH_IMAGE029
；
Wherein z is estimated failure time of oil cleanliness,
Figure 721667DEST_PATH_IMAGE030
is the inverse of the time of use of the equipment,
Figure 166293DEST_PATH_IMAGE031
is the inverse of the actual working time of the equipment,
Figure 219700DEST_PATH_IMAGE032
in order to meet the environmental pollution degree of the working condition,
Figure 325059DEST_PATH_IMAGE033
in order to be a bias term, the bias term,
Figure 71298DEST_PATH_IMAGE034
for the coefficient of the time-of-use term of the equipment,
Figure 629318DEST_PATH_IMAGE035
in order to equip the actual working time term coefficients,
Figure 486416DEST_PATH_IMAGE036
is a coefficient of environmental pollution degree under working conditions.
The 7000 ton free forging hydraulic press uses an oil contamination detector to continuously detect particulate contamination in the hydraulic system. And when the equipment is used to the time t1, the oil cleanliness grade reaches ISO15. The system prompts according to the function: the oil cleanliness reaches ISO15, the estimated failure time is 153 days, the oil cleanliness reaches ISO19 after the failure time is up, and the equipment is failed and stopped. The system gives suggested repair or maintenance measures: starting an online oil filtering system for circulating filtration; temporary solution measures: and replacing a filter of a cooling circulation part of the system, and starting the circulation system to delay the deterioration of the cleanliness of the oil.
And the user starts the online oil filtering device to filter oil to a desired state.
When the equipment runs to the time t2, the system warns and prompts again, but the online oil filtering system is not started at the time, and a temporary solution is not taken. When the equipment continues to operate to the time t3, the equipment is in failure shutdown, and the system sends a shutdown alarm prompt: the cleanliness of oil liquid exceeds the allowable use state, the online oil filtering device needs to be opened for oil filtering, and the equipment cannot be used before the standard of use is not recovered. The system records the failure time as the sample data of parameter iteration.
And (4) starting the online oil filtering device at the moment of t4 to filter oil, and continuing to operate the equipment after the oil is filtered to an ideal state.
If special conditions (such as the damage of an oil pump causes the severe damage of the oil cleanliness), the oil cleanliness exceeds the allowable use range of the system after the system operates to the time t5, and the system can be normally used after shutdown maintenance is required.
A systematic failure prediction system for a forging press, comprising:
a parameter acquisition module: collecting state data of a hydraulic forging press: the controller of each hydraulic forging press collects the operation data of the equipment through a sensor, and the method comprises the following steps: the cleanliness of oil, the degree of working condition environmental pollution, the temperature of the oil, the environmental temperature and the like; the controller of each hydraulic forging press counts the operation data, and comprises the following steps: the equipment investment time, the equipment actual working time, the maximum tonnage use times in the current time period and the maximum tonnage use times of the equipment.
Setting a module: the pre-set early warning rule of the predictive maintenance system of the forging hydraulic press comprises the following steps: the oil cleanliness reaches ISO15 early warning, the oil temperature reaches 45 ℃ early warning, strain early warning and the like.
The system presets failure rules, which comprise: the oil cleanliness reaches ISO19 alarm, the oil temperature reaches 60 ℃ alarm, strain alarm and the like. The system records the time difference between alarm failure and early warning as the estimated failure time, and comprises the following steps: the estimated failure time of the oil cleanliness, the estimated failure time of the oil temperature, the estimated failure time of strain and the like. And the estimated failure time is used as the actual failure time value y of the regression analysis training sample.
A sample data generation module: when the preset early warning rule is triggered, the predictive maintenance system receives the data of each forging hydraulic press as the interpretation variable x of the regression analysis training sample,
a data fitting module: the estimated time to failure of the regression analysis is taken as the interpreted variable z. The interpreted variable z of the training set is expressed as:
Figure 446282DEST_PATH_IMAGE037
；
wherein:
Figure 238788DEST_PATH_IMAGE002
，
Figure 18525DEST_PATH_IMAGE038
，……
Figure 679314DEST_PATH_IMAGE004
for the explanation variables, n is the number of the explanation variables, and n is more than or equal to 2;
Figure 759265DEST_PATH_IMAGE005
in order to be a term of the offset,
Figure 581728DEST_PATH_IMAGE025
，
Figure 114340DEST_PATH_IMAGE018
，……
Figure 313241DEST_PATH_IMAGE008
for each explanatory variable coefficient;
derived from the z-expressions of the interpreted variables of the training set
Figure 123065DEST_PATH_IMAGE039
The expression of (a) is:
Figure 116428DEST_PATH_IMAGE019
；
wherein the content of the first and second substances,
Figure 136337DEST_PATH_IMAGE026
to train the interpreted variable matrix of the set,
Figure 138928DEST_PATH_IMAGE012
in order to explain the matrix of coefficients of the variables,
Figure 927893DEST_PATH_IMAGE021
to explain the variable matrix;
Figure 92158DEST_PATH_IMAGE014
a transpose matrix for interpreting the variable coefficients; i is the number of training set samples;
actual value matrix y of training set (i) The expression of (a) is:
Figure 599362DEST_PATH_IMAGE015
；
wherein, 400 (i) Is an error matrix;
Ɛ (i) are independent and have the same distribution and follow a normal distribution with a mean of 0, i.e.:
Ɛ (i) ～N(0,σ)。
to make the predicted values (output values of the fitting function) and the true values (output values of the true value function) closest, it is necessary to minimize the loss function. The average loss function over m samples defines the loss function as:
Figure 281011DEST_PATH_IMAGE040
；
according to the small batch gradient descent method, the t-th iteration machine selects a subset D containing K samples t The gradient of the loss function of each sample of the subset is calculated and averaged, and then the parameter is updated.
Figure 658902DEST_PATH_IMAGE041
；
Wherein: theta is the fitting function parameter result, and alpha is the learning factor.
A correction module: and modifying the fitting function z according to actual experience and the weight of the calculated value of theta. Such as: the estimated failure time of the oil cleanliness is a function of the equipment use time, the equipment actual working time and the working condition environmental pollution degree; the predicted failure time of the oil temperature is a function of the maximum tonnage use times and the environmental temperature in the current time period; the strain prediction failure time is a function of the maximum tonnage use times of the equipment and the actual working time of the equipment.
The early warning module: when a certain hydraulic forging press applies the established predictive maintenance system, a preset early warning rule is triggered after a certain fault occurs, the system calculates the estimated failure time of the corresponding early warning according to the regression function, and preventive maintenance measures are given.
A processing module: if the user takes preventive maintenance measures, the early warning is reset, and if the user does not take preventive maintenance measures, the system fails after preset failure rule triggering.
The above description is only a preferred embodiment of the present invention, and is not intended to limit the present invention in any way, and any simple modifications, equivalent variations and modifications made to the above embodiment according to the technical spirit of the present invention are within the scope of the technical solution of the present invention.

Claims (10)

1. A method for predicting systematic failure of a forging hydraulic press, comprising:
step one, collecting state data of a hydraulic forging press;
step two, setting a preset early warning rule and a preset failure rule of the predictive maintenance system;
step three, after triggering of a preset early warning rule, receiving state data of each forging hydraulic press, and taking the state data as an interpretation variable x of a regression analysis training sample;
recording the time difference between alarm failure and early warning, and defining the time difference as estimated failure time; taking the estimated failure time as the failure time actual value y of a regression analysis training sample, taking the estimated failure time of the regression analysis as an interpreted variable z, wherein the interpreted variable z of the training set has an expression as follows:
Figure 732989DEST_PATH_IMAGE001
；
wherein:
Figure 55517DEST_PATH_IMAGE002
，
Figure 510769DEST_PATH_IMAGE003
，……
Figure 676171DEST_PATH_IMAGE004
for the explanation variables, n is the number of the explanation variables, and n is more than or equal to 2;
Figure 507861DEST_PATH_IMAGE005
in order to be a term of the offset,
Figure 543950DEST_PATH_IMAGE006
，
Figure 435682DEST_PATH_IMAGE007
，……
Figure 822801DEST_PATH_IMAGE008
coefficients for each of the explanatory variables;
derived from the z-expressions of the interpreted variables of the training set
Figure 67969DEST_PATH_IMAGE009
The expression of (a) is:
Figure 489723DEST_PATH_IMAGE010
；
wherein, the first and the second end of the pipe are connected with each other,
Figure 286778DEST_PATH_IMAGE011
to train the interpreted variable matrix of the set,
Figure 161193DEST_PATH_IMAGE012
in order to interpret the matrix of coefficients of the variables,
Figure 6789DEST_PATH_IMAGE013
to explain the variable matrix;
Figure 17471DEST_PATH_IMAGE014
a transpose matrix for interpreting the variable coefficients; i is the number of training set samples;
actual value matrix y of training set (i) The expression of (a) is:
Figure 251006DEST_PATH_IMAGE015
；
wherein, 400 (i) Is an error matrix;
the loss function defined after averaging the loss functions of m samples is:
Figure 612717DEST_PATH_IMAGE016
；
according to the small batch gradient descent method, the t-th iteration machine selects a subset D containing K samples t Calculating the gradient of the loss function of each sample in the subset and averaging, and then updating the parameters:
Figure 58742DEST_PATH_IMAGE017
；
wherein: theta is a fitting function parameter result, and alpha is a learning factor;
fifthly, correcting the explained variable z according to the weight of the value calculated by theta;
step six, triggering a preset early warning rule after a certain fault of a certain forging hydraulic machine occurs, calculating the estimated failure time of corresponding early warning according to a regression function, and giving preventive maintenance measures;
and seventhly, if preventive maintenance measures are taken, early warning reset is carried out, and if the preventive maintenance measures are not taken, a preset failure rule triggers the system to fail.
2. The method for predicting systematic failure of a forging press as set forth in claim 1, wherein the first step is specifically: the controller of each hydraulic forging press collects the following data of the equipment operation through a sensor: the oil cleanliness, the working condition environmental pollution degree, the oil temperature and the environmental temperature; the controller of each forging hydraulic press counts the following operating data: the equipment investment and use time, the equipment actual working time, the maximum tonnage use times in the current time period and the total tonnage use times of the equipment.
3. The method of predicting systematic failure of a forging hydraulic press as set forth in claim 1, wherein the preset early warning rules include an oil cleanliness reaching ISO15 early warning, an oil temperature reaching 45 ℃ early warning, and a strain early warning.
4. The method of predicting systematic failure of a forging hydraulic press of claim 1, wherein the preset failure rules include an oil cleanliness up to ISO19 alarm, an oil temperature up to 60 ℃ alarm, and a strain alarm.
5. The method of claim 1, wherein the estimated time to failure comprises: the oil cleanliness pre-estimation failure time, the oil temperature pre-estimation failure time and the strain pre-estimation failure time.
6. A systematic failure prediction system for a forging press, comprising:
a parameter acquisition module: collecting state data of a hydraulic forging press;
setting a module: setting a preset early warning rule and a preset failure rule of the predictive maintenance system;
a sample data generation module: when a preset early warning rule is triggered, the predictive maintenance system receives the state data of each hydraulic forging press, the state data is used as an explanatory variable x of a regression analysis training sample,
a data fitting module: recording the time difference between alarm failure and early warning, and defining the time difference as estimated failure time; taking the estimated failure time as the failure time actual value y of a regression analysis training sample, taking the estimated failure time of the regression analysis as an interpreted variable z, wherein the interpreted variable z of the training set has an expression as follows:
Figure 828989DEST_PATH_IMAGE018
；
wherein:
Figure 967847DEST_PATH_IMAGE002
，
Figure 816854DEST_PATH_IMAGE019
，……
Figure 332149DEST_PATH_IMAGE004
for the explanation variables, n is the number of the explanation variables, and n is more than or equal to 2;
Figure 317423DEST_PATH_IMAGE005
in order to be a bias term, the bias term,
Figure 892760DEST_PATH_IMAGE020
，
Figure 963485DEST_PATH_IMAGE007
，……
Figure 157837DEST_PATH_IMAGE008
for each explanatory variable coefficient;
derived from the z-expressions of the interpreted variables of the training set
Figure 997617DEST_PATH_IMAGE021
The expression of (c) is:
Figure 743856DEST_PATH_IMAGE010
；
wherein the content of the first and second substances,
Figure 301876DEST_PATH_IMAGE022
to train the interpreted variable matrix of the set,
Figure 158974DEST_PATH_IMAGE012
in order to interpret the matrix of coefficients of the variables,
Figure 118839DEST_PATH_IMAGE013
to explain the variable matrix;
Figure 770401DEST_PATH_IMAGE014
a transpose matrix for interpreting the variable coefficients; i is the number of training set samples;
actual value matrix y of training set (i) The expression of (c) is:
Figure 956662DEST_PATH_IMAGE015
；
wherein, 400 (i) Is an error matrix;
the loss function defined after averaging the loss functions of m samples is:
Figure 617451DEST_PATH_IMAGE023
；
according to the small batch gradient descent method, the t-th iteration machine selects a subset D containing K samples t Calculating the gradient of the loss function of each sample of the subset and averaging, and then updating the parameters:
Figure 166244DEST_PATH_IMAGE024
；
wherein: theta is a fitting function parameter result, and alpha is a learning factor;
a correction module: correcting the explained variable z according to the weight of the value theta;
the early warning module: when a certain fault of a certain hydraulic forging press occurs, triggering a preset early warning rule, calculating the estimated failure time of corresponding early warning according to a regression function, and giving preventive maintenance measures;
a processing module: if preventive maintenance measures are taken, the early warning is reset, and if the preventive maintenance measures are not taken, the preset failure rule triggers the system to fail.
7. The system for systematic failure prediction of a forging press of claim 6, wherein the parameter acquisition process is: the controller of each hydraulic forging press acquires the following data of equipment operation through a sensor: the oil cleanliness, the working condition environmental pollution degree, the oil temperature and the environmental temperature; the controller of each forging hydraulic press counts the following operating data: the equipment investment time, the equipment actual working time, the maximum tonnage use times in the current time period and the total tonnage use times of the equipment.
8. The system for predicting systematic failure of a forging hydraulic press of claim 6, wherein the preset early warning rules include an oil cleanliness reaching ISO15 early warning, an oil temperature reaching 45 ℃ early warning, and a strain early warning.
9. The system for systematic failure prediction of hydraulic forging presses of claim 6, wherein the preset failure rules include an oil cleanliness up to ISO19 alarm, an oil temperature up to 60 ℃ alarm, and a strain alarm.
10. The system of claim 6, wherein the estimated time to failure comprises: the oil cleanliness estimated failure time, the oil temperature estimated failure time and the strain estimated failure time.
CN202210780611.7A 2022-07-05 2022-07-05 Systematic failure prediction method and system for hydraulic forging press Active CN114862064B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210780611.7A CN114862064B (en) 2022-07-05 2022-07-05 Systematic failure prediction method and system for hydraulic forging press

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210780611.7A CN114862064B (en) 2022-07-05 2022-07-05 Systematic failure prediction method and system for hydraulic forging press

Publications (2)

Publication Number Publication Date
CN114862064A CN114862064A (en) 2022-08-05
CN114862064B true CN114862064B (en) 2022-11-22

Family

ID=82625960

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210780611.7A Active CN114862064B (en) 2022-07-05 2022-07-05 Systematic failure prediction method and system for hydraulic forging press

Country Status (1)

Country Link
CN (1) CN114862064B (en)

Family Cites Families (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP4542819B2 (en) * 2004-05-21 2010-09-15 株式会社小松製作所 Hydraulic machine, system and method for monitoring the health status of a hydraulic machine
US20200010101A1 (en) * 2018-07-06 2020-01-09 Alstom Transport Technologies Method and system for monitoring a point system of a railway network, and point system of a railway network
CN111765075B (en) * 2020-05-20 2022-05-03 天津市天锻压力机有限公司 Hydraulic forging press pump source fault prediction method and system
CN112016193B (en) * 2020-08-07 2021-05-04 震坤行工业超市(上海)有限公司 Online prediction method and system for lubrication failure of shield tunneling machine system
CN112819293B (en) * 2021-01-14 2023-01-06 中国长江三峡集团有限公司 Failure early warning analysis method for water reservoir scheduling rule under influence of climate change

Also Published As

Publication number Publication date
CN114862064A (en) 2022-08-05

Similar Documents

Publication Publication Date Title
CN112527788A (en) Method and device for detecting and cleaning abnormal value of transformer monitoring data
RU2649542C1 (en) Method and system of remote monitoring of objects
RU2626780C1 (en) Method and system of remote monitoring energy installations
CN112713649A (en) Power equipment residual life prediction method based on extreme learning machine
CN112598144B (en) CNN-LSTM burst fault early warning method based on correlation analysis
CN112904810B (en) Process industry nonlinear process monitoring method based on effective feature selection
CN110209144B (en) Two-layer real-time monitoring and alarm tracing method based on dynamic and static cooperative difference analysis
Akram et al. Quality monitoring and process adjustment by integrating SPC and APC: a review
CN111797533B (en) Nuclear power device operation parameter abnormity detection method and system
CN114862064B (en) Systematic failure prediction method and system for hydraulic forging press
CN113887606A (en) Electronic equipment control system fault diagnosis method based on fault tree establishment
CN114519382A (en) Nuclear power plant key operation parameter extraction and abnormity monitoring method
CN117318310A (en) Remote unmanned intelligent monitoring system for explosion-proof power distribution cabinet
CN111652393A (en) Power equipment abnormity early warning method based on big data technology
CN115909535A (en) Blast furnace ironmaking process monitoring method based on RBNCVA
Codoban et al. Monitoring system of production installation for the separation of isotope 18 O
CN112598030A (en) Non-stationary process monitoring method based on recursive covariance analysis and elastic weight consolidation
Galicia et al. A comprehensive evaluation of Statistics Pattern Analysis based process monitoring
Liu et al. On-line process monitoring and fault isolation using pca
CN111474911B (en) Gaussian non-Gaussian characteristic collaborative analysis and monitoring method for non-steady operation of high-end coal-fired power generation equipment
Zhang et al. A novel method of fault detection method for tep based mipcr
Xuefeng et al. Study on Multi-Sensor Parameter Warning Based on Gaussian Mixed Model and Nonlinear State Estimation Technique
Attieh et al. Pattern recognition techniques for transient detection to enhance nuclear reactors' operational safety
Kazemi et al. Advanced Online Monitoring of Fars Province Gas Pressure Reduction Stations Based on Distributed Control System of PCS7
CN115809941A (en) Individualized accurate maintenance strategy based on transformer equipment operation situation perception analysis

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