CN115718466B - Digital twin workshop fault prediction method based on random forest and analytic hierarchy process - Google Patents
Digital twin workshop fault prediction method based on random forest and analytic hierarchy process Download PDFInfo
- Publication number
- CN115718466B CN115718466B CN202211457614.3A CN202211457614A CN115718466B CN 115718466 B CN115718466 B CN 115718466B CN 202211457614 A CN202211457614 A CN 202211457614A CN 115718466 B CN115718466 B CN 115718466B
- Authority
- CN
- China
- Prior art keywords
- production
- judgment matrix
- predictive maintenance
- digital twin
- layer
- 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
Links
- 238000000034 method Methods 0.000 title claims abstract description 85
- 238000007637 random forest analysis Methods 0.000 title claims abstract description 59
- 230000008569 process Effects 0.000 title claims abstract description 37
- 238000004519 manufacturing process Methods 0.000 claims abstract description 189
- 239000011159 matrix material Substances 0.000 claims abstract description 150
- 238000012423 maintenance Methods 0.000 claims abstract description 124
- 239000013598 vector Substances 0.000 claims abstract description 84
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 24
- 238000012545 processing Methods 0.000 claims abstract description 14
- 230000000694 effects Effects 0.000 claims abstract description 11
- 238000003066 decision tree Methods 0.000 claims description 27
- 230000009471 action Effects 0.000 claims description 10
- 238000005070 sampling Methods 0.000 claims description 5
- 230000003993 interaction Effects 0.000 claims description 4
- 230000007246 mechanism Effects 0.000 claims description 4
- 238000010606 normalization Methods 0.000 claims description 4
- 238000005516 engineering process Methods 0.000 description 8
- 238000012549 training Methods 0.000 description 8
- 230000007547 defect Effects 0.000 description 5
- 238000010586 diagram Methods 0.000 description 4
- 230000000007 visual effect Effects 0.000 description 4
- 238000009776 industrial production Methods 0.000 description 3
- 238000007726 management method Methods 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 2
- 238000010276 construction Methods 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 239000000463 material Substances 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 238000012544 monitoring process Methods 0.000 description 2
- 230000002159 abnormal effect Effects 0.000 description 1
- 238000005299 abrasion Methods 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000001808 coupling effect Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000036541 health Effects 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 238000003754 machining Methods 0.000 description 1
- 238000013386 optimize process Methods 0.000 description 1
- 238000012163 sequencing technique Methods 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 239000002699 waste material Substances 0.000 description 1
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention provides a digital twin workshop fault prediction method based on a random forest and analytic hierarchy process, which relates to the technical field of intelligent manufacturing and electronic information and comprises the steps of obtaining production element data and real-time production data of a production workshop, constructing a digital twin model of the production workshop according to the obtained production element data, carrying out digital twin processing on the real-time production data to obtain the digital twin data of the production workshop, obtaining a predictive maintenance hierarchy judgment matrix by using a random forest algorithm, solving the predictive maintenance hierarchy judgment matrix to obtain a judgment matrix feature vector, carrying out fault prediction on the production workshop by using the judgment matrix feature vector to obtain a fault prediction result of the production workshop, and controlling and managing production activities of the production workshop according to the fault prediction result of the production workshop; according to the method, the random forest and the analytic hierarchy process are combined to conduct real-time fault prediction of the production workshop, the prediction accuracy is high, the speed is high, the adaptability is high, and the working difficulty of maintenance personnel is remarkably reduced.
Description
Technical Field
The invention relates to the technical field of intelligent manufacturing and electronic information, in particular to a digital twin workshop fault prediction method based on a random forest and analytic hierarchy process.
Background
Digital twin technology is widely used in more and more enterprises, particularly in the industry of shifting from product sales to product service binding sales or as service sales; with the improvement of enterprise capability and maturity, more enterprises will use digital twin technology to optimize processes, decide data drives, revise new products, new services and business modes in the future.
As a basic unit of industrial production, the level of digitization and intelligence of the workshops has an important influence on the quality, safety and efficiency of industrial production. Meanwhile, large industrial equipment running at high speed, mobile personnel and complex operation environment also make workshops become high places for enterprise safety accidents. However, at present, the digitization level of workshop safety management and equipment health management is still to be improved, but at present, no model based on real-time data is available, which can comprehensively consider the coupling effect of multiple risk factors such as equipment, personnel, environment and the like.
The aim of utilizing the digital twin technology is to reduce the investment of manpower and material resources and improve the working efficiency. The manual monitoring has a large number of defects, the manual monitoring equipment is time-consuming and labor-consuming, and the wrong judgment is very easy to occur. The digital twin technology is utilized to create a virtual model for a physical object in a digital mode to simulate the behavior of the physical object in a real environment, and the whole running process of the equipment is monitored in real time during fault prediction, so that the real-time fault prediction of the equipment is facilitated, and the prediction efficiency and accuracy are improved.
The analytic hierarchy process mainly solves the problem of fault prediction. During fault prediction, the collected data are processed by using a hierarchical analysis method, the decision-making problem is decomposed into different hierarchical structures according to the sequence of a total target, sub-targets of each layer and evaluation criteria until a specific spare power switching scheme is achieved, then a method for solving and judging matrix characteristic vectors is used, wherein elements in the matrix are data collected from equipment, the priority weight of each element of each layer on a certain element of the previous layer is obtained, and finally the final weight of each alternative scheme on the total target is reduced by a weighted sum method, and the final weight with the maximum weight is the optimal scheme, so that the fault prediction of the equipment is completed.
Because the analytic hierarchy process has the defect of strong subjectivity, when the defect is overcome, a new prediction module is constructed by combining a random forest algorithm and the analytic hierarchy process, and the accuracy of prediction is improved. The random forest is composed of a plurality of decision trees, each decision tree is different, when the decision tree is constructed, a part of samples are randomly selected from training data, all the characteristics of the data are not used, and part of the characteristics are randomly selected for training. The samples and features used for each tree are different, and the training results are different. Because of the randomness of the random forest algorithm, the subjectivity of the prediction can be well eliminated, so that the combination of the random forest algorithm and the analytic hierarchy process greatly improves the accuracy of the prediction.
The prior art discloses a digital twin system for industrial production, comprising: the data acquisition module comprises a dynamic data information unit and a static data information unit; a model building module; the model building module comprises a space model unit, a characteristic model unit, a production flow model unit and a model integration unit; a digital twinning module; the digital twin module comprises an information extraction unit and a digital twin unit; in the method in the prior art, the production scheme is optimized, predicted and managed only by a digital twin system in the production process, and the excessive maintenance is easy to cause due to the non-sampling analytic hierarchy process, namely, unnecessary disassembly, component replacement and the like; in addition, the method and the equipment in the prior art have longer shutdown maintenance time, so that the waste of manpower and material resources is caused; in addition, the existing predictive maintenance has the defect of strong subjectivity, the accuracy of fault prediction is low, and the effect is poor when the complex model fault prediction is faced.
Disclosure of Invention
The invention provides the digital twin workshop fault prediction method based on the random forest and analytic hierarchy process, which overcomes the defects of low accuracy and poor adaptability in the fault prediction in the prior art, has the advantages of high accuracy, high training speed, strong adaptability and the like, and can be better suitable for workshop fault prediction.
In order to solve the technical problems, the technical scheme of the invention is as follows:
a digital twin workshop fault prediction method based on random forest and analytic hierarchy process comprises the following steps:
S1: acquiring production element data and real-time production data of a production workshop;
s2: constructing a digital twin model of the production workshop according to the production element data of the production workshop;
S3: carrying out digital twin processing on the real-time production data by using a digital twin model of a production workshop to obtain digital twin data of the production workshop;
S4: inputting digital twin data of a production workshop into a constructed predictive maintenance hierarchical structure model, obtaining a predictive maintenance hierarchical judgment matrix by utilizing a random forest algorithm, solving the predictive maintenance hierarchical judgment matrix, and obtaining a judgment matrix feature vector;
S5: performing fault prediction on the production workshop by using the judging matrix feature vector to obtain a fault prediction result of the production workshop;
S6: and controlling and managing the production activities of the production workshops according to the fault prediction results of the production workshops.
Preferably, in the step S1, the production factor data of the production plant includes: device entity element data and information element data;
the equipment entity element data includes: geometric dimensions, physical mechanisms, behavior characteristics, position information and interaction relation among all devices in the production workshop;
the information element data includes: information on the status of various equipment and personnel in the production plant, task order processing information and scheduling decision information.
Preferably, the step S5 further includes:
the predictive maintenance hierarchical structure model uploads the fault prediction result of the production workshop to the database, the digital twin model of the production workshop interacts with the database, the fault prediction result is obtained, and the predicted fault position information is displayed.
Preferably, in the step S4, the constructed predictive maintenance hierarchical structure model is specifically:
the predictive maintenance hierarchical structure model comprises a target layer, a criterion layer and a measure layer;
The target layer represents predictive maintenance of target equipment and is marked as A;
the criterion layer comprises a plurality of criterion layer elements, which are marked as B k and represent the kth criterion layer elements;
the measure layer comprises a plurality of measure layer elements, which are marked as C n and represent the nth measure layer element.
Preferably, in the step S4, the specific method for obtaining the predictive maintenance level judgment matrix by using the random forest algorithm is as follows:
the random forest algorithm comprises a plurality of decision trees;
taking digital twin data of a production workshop as an original data set;
s4.1.1: randomly sampling the original data set for a plurality of times with a replacement to obtain a plurality of sub data sets, wherein the element number of each sub data set is consistent with the element number of the original data set;
S4.1.2: for each sub-data set, randomly selecting a plurality of features from all preset features, wherein each feature is used as an input feature of one decision tree to obtain a plurality of decision trees;
S4.1.3: setting a decision tree with a large information gain value at the top of a decision tree with a small information gain value to form a random forest;
S4.1.4: and obtaining a predictive maintenance level judgment matrix according to the output result of the random forest.
Preferably, the predictive maintenance level judgment matrix in step S4.1.4 is specifically:
For each criterion layer element, a predictive maintenance level judgment matrix is provided, and the predictive maintenance level judgment matrix corresponding to the kth criterion layer element is as follows:
Wherein C ij represents the importance degree of the ith action layer element to the kth criterion layer element compared with the jth action layer element;
and obtaining different values of C ij according to the output result of the random forest to form different predictive maintenance level judgment matrixes.
Preferably, in the step S4, the specific method for solving the predictive maintenance level judgment matrix to obtain the feature vector of the judgment matrix is as follows:
the method for solving the predictive maintenance level judgment matrix corresponding to the k criterion layer element comprises the following steps:
S4.2.1: squaring each element in the predictive maintenance level judgment matrix corresponding to the k criterion layer element to obtain a new judgment matrix C 2(Bk):
wherein i and j satisfy 1.ltoreq.i.ltoreq.n, 1.ltoreq.j.ltoreq.n;
S4.2.2: calculate the sum of all elements of each row in the new decision matrix C 2(Bk):
wherein sum i represents the sum of all elements of row i in the new decision matrix C 2(Bk);
s4.2.3: calculate the sum of all elements in the new decision matrix C 2(Bk):
Where sum represents the sum of all elements in the new decision matrix C 2(Bk);
S4.2.4: calculate a new decision matrix C 2(Bk) normalized vector of all element sums per row:
Wherein W i (q) represents the new judgment matrix C 2(Bk) calculated at the q-th time, the normalized vector of the sum of all elements of the i-th row;
S4.2.5: updating digital twin data of a production plant, executing steps S4.2.1-S4.2.4 again to obtain a new judgment matrix C 2(Bk) of each row of all element sums in a new round, and calculating error values of normalization vectors obtained in two consecutive rounds:
where T represents the error value of the normalized vector obtained twice in succession.
S4.2.6: comparing the error value T of the normalized vector obtained twice continuously with a preset threshold value; if the error value T is smaller than the preset threshold value, taking the normalized vector at the moment as the judgment matrix characteristic vector of the k criterion layer element, and marking asOtherwise, step S4.2.5 is performed until the error value T is less than the preset threshold.
Preferably, in the step S5, fault prediction is performed on the production plant by using the feature vector of the judgment matrix to obtain a fault prediction result of the production plant, and the specific method is as follows:
S5.1: solving the predictive maintenance level judgment matrix corresponding to each criterion layer element to obtain the judgment matrix characteristic vector of the measure layer relative to each criterion layer element, and marking as
S5.2: obtaining a predictive maintenance level judgment matrix of the criterion layer relative to the target layer, solving the predictive maintenance level judgment matrix, and obtaining a judgment matrix characteristic vector of the criterion layer relative to the target layer, wherein the judgment matrix characteristic vector is marked as W B-A;
S5.3: and according to the obtained feature vector of the judgment matrix, obtaining the final priority ranking of predictive maintenance, and taking the final priority ranking of predictive maintenance as a fault prediction result of a production workshop.
Preferably, in the step S5.2, the predictive maintenance level judgment matrix of the criterion layer with respect to the target layer is specifically:
Wherein B ab represents the importance level of the a-th criterion layer element to the target layer compared to the B-th criterion layer element.
Preferably, in step S5.3, the final priority ranking of predictive maintenance is obtained according to the obtained feature vector of the judgment matrix, and the specific method is as follows:
the feature vector of the measure layer relative to the target layer is obtained according to the following formula:
wherein W C-A represents a feature vector of the measure layer relative to the target layer;
And obtaining the importance degree of each measure layer element relative to the target layer according to the feature vector W C-A of the measure layer relative to the target layer, and sorting all the measure layer elements according to the importance degree to obtain the final priority sorting of predictive maintenance.
Compared with the prior art, the technical scheme of the invention has the beneficial effects that:
The invention provides a digital twin workshop fault prediction method based on a random forest and an analytic hierarchy process, which comprises the steps of constructing a digital twin model of a production workshop according to the acquired production element data by acquiring the production element data and real-time production data of the production workshop, carrying out digital twin processing on the real-time production data by utilizing the digital twin model of the production workshop to acquire the digital twin data of the production workshop, inputting the digital twin data of the production workshop into a constructed predictive maintenance hierarchy model, acquiring a predictive maintenance hierarchy judgment matrix by utilizing a random forest algorithm, solving the predictive maintenance hierarchy judgment matrix to acquire a judgment matrix feature vector, carrying out fault prediction on the production workshop by utilizing the judgment matrix feature vector to acquire a fault prediction result of the production workshop, and finally controlling and managing production activities of the production workshop according to the fault prediction result of the production workshop;
The method can not only observe the running condition of workshop equipment in real time under the addition of a visual and real-time digital twin technology, but also directly observe the position of the fault when the fault occurs, thereby saving the time cost of equipment fault retrieval of maintenance personnel, facilitating the maintenance and reducing the working difficulty of the maintenance personnel; in addition, a random forest algorithm and an analytic hierarchy process are combined to conduct production workshop fault prediction, an objective and effective judgment matrix can be obtained, the prediction result is enabled to have confidence, and the method has the advantages of being high in accuracy, high in training speed and high in adaptability.
Drawings
Fig. 1 is a flowchart of a method for predicting faults in a digital twin plant based on random forest and analytic hierarchy process provided in embodiment 1.
Fig. 2 is a flowchart for obtaining a prediction result by using random forest and analytic hierarchy process as provided in example 2.
Fig. 3 is a schematic diagram of a machine tool temperature decision tree provided in example 2.
Fig. 4 is a schematic diagram of a tool wear condition decision tree provided in example 2.
Fig. 5 is a schematic diagram of a random forest as provided in example 2.
Fig. 6 is a block diagram of a digital twin plant fault prediction system based on random forest and analytic hierarchy process provided in embodiment 3.
Detailed Description
The drawings are for illustrative purposes only and are not to be construed as limiting the present patent;
For the purpose of better illustrating the embodiments, certain elements of the drawings may be omitted, enlarged or reduced and do not represent the actual product dimensions;
it will be appreciated by those skilled in the art that certain well-known structures in the drawings and descriptions thereof may be omitted.
The technical scheme of the invention is further described below with reference to the accompanying drawings and examples.
Example 1
As shown in fig. 1, the present embodiment provides a digital twin workshop fault prediction method based on random forest and analytic hierarchy process, which includes the following steps:
S1: acquiring production element data and real-time production data of a production workshop;
s2: constructing a digital twin model of the production workshop according to the production element data of the production workshop;
S3: carrying out digital twin processing on the real-time production data by using a digital twin model of a production workshop to obtain digital twin data of the production workshop;
S4: inputting digital twin data of a production workshop into a constructed predictive maintenance hierarchical structure model, obtaining a predictive maintenance hierarchical judgment matrix by utilizing a random forest algorithm, solving the predictive maintenance hierarchical judgment matrix, and obtaining a judgment matrix feature vector;
S5: performing fault prediction on the production workshop by using the judging matrix feature vector to obtain a fault prediction result of the production workshop;
S6: and controlling and managing the production activities of the production workshops according to the fault prediction results of the production workshops.
In the specific implementation process, firstly, production element data and real-time production data of a production workshop are obtained, a digital twin model of the production workshop is constructed according to the obtained production element data, the digital twin model of the production workshop is utilized to carry out digital twin processing on the real-time production data to obtain digital twin data of the production workshop, then the digital twin data of the production workshop is input into a constructed predictive maintenance hierarchical structure model, a predictive maintenance hierarchical judgment matrix is obtained by utilizing a random forest algorithm, a judgment matrix feature vector is obtained, fault prediction is carried out on the production workshop by utilizing the judgment matrix feature vector to obtain a fault prediction result of the production workshop, and finally, production activities of the production workshop are controlled and managed according to the fault prediction result of the production workshop;
The method can not only observe the running condition of workshop equipment in real time under the addition of a visual and real-time digital twin technology, but also directly observe the position of the fault when the fault occurs, thereby saving the time cost of equipment fault retrieval of maintenance personnel, facilitating the maintenance and reducing the working difficulty of the maintenance personnel; in addition, a random forest algorithm and an analytic hierarchy process are combined to conduct production workshop fault prediction, an objective and effective judgment matrix can be obtained, the prediction result is enabled to have confidence, and the method has the advantages of being high in accuracy, high in training speed and high in adaptability.
Example 2
The embodiment provides a digital twin workshop fault prediction method based on a random forest and analytic hierarchy process, which comprises the following steps:
S1: acquiring production element data and real-time production data of a production workshop;
s2: constructing a digital twin model of the production workshop according to the production element data of the production workshop;
S3: carrying out digital twin processing on the real-time production data by using a digital twin model of a production workshop to obtain digital twin data of the production workshop;
S4: as shown in fig. 2, inputting digital twin data of a production plant into a constructed predictive maintenance hierarchical structure model, obtaining a predictive maintenance hierarchical judgment matrix by using a random forest algorithm, and solving the predictive maintenance hierarchical judgment matrix to obtain a judgment matrix feature vector;
S5: predicting faults of the production workshop by utilizing the judging matrix feature vector to obtain a fault prediction result of the production workshop, uploading the fault prediction result of the production workshop to a database by using a predictive maintenance hierarchical structure model, interacting a digital twin model of the production workshop with the database, obtaining the fault prediction result and displaying the predicted fault position information;
S6: and controlling and managing the production activities of the production workshops according to the fault prediction results of the production workshops.
In the step S1, the production factor data of the production plant includes: device entity element data and information element data;
the equipment entity element data includes: geometric dimensions, physical mechanisms, behavior characteristics, position information and interaction relation among all devices in the production workshop;
the information element data includes: information on the status of various equipment and personnel in the production plant, task order processing information and scheduling decision information.
In the step S4, the constructed predictive maintenance hierarchical structure model specifically includes:
the predictive maintenance hierarchical structure model comprises a target layer, a criterion layer and a measure layer;
The target layer represents predictive maintenance of target equipment and is marked as A;
the criterion layer comprises a plurality of criterion layer elements, which are marked as B k and represent the kth criterion layer elements;
the measure layer comprises a plurality of measure layer elements, which are marked as C n and represent the nth measure layer element.
In the step S4, the specific method for obtaining the predictive maintenance level judgment matrix by using the random forest algorithm is as follows:
the random forest algorithm comprises a plurality of decision trees;
taking digital twin data of a production workshop as an original data set;
s4.1.1: randomly sampling the original data set for a plurality of times with a replacement to obtain a plurality of sub data sets, wherein the element number of each sub data set is consistent with the element number of the original data set;
S4.1.2: for each sub-data set, randomly selecting a plurality of features from all preset features, wherein each feature is used as an input feature of one decision tree to obtain a plurality of decision trees;
S4.1.3: setting a decision tree with a large information gain value at the top of a decision tree with a small information gain value to form a random forest;
S4.1.4: and obtaining a predictive maintenance level judgment matrix according to the output result of the random forest.
The predictive maintenance level judgment matrix in step S4.1.4 specifically includes:
For each criterion layer element, a predictive maintenance level judgment matrix is provided, and the predictive maintenance level judgment matrix corresponding to the kth criterion layer element is as follows:
Wherein C ij represents the importance degree of the ith action layer element to the kth criterion layer element compared with the jth action layer element;
and obtaining different values of C ij according to the output result of the random forest to form different predictive maintenance level judgment matrixes.
In the step S4, the specific method for obtaining the feature vector of the judgment matrix by solving the predictive maintenance level judgment matrix is as follows:
the method for solving the predictive maintenance level judgment matrix corresponding to the k criterion layer element comprises the following steps:
S4.2.1: squaring each element in the predictive maintenance level judgment matrix corresponding to the k criterion layer element to obtain a new judgment matrix C 2(Bk):
wherein i and j satisfy 1.ltoreq.i.ltoreq.n, 1.ltoreq.j.ltoreq.n;
S4.2.2: calculate the sum of all elements of each row in the new decision matrix C 2(Bk):
wherein sum i represents the sum of all elements of row i in the new decision matrix C 2(Bk);
s4.2.3: calculate the sum of all elements in the new decision matrix C 2(Bk):
Where sum represents the sum of all elements in the new decision matrix C 2(Bk);
S4.2.4: calculate a new decision matrix C 2(Bk) normalized vector of all element sums per row:
Wherein W i (q) represents the new judgment matrix C 2(Bk) calculated at the q-th time, the normalized vector of the sum of all elements of the i-th row;
S4.2.5: updating digital twin data of a production plant, executing steps S4.2.1-S4.2.4 again to obtain a new judgment matrix C 2(Bk) of each row of all element sums in a new round, and calculating error values of normalization vectors obtained in two consecutive rounds:
where T represents the error value of the normalized vector obtained twice in succession.
S4.2.6: comparing the error value T of the normalized vector obtained twice continuously with a preset threshold value; if the error value T is smaller than the preset threshold value, taking the normalized vector at the moment as the judgment matrix characteristic vector of the k criterion layer element, and marking asOtherwise, step S4.2.5 is performed until the error value T is less than the preset threshold.
In the step S5, the fault prediction is performed on the production workshop by using the feature vector of the judgment matrix, so as to obtain a fault prediction result of the production workshop, and the specific method comprises the following steps:
S5.1: solving the predictive maintenance level judgment matrix corresponding to each criterion layer element to obtain the judgment matrix characteristic vector of the measure layer relative to each criterion layer element, and marking as
S5.2: obtaining a predictive maintenance level judgment matrix of the criterion layer relative to the target layer, solving the predictive maintenance level judgment matrix, and obtaining a judgment matrix characteristic vector of the criterion layer relative to the target layer, wherein the judgment matrix characteristic vector is marked as W B-A;
S5.3: and according to the obtained feature vector of the judgment matrix, obtaining the final priority ranking of predictive maintenance, and taking the final priority ranking of predictive maintenance as a fault prediction result of a production workshop.
In the step S5.2, the predictive maintenance level judgment matrix of the criterion layer relative to the target layer is specifically:
Wherein B ab represents the importance level of the a-th criterion layer element to the target layer compared to the B-th criterion layer element.
In the step S5.3, a final priority ranking of predictive maintenance is obtained according to the obtained feature vector of the judgment matrix, and the specific method is as follows:
the feature vector of the measure layer relative to the target layer is obtained according to the following formula:
wherein W C-A represents a feature vector of the measure layer relative to the target layer;
And obtaining the importance degree of each measure layer element relative to the target layer according to the feature vector W C-A of the measure layer relative to the target layer, and sorting all the measure layer elements according to the importance degree to obtain the final priority sorting of predictive maintenance.
In a specific implementation process, the embodiment takes a numerical control machine tool in a production workshop as an example to describe the specific implementation process;
Firstly, acquiring production element data and real-time production data of a numerical control machine tool, wherein the production element data of the numerical control machine tool comprises: machine tool entity element data and machine tool information element data; the machine tool entity element data includes: parameter information such as geometric dimensions, physical mechanisms, behavior characteristics, position information and interaction relation among machine tools; the information element data of the machine tool comprises state information of the machine tool and task decision scheduling information of the machine tool;
Constructing a digital twin model of the numerical control machine according to the production element data of the numerical control machine;
Then, carrying out digital twin processing on real-time production data of the numerical control machine by using a digital twin model of the numerical control machine to obtain digital twin data of the numerical control machine, wherein the digital twin data of the numerical control machine comprises machine temperature and cutter abrasion conditions;
inputting digital twin data of a numerical control machine tool of a production workshop into a constructed predictive maintenance hierarchical structure model, wherein in the embodiment, the predictive maintenance hierarchical structure model specifically comprises:
the predictive maintenance hierarchical structure model comprises a target layer, a criterion layer and a measure layer;
The target layer represents predictive maintenance of target equipment and is marked as A;
the criterion layer comprises a plurality of criterion layer elements, which are marked as B k and represent the kth criterion layer elements;
In this embodiment, the criterion layer includes 4 criterion layer elements, where the first criterion layer element B 1 represents maintaining processing accuracy, the second criterion layer element B 2 represents saving running cost, the third criterion layer element B 3 represents reducing equipment damage, and the fourth criterion layer element B 4 represents avoiding casualties;
the measure layer comprises a plurality of measure layer elements, which are marked as C n and represent the nth measure layer element;
In this embodiment, the measure layer includes 5 criterion layer elements, where the first measure layer element C 1 represents a feed system, the second measure layer element C 2 represents a spindle system, the third measure layer element C 3 represents a tool machining system, the fourth measure layer element C 4 represents a CNC control system, and the fifth measure layer element C 5 represents an automatic tool changing system;
The predictive maintenance level judgment matrix is obtained by utilizing a random forest algorithm, and the specific method comprises the following steps:
the random forest algorithm comprises a plurality of decision trees;
taking digital twin data of a numerical control machine tool of a production workshop as an original data set;
s4.1.1: randomly sampling the original data set for a plurality of times with a replacement to obtain a plurality of sub data sets, wherein the element number of each sub data set is consistent with the element number of the original data set;
S4.1.2: for each sub-data set, randomly selecting a plurality of features from all preset features, wherein each feature is used as an input feature of one decision tree to obtain a plurality of decision trees;
In this embodiment, the preset features of the machine tool temperature include abnormal temperature and normal temperature, and the preset features of the tool wear condition include normal tool, slight tool wear, and excessive tool wear;
as shown in fig. 3 and 4, the machine tool temperature decision tree and the tool wear condition decision tree are respectively;
S4.1.3: as shown in fig. 5, a decision tree with a large information gain value is arranged at the top of a decision tree with a small information gain value to form a random forest, in this embodiment, the information gain value of the machine tool temperature is 0, and the information gain value of the tool wear condition is 0.92;
s4.1.4: obtaining a predictive maintenance level judgment matrix according to the output result of the random forest, wherein the predictive maintenance level judgment matrix comprises the following specific steps:
For each criterion layer element, a predictive maintenance level judgment matrix is provided, and the predictive maintenance level judgment matrix corresponding to the kth criterion layer element is as follows:
Wherein C ij represents the importance degree of the ith action layer element to the kth criterion layer element compared with the jth action layer element;
The greater the value of C ij satisfies 1.ltoreq.C ij≤9,Cij, the higher the importance of the ith action layer element to the kth criterion layer element compared with the jth action layer element;
different values of C ij are obtained according to the output result of the random forest to form different predictive maintenance level judgment matrixes;
Solving a predictive maintenance level judgment matrix to obtain a judgment matrix feature vector, wherein the specific method comprises the following steps of:
the method for solving the predictive maintenance level judgment matrix corresponding to the k criterion layer element comprises the following steps:
S4.2.1: squaring each element in the predictive maintenance level judgment matrix corresponding to the k criterion layer element to obtain a new judgment matrix C 2(Bk):
wherein i and j satisfy 1.ltoreq.i.ltoreq.5, 1.ltoreq.j.ltoreq.5;
S4.2.2: calculate the sum of all elements of each row in the new decision matrix C 2(Bk):
wherein sum i represents the sum of all elements of row i in the new decision matrix C 2(Bk);
s4.2.3: calculate the sum of all elements in the new decision matrix C 2(Bk):
Where sum represents the sum of all elements in the new decision matrix C 2(Bk);
S4.2.4: calculate a new decision matrix C 2(Bk) normalized vector of all element sums per row:
Wherein W i (q) represents the new judgment matrix C 2(Bk) calculated at the q-th time, the normalized vector of the sum of all elements of the i-th row;
S4.2.5: updating digital twin data of a production plant, executing steps S4.2.1-S4.2.4 again to obtain a new judgment matrix C 2(Bk) of each row of all element sums in a new round, and calculating error values of normalization vectors obtained in two consecutive rounds:
where T represents the error value of the normalized vector obtained twice in succession.
S4.2.6: comparing the error value T of the normalized vector obtained twice continuously with a preset threshold value; if the error value T is smaller than the preset threshold value, taking the normalized vector at the moment as the judgment matrix characteristic vector of the k criterion layer element, and marking asOtherwise, executing step S4.2.5 until the error value T is smaller than a preset threshold, where in this embodiment, the preset threshold is a constant in the debugging process of the numerically-controlled machine tool;
The fault prediction is carried out on the production workshop by utilizing the feature vector of the judgment matrix, and the fault prediction result of the production workshop is obtained, and the specific method comprises the following steps:
S5.1: solving the predictive maintenance level judgment matrix corresponding to each criterion layer element to obtain the judgment matrix characteristic vector of the measure layer relative to each criterion layer element, and marking as
S5.2: the method comprises the steps of obtaining a predictive maintenance level judgment matrix of a criterion layer relative to a target layer, wherein the predictive maintenance level judgment matrix specifically comprises the following steps:
/>
Wherein B ab represents the importance level of the a-th criterion layer element to the target layer compared with the B-th criterion layer element;
solving a predictive maintenance level judgment matrix of the criterion layer relative to the target layer, and obtaining a judgment matrix characteristic vector of the criterion layer relative to the target layer, which is marked as W B-A;
s5.3: according to the obtained judgment matrix feature vector, obtaining the final priority ranking of predictive maintenance, specifically:
the feature vector of the measure layer relative to the target layer is obtained according to the following formula:
wherein W C-A represents a feature vector of the measure layer relative to the target layer;
Obtaining importance degrees of elements of each measure layer relative to the target layer according to feature vectors W C-A of the measure layer relative to the target layer, and sorting all the elements of the measure layer according to the importance degrees to obtain final priority sorting of predictive maintenance;
Sequencing the final priority of predictive maintenance as a fault prediction result of the production workshop;
finally, uploading a fault prediction result of the production workshop to a database by the predictive maintenance hierarchical structure model, interacting the digital twin model of the production workshop with the database, acquiring the fault prediction result and displaying the predicted fault position information;
controlling and managing production activities of the production workshops according to the fault prediction results of the production workshops;
The method can not only observe the running condition of workshop equipment in real time under the addition of a visual and real-time digital twin technology, but also directly observe the position of the fault when the fault occurs, thereby saving the time cost of equipment fault retrieval of maintenance personnel, facilitating the maintenance and reducing the working difficulty of the maintenance personnel; in addition, a random forest algorithm and an analytic hierarchy process are combined to conduct production workshop fault prediction, an objective and effective judgment matrix can be obtained, the prediction result is enabled to have confidence, and the method has the advantages of being high in accuracy, high in training speed and high in adaptability.
Example 3
As shown in fig. 6, the present embodiment provides a digital twin plant fault prediction system based on random forest and analytic hierarchy process, including:
The data acquisition unit 301: the method comprises the steps of obtaining production element data and real-time production data of a production workshop;
digital twin model construction unit 302: the digital twin model is used for constructing a production workshop according to the production factor data of the production workshop;
Digital twin data acquisition unit 303: carrying out digital twin processing on the real-time production data by using a digital twin model of a production workshop to obtain digital twin data of the production workshop;
Judgment matrix calculation unit 304: the method comprises the steps of inputting digital twin data of a production workshop into a constructed predictive maintenance hierarchical structure model, obtaining a predictive maintenance hierarchical judgment matrix by utilizing a random forest algorithm, solving the predictive maintenance hierarchical judgment matrix, and obtaining a judgment matrix feature vector;
The failure prediction unit 305: the method comprises the steps of performing fault prediction on a production workshop by utilizing a judgment matrix feature vector to obtain a fault prediction result of the production workshop;
prediction result output unit 306: and the production activity of the production workshop is controlled and managed according to the fault prediction result of the production workshop.
In a specific implementation process, firstly, a data acquisition unit 301 acquires production element data and real-time production data of a production workshop, a digital twin model construction unit 302 constructs a digital twin model of the production workshop according to the acquired production element data, a digital twin data acquisition unit 303 performs digital twin processing on the real-time production data by using the digital twin model of the production workshop to acquire digital twin data of the production workshop, a judgment matrix calculation unit 304 inputs the digital twin data of the production workshop into a constructed predictive maintenance hierarchical structure model, a predictive maintenance hierarchical judgment matrix is acquired by using a random forest algorithm, a predictive maintenance hierarchical judgment matrix is solved, a judgment matrix feature vector is acquired, a fault prediction unit 305 performs fault prediction on the production workshop by using the judgment matrix feature vector to acquire a fault prediction result of the production workshop, and finally a prediction result output unit 306 performs control management on production activities of the production workshop according to the fault prediction result of the production workshop;
Under the addition of a visual and real-time digital twin technology, the system can not only observe the running condition of workshop equipment in real time, but also directly observe the position of the fault when the fault occurs, thereby saving the time cost of equipment fault retrieval of maintenance personnel, facilitating maintenance, and reducing the working difficulty of the maintenance personnel; in addition, a random forest algorithm and an analytic hierarchy process are combined to conduct production workshop fault prediction, an objective and effective judgment matrix can be obtained, the prediction result is enabled to have confidence, and the system has the advantages of being high in accuracy, high in training speed and high in adaptability.
The same or similar reference numerals correspond to the same or similar components;
the terms describing the positional relationship in the drawings are merely illustrative, and are not to be construed as limiting the present patent;
It is to be understood that the above examples of the present invention are provided by way of illustration only and not by way of limitation of the embodiments of the present invention. Other variations or modifications of the above teachings will be apparent to those of ordinary skill in the art. It is not necessary here nor is it exhaustive of all embodiments. Any modification, equivalent replacement, improvement, etc. which come within the spirit and principles of the invention are desired to be protected by the following claims.
Claims (7)
1. A digital twin workshop fault prediction method based on random forest and analytic hierarchy process is characterized by comprising the following steps:
S1: acquiring production element data and real-time production data of a production workshop;
s2: constructing a digital twin model of the production workshop according to the production element data of the production workshop;
S3: carrying out digital twin processing on the real-time production data by using a digital twin model of a production workshop to obtain digital twin data of the production workshop;
S4: inputting digital twin data of a production plant into a constructed predictive maintenance hierarchical structure model,
The constructed predictive maintenance hierarchical structure model is specifically:
the predictive maintenance hierarchical structure model comprises a target layer, a criterion layer and a measure layer;
The target layer represents predictive maintenance of target equipment and is marked as A;
the criterion layer comprises a plurality of criterion layer elements, which are marked as B k and represent the kth criterion layer elements;
the measure layer comprises a plurality of measure layer elements, which are marked as C n and represent the nth measure layer element;
Obtaining a predictive maintenance level judgment matrix by utilizing a random forest algorithm, solving the predictive maintenance level judgment matrix, and obtaining a judgment matrix feature vector;
the step of obtaining the predictive maintenance level judgment matrix by utilizing the random forest algorithm comprises the following steps:
the random forest algorithm comprises a plurality of decision trees;
taking digital twin data of a production workshop as an original data set;
s4.1.1: randomly sampling the original data set for a plurality of times with a replacement to obtain a plurality of sub data sets, wherein the element number of each sub data set is consistent with the element number of the original data set;
S4.1.2: for each sub-data set, randomly selecting a plurality of features from all preset features, wherein each feature is used as an input feature of one decision tree to obtain a plurality of decision trees;
S4.1.3: setting a decision tree with a large information gain value at the top of a decision tree with a small information gain value to form a random forest;
S4.1.4: obtaining a predictive maintenance level judgment matrix according to the output result of the random forest;
the step of solving the predictive maintenance level judgment matrix corresponding to the k criterion layer element is as follows:
S4.2.1: squaring each element in the predictive maintenance level judgment matrix corresponding to the k criterion layer element to obtain a new judgment matrix C 2(Bk):
wherein i and j satisfy 1.ltoreq.i.ltoreq.n, 1.ltoreq.j.ltoreq.n;
S4.2.2: calculate the sum of all elements of each row in the new decision matrix C 2(Bk):
wherein sum i represents the sum of all elements of row i in the new decision matrix C 2(Bk);
s4.2.3: calculate the sum of all elements in the new decision matrix C 2(Bk):
Where sum represents the sum of all elements in the new decision matrix C 2(Bk);
S4.2.4: calculate a new decision matrix C 2(Bk) normalized vector of all element sums per row:
Wherein W i (q) represents the new judgment matrix C 2(Bk) calculated at the q-th time, the normalized vector of the sum of all elements of the i-th row;
S4.2.5: updating digital twin data of a production plant, executing steps S4.2.1-S4.2.4 again to obtain a new judgment matrix C 2(Bk) of each row of all element sums in a new round, and calculating error values of normalization vectors obtained in two consecutive rounds:
wherein T represents the error value of the normalized vector obtained twice in succession;
s4.2.6: comparing the error value T of the normalized vector obtained twice continuously with a preset threshold value; if the error value T is smaller than the preset threshold value, taking the normalized vector at the moment as the judgment matrix characteristic vector of the k criterion layer element, and marking as Otherwise, step S4.2.5 is executed until the error value T is less than the preset threshold;
S5: performing fault prediction on the production workshop by using the judging matrix feature vector to obtain a fault prediction result of the production workshop;
S6: and controlling and managing the production activities of the production workshops according to the fault prediction results of the production workshops.
2. The method for predicting digital twin plant failure based on random forest and analytic hierarchy process of claim 1, wherein in step S1, the production factor data of the production plant comprises: device entity element data and information element data;
the equipment entity element data includes: geometric dimensions, physical mechanisms, behavior characteristics, position information and interaction relation among all devices in the production workshop;
the information element data includes: information on the status of various equipment and personnel in the production plant, task order processing information and scheduling decision information.
3. The method for predicting digital twin plant failure based on random forest and analytic hierarchy process of claim 1, wherein step S5 further comprises:
the predictive maintenance hierarchical structure model uploads the fault prediction result of the production workshop to the database, the digital twin model of the production workshop interacts with the database, the fault prediction result is obtained, and the predicted fault position information is displayed.
4. The method for predicting a digital twin plant fault based on random forest and analytic hierarchy process of claim 1, wherein the predictive maintenance analytic hierarchy process of step S4.1.4 is specifically:
For each criterion layer element, a predictive maintenance level judgment matrix is provided, and the predictive maintenance level judgment matrix corresponding to the kth criterion layer element is as follows:
Wherein C ij represents the importance degree of the ith action layer element to the kth criterion layer element compared with the jth action layer element;
and obtaining different values of C ij according to the output result of the random forest to form different predictive maintenance level judgment matrixes.
5. The method for predicting the fault of the digital twin workshop based on the random forest and analytic hierarchy process of claim 1, wherein in the step S5, the fault prediction is performed on the production workshop by using the feature vector of the judgment matrix, so as to obtain the result of the fault prediction of the production workshop, and the specific method comprises the following steps:
S5.1: solving the predictive maintenance level judgment matrix corresponding to each criterion layer element to obtain the judgment matrix characteristic vector of the measure layer relative to each criterion layer element, and marking as
S5.2: obtaining a predictive maintenance level judgment matrix of the criterion layer relative to the target layer, solving the predictive maintenance level judgment matrix, and obtaining a judgment matrix characteristic vector of the criterion layer relative to the target layer, wherein the judgment matrix characteristic vector is marked as W B-A;
S5.3: and according to the obtained feature vector of the judgment matrix, obtaining the final priority ranking of predictive maintenance, and taking the final priority ranking of predictive maintenance as a fault prediction result of a production workshop.
6. The method for predicting digital twin plant faults based on random forest and analytic hierarchy process as claimed in claim 5, wherein in the step S5.2, the predictive maintenance level judgment matrix of the criterion layer relative to the target layer is specifically:
Wherein B ab represents the importance level of the a-th criterion layer element to the target layer compared to the B-th criterion layer element.
7. The method for predicting the fault of the digital twin workshop based on the random forest and analytic hierarchy process of claim 5, wherein in step S5.3, the final priority order of predictive maintenance is obtained according to the obtained feature vector of the judgment matrix, and the specific method is as follows:
the feature vector of the measure layer relative to the target layer is obtained according to the following formula:
wherein W C-A represents a feature vector of the measure layer relative to the target layer;
And obtaining the importance degree of each measure layer element relative to the target layer according to the feature vector W C-A of the measure layer relative to the target layer, and sorting all the measure layer elements according to the importance degree to obtain the final priority sorting of predictive maintenance.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211457614.3A CN115718466B (en) | 2022-11-21 | 2022-11-21 | Digital twin workshop fault prediction method based on random forest and analytic hierarchy process |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211457614.3A CN115718466B (en) | 2022-11-21 | 2022-11-21 | Digital twin workshop fault prediction method based on random forest and analytic hierarchy process |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115718466A CN115718466A (en) | 2023-02-28 |
| CN115718466B true CN115718466B (en) | 2024-05-14 |
Family
ID=85255729
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211457614.3A Active CN115718466B (en) | 2022-11-21 | 2022-11-21 | Digital twin workshop fault prediction method based on random forest and analytic hierarchy process |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115718466B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116224905B (en) * | 2023-03-02 | 2023-10-20 | 广东工业大学 | Machine tool thermal error prediction method and system based on joint distribution self-adaption |
| CN116861204B (en) * | 2023-09-05 | 2023-12-08 | 山东山森数控技术有限公司 | Intelligent manufacturing equipment data management system based on digital twinning |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114311609A (en) * | 2022-01-17 | 2022-04-12 | 安徽理工大学 | Extruding machine fault diagnosis system based on digital twinning |
| CN114519302A (en) * | 2022-01-25 | 2022-05-20 | 江苏科技大学 | Road traffic situation simulation method based on digital twin |
| CN114563988A (en) * | 2022-01-26 | 2022-05-31 | 浙江中控信息产业股份有限公司 | Water plant intelligent PAC adding method and system based on random forest algorithm |
| CN115081508A (en) * | 2022-04-26 | 2022-09-20 | 同济大学 | Traffic running risk parallel simulation system based on traffic digital twins |
| CN115329493A (en) * | 2022-08-17 | 2022-11-11 | 兰州理工大学 | Impeller mechanical fault detection method based on centrifugal pump digital twin model |
-
2022
- 2022-11-21 CN CN202211457614.3A patent/CN115718466B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114311609A (en) * | 2022-01-17 | 2022-04-12 | 安徽理工大学 | Extruding machine fault diagnosis system based on digital twinning |
| CN114519302A (en) * | 2022-01-25 | 2022-05-20 | 江苏科技大学 | Road traffic situation simulation method based on digital twin |
| CN114563988A (en) * | 2022-01-26 | 2022-05-31 | 浙江中控信息产业股份有限公司 | Water plant intelligent PAC adding method and system based on random forest algorithm |
| CN115081508A (en) * | 2022-04-26 | 2022-09-20 | 同济大学 | Traffic running risk parallel simulation system based on traffic digital twins |
| CN115329493A (en) * | 2022-08-17 | 2022-11-11 | 兰州理工大学 | Impeller mechanical fault detection method based on centrifugal pump digital twin model |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115718466A (en) | 2023-02-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN115718466B (en) | Digital twin workshop fault prediction method based on random forest and analytic hierarchy process | |
| Nurprihatin et al. | Total productive maintenance policy to increase effectiveness and maintenance performance using overall equipment effectiveness | |
| Sharma et al. | A literature review and future perspectives on maintenance optimization | |
| CN109902954B (en) | Flexible job shop dynamic scheduling method based on industrial big data | |
| CN106682781A (en) | Power equipment multi-index prediction method | |
| US7966151B2 (en) | Method for analyzing operation of a machine | |
| DE102016015332A1 (en) | A preventive maintenance management system and method for creating a maintenance plan of a machine and cell controller | |
| CN107862324B (en) | MWSPCA-based CBR prediction model intelligent early warning method | |
| Bastos et al. | A maintenance prediction system using data mining techniques | |
| CN114118673A (en) | Workshop intelligent fault diagnosis early warning method based on digital twin technology | |
| CN113469241A (en) | Product quality control method based on process network model and machine learning algorithm | |
| CN117389236B (en) | Propylene oxide production process optimization method and system | |
| Aleš et al. | Methodology of overall equipment effectiveness calculation in the context of industry 4.0 environment | |
| CN114749996A (en) | Tool residual life prediction method based on deep learning and time sequence regression model | |
| CN114418177B (en) | New product material distribution prediction method based on digital twin workshops for generating countermeasure network | |
| Kodali et al. | Performance value analysis for the justification of world-class manufacturing systems | |
| Ståhl et al. | Dynamic manufacturing costs—Describing the dynamic behavior of downtimes from a cost perspective | |
| Kikolski | Determination of ISO 22400 Key Performance Indicators using Simulation Models: The Concept and Methodology. | |
| CN108664696B (en) | Method and device for evaluating running state of water chiller | |
| CN111091292A (en) | Real-time risk dynamic modeling analysis system for nuclear power station | |
| CN115688016A (en) | Integrated method for intelligent operation and maintenance of large database | |
| Wittbrodt et al. | Identification of the impact of the availability factor on the efficiency of production processes using the AHP and fuzzy AHP | |
| Dobra et al. | Correlation between Overall Equipment Effectiveness (OEE) and Failure Mode and Effect Analysis (FMEA) at the semi-automatic assembly lines | |
| Noterdaeme et al. | Mapping Heavy Industry's Digital-Manufacturing Opportunities | |
| CN116551467B (en) | Cutter abrasion state identification method based on internet of things monitoring data |
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 |