Production workshop abnormal reason tracing method based on digital twinning
By constructing a digital twin-based production workshop anomaly tracing model and utilizing Bayesian networks and decision trees for data-driven anomaly tracing, the accuracy and efficiency issues of anomaly tracing in complex product production workshops are solved, achieving more efficient anomaly handling.
Patent Information
- Application Number
- CN202410966887.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-18
- Publication Date
- 2024-11-15
AI Technical Summary
In complex product manufacturing workshops, existing technologies struggle to trace the causes of production anomalies in a timely and accurate manner, resulting in low efficiency in handling abnormal events.
A production workshop anomaly tracing model based on digital twins is constructed, including a physical production workshop, a virtual production workshop, a full-element twin data platform, and a production anomaly event tracing service system. Bayesian networks and decision trees are used for data-driven anomaly tracing, and anomaly handling is optimized by comparing the simulation results of the virtual production workshop with the actual results.
It improves the accuracy and speed of tracing the causes of abnormal events in the production workshop, ensuring the stability and controllability of the production process.
Smart Images

Figure CN118966529A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of production abnormality tracing, and in particular to a method for tracing the causes of abnormalities in a production workshop based on digital twins. Background Art
[0002] With the transformation of the manufacturing industry driven by technological innovation, the rapid development of technologies such as the Internet of Things and cloud computing, and their continuous penetration in the manufacturing industry, the production process has evolved from the traditional black box model to a multi-dimensional, transparent and intelligent model. By using intelligent technology, the production workshop has strengthened the control of the entire production process, greatly improved the level of production management, and reduced production costs. With the promotion of intelligent manufacturing technology, higher requirements are put forward for the timely and accurate acquisition of information in the manufacturing system. Whether the manufacturing system can trace and feedback the manufacturing process information in real time has become a bottleneck for workshop and even enterprise management. The complex product production workshop belongs to a discrete manufacturing workshop. There are too many factors affected by abnormalities in the production process. It is inevitable that production abnormalities such as processing timeouts, delayed arrival of work-in-progress, and failure to deliver materials on time will occur. After the production abnormality occurs, the on-site management personnel use the instructions and abnormality resolution experience as the basis for speculation, and there is a situation where the efficiency of handling production abnormalities is low. Therefore, in the production process of complex products, how to trace the abnormal causes in a timely and accurate manner and feedback the tracing results is one of the important issues that enterprises need to solve urgently.
[0003] With the deepening of research, the tracing of abnormal causes is no longer limited to a single production factor, but has been expanded to the entire production process. When the manufacturing system identifies an abnormal production performance event, the abnormal cause is obtained from certain rules according to the type of abnormal event. At present, the solutions to the problem of intelligent tracing of abnormal causes in the production process can be divided into three categories: analytical model-based, signal processing-based, and knowledge-based. The knowledge-based method is a combination of artificial intelligence and abnormal tracing technology, and is currently a more respected method. Common knowledge-based production abnormality intelligent tracing methods include those based on artificial neural networks, support vector machines, decision trees, Bayesian networks, and intelligent tracing methods based on Petri nets. In addition, digital twin technology has been explored and applied in the twin construction, technical architecture, visual monitoring, and abnormal cause tracing integration of production workshops. The present invention is aimed at a production abnormal event tracing method for complex product production workshops, establishes a complex product production abnormal event tracing model based on digital twins, uses workshop twin data to realize the analysis and tracing of abnormal events in the production workshop, and provides real-time information for the production workshop. Summary of the invention
[0004] In order to solve the above technical problems, the present invention proposes a method for tracing the causes of abnormalities in production workshops based on digital twins. In the method, all-factor production abnormalities in the production workshop can be intelligently traced, providing a reference value for the application of digital twins in production workshops.
[0005] In order to achieve the above object, the technical solution of the present invention is as follows:
[0006] A method for tracing abnormal causes of production workshops based on digital twins includes the following steps:
[0007] Construct a digital twin production workshop abnormality cause tracing model, the production workshop abnormality cause tracing model includes a physical production workshop, a virtual production workshop, a full-factor twin data platform and a production abnormal event cause tracing service system, wherein the physical production workshop and the virtual production workshop interact, and the full-factor twin data platform is used to drive the virtual model in the virtual production workshop to perform iterative optimization, and the production abnormal event cause tracing service system is used to trace the abnormal cause based on the data-driven production abnormality tracing Bayesian network and the abnormal cause tracing decision tree and display the tracing results;
[0008] Create a production abnormality cause tracing mechanism: the production abnormal event cause tracing service system calls the full-factor twin data platform to trace the abnormal events in the physical production workshop to obtain the tracing results; the virtual production workshop simulates the data of the physical production workshop to obtain the simulation results, and compares the tracing results with the simulation results. If the comparison results are consistent, an abnormality sorting solution is provided to the physical production workshop based on the tracing results; if the comparison results are inconsistent, the full-factor twin data platform is optimized.
[0009] Preferably, the virtual model in the virtual production workshop includes a personnel model, a material model, a method model, an equipment model and an environment model.
[0010] Preferably, the full-factor twin data platform includes physical data, virtual data, a knowledge base, an intelligent algorithm library, a production database and a model library.
[0011] Preferably, the production abnormal event cause tracing service system includes a production abnormality tracing Bayesian network, an abnormality cause tracing decision tree and a visualization tool, wherein the production abnormality tracing Bayesian network is used to reflect the probabilistic relationship between abnormal events and abnormal influencing factors; the abnormality cause tracing decision tree is constructed based on the historical abnormal data statistically collected from the physical production workshop and the constructed Bayesian network, each decision node, state node and result node in the abnormality cause tracing decision tree corresponds to the decision node, state node and benefit node of the Bayesian network respectively, the same event is merged into one node, each path in the abnormality cause tracing decision tree is converted into a cause tracing rule, and an abnormality tracing rule base is obtained, and the abnormality tracing rule base is stored in the knowledge base; the visualization tool is used to display the tracing results.
[0012] Based on the above technical scheme, the beneficial effects of the present invention are as follows: the present invention provides a method for tracing the causes of abnormalities in a production workshop based on digital twins, and constructs a tracing model for the causes of abnormalities in a production workshop based on digital twins, wherein the tracing model for the causes of abnormalities in the production workshop includes a physical production workshop, a virtual production workshop, a full-factor twin data platform, and a tracing service system for the causes of abnormal production events, wherein the physical production workshop interacts with the virtual production workshop, and the full-factor twin data platform is used to drive the virtual model in the virtual production workshop to perform iterative optimization, and the tracing service system for the causes of abnormal production events is used to trace the causes of abnormal production events based on a data-driven Bayesian network for tracing the causes of abnormal production events and to display the tracing results; a tracing mechanism for the causes of abnormal production events is created: the tracing service system for the causes of abnormal production events calls the full-factor twin data platform to trace the causes of abnormal events in the physical production workshop and obtain tracing results; the virtual production workshop simulates the data of the physical production workshop to obtain simulation results, and compares the tracing results with the simulation results. If the comparison results are consistent, an abnormal combing solution is provided to the physical production workshop based on the tracing results; if the comparison results are inconsistent, the full-factor twin data platform is optimized. The present invention improves the accuracy of tracing the causes of abnormal events in a production workshop and also improves the execution speed to a certain extent. BRIEF DESCRIPTION OF THE DRAWINGS
[0013] Figure 1 is a structural schematic diagram of a production workshop abnormality cause tracing model in a production workshop abnormality cause tracing method based on digital twin in an embodiment;
[0014] Figure 2 It is a flow chart of a production abnormality cause tracing mechanism in a production workshop abnormality cause tracing method based on digital twin in an embodiment;
[0015] Figure 3 is a graph of association between abnormal events and abnormal factors in one embodiment;
[0016] Figure 4 is an embodiment of producing an abnormal Bayesian network distribution map;
[0017] Figure 5 The figure is a comparative staining diagram of abnormal cause tracing results of an experiment in an embodiment. DETAILED DESCRIPTION
[0018] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the accompanying drawings in the embodiments of the present invention.
[0019] This embodiment provides a method for tracing abnormal causes of a production workshop based on digital twins, including the following steps:
[0020] Step 1: Build a digital twin production workshop abnormality cause tracing model. The production workshop abnormality cause tracing model consists of four parts: physical production workshop, virtual production workshop, full-factor twin data platform and production abnormal event cause tracing service system. Figure 1 .
[0021] The physical production workshop is a production space unit located in the production workshop. It is set up according to the various stages of complex product production and various auxiliary production activities, and has entities such as complex products and their manufacturing process resource ends, such as production personnel, management personnel, processing equipment, auxiliary tools, materials, RFID and sensors. The physical production workshop includes the entire production process of complex products from raw materials to finished products, during which all factors are integrated to provide relevant status information and manufacturing data for the virtual production workshop and the production workshop twin data platform.
[0022] The virtual production workshop is a mapping of the physical production workshop entity and the production process, and includes a collection of models at the three levels of elements, behaviors, and rules. Based on the twin data platform of the complex product production workshop, by building a virtual production workshop, the attributes, states, and relationships of personnel, materials, methods, equipment, and production plans in the physical production workshop are timely and accurately mapped. The virtual model of the virtual production workshop includes personnel models, material models, method models, equipment models, and environmental models, etc., which simulate the pre-operation strategies and internal logical relationships of personnel, equipment, and production processes. Driven by the twin model of the workshop and the twin data, production-related activities such as production factor management and cause tracing of production abnormalities in the physical workshop are continuously updated and improved.
[0023] The full-factor twin data platform includes physical data, virtual data, knowledge base, intelligent algorithm library, production database and model library. Physical data mainly refers to the real-time production data of the physical production workshop. Virtual data mainly refers to the simulation data and virtual model data of the virtual production workshop. The knowledge base contains rule information, simulation logic, etc., providing knowledge support for digital twin modeling and simulation analysis. The intelligent algorithm library provides support for knowledge reasoning, intelligent optimization algorithms, such as Bayesian network algorithm and C5.0 decision tree algorithm and the update of their rules. The production database contains complex product machining process data, production abnormality history data, etc. The model library provides templates for the virtual production workshop, instantiates and digitizes the production process and its production factors. Through the intelligent IoT device, the real-time manufacturing data of the personnel, materials, methods, equipment, environment, etc. of the physical production workshop is perceived, and then converted into corresponding production events, while continuously updating and expanding its own data. Through the simulation data of the virtual production workshop, the difference with the physical production workshop is compared to drive the optimization of the virtual model in the virtual production workshop. By analyzing the service data of the production abnormality event cause tracing service system, it provides intelligent algorithm support for it.
[0024] The production abnormal event cause tracing service system is a combination of data-driven Bayesian network and decision tree tracing technology. It mainly includes modules such as causal relationship between abnormal events and abnormal influencing factors, construction of abnormal cause tracing decision tree, rule matching cause tracing to obtain abnormal causes, result feedback and rule update.
[0025] Step 2: Create a production abnormality cause tracing mechanism
[0026] With the help of digital twin technology, twin production workshops (physical production workshops, virtual production workshops) can ensure the transparency and sharing of workshop information. The production abnormal event cause tracing service system can realize the functions of tracing the abnormal causes of personnel, materials, and production progress based on the full-factor twin data platform. The digital twin-driven production workshop abnormal cause tracing operation mechanism, such as Figure 2 shown.
[0027] In the production process, by reasoning and judging the production abnormal events, the influencing causes can be traced, effectively improving the stability and controllability of the entire production process. First, the real-time resource status of the physical production workshop and the model data and simulation data of the virtual production workshop are transmitted to the full-factor twin data platform. The platform converts the data into corresponding production events, inputs them into the production abnormal event diagnosis service system, and provides support such as knowledge base and intelligent algorithm library. Secondly, according to the correlation between abnormal events and abnormal factors, a Bayesian network is constructed, and a production abnormality C5.0 decision tree is established, and then the corresponding abnormal tracing rule library is extracted. Through instance matching rules, the abnormal causes are traced and inferred. Finally, the cause tracing results are simulated and verified to verify whether the cause tracing results are consistent with the simulation results. If the consistency test is no, the cause tracing service system is returned to continuously update the abnormal tracing rule library; if the consistency test is yes, real-time information is provided to the abnormality handling center. The abnormality handling center can re-allocate resources according to the feedback of the cause tracing results, and make corresponding adjustments to the workshop environment, processing equipment, processing personnel, processing materials, processing technology, etc. in a timely manner, and feedback bottleneck processes and weak links to the physical production workshop to prevent the same abnormality from happening again.
[0028] Through the establishment of the abnormal cause tracing model and abnormal cause tracing mechanism of the production workshop, combined with the status information of "people, machines, materials, methods, and environment" in the workshop, the abnormal cause tracing of the production workshop needs to regard the complex system as a whole, focus on the local phenomena, and then apply it to the entire multi-level, multi-factor, and multi-variable production process. Figure 3 These are the common abnormal events and abnormal factors in the production workshop. Tracing the causes of production abnormalities belongs to the problem of a posteriori probability, that is, under the premise of a known conclusion, inferring the cause of the conclusion. The specific modeling process and calculation steps are as follows:
[0029] Step 1: For each event in the production abnormality event sequence, create a node in the Bayesian network and name it according to the event name. For repeated events, only one node is created. Y is called a Bayesian network relative to a directed acyclic graph G, and the nodes represent random variables. Let Y = (Ya), a∈I is the random variable represented by a node a in its directed acyclic graph,
[0030] Ya={Y1,Y2,…,Y n} (1)
[0031] Step 2: For the influencing factors, establish a child node and determine the state space of the node according to the number of factors. The arrow connecting two nodes represents that the two random variables have a causal relationship or are unconditionally independent. Assuming that node F directly affects node H, that is, F→H, a directed arc (F, H) from node F to node H is established with an arrow from E to H, and the weight (that is, the connection strength) is represented by the conditional probability P(H|F), as shown in the following figure:
[0032]
[0033] Step 3: Connect each node with directed arcs to form the topological structure of the Bayesian network, and obtain a directed acyclic Bayesian network graph. Let G = (I, E) represent a directed acyclic graph (DAG), where I represents the set of all nodes in the graph and E represents the set of directed connecting segments.
[0034] Step 4: Determine the prior probability of the parent node and the conditional probability of each child node based on the sequence of production abnormal events and the probability of occurrence. For any random variable x, p(a) represents the "cause" of node a, or p(a) is the parents of a, and the joint probability of node Y can be expressed as:
[0035]
[0036] Their joint probability can be obtained by multiplying their respective local conditional probability distributions:
[0037] p(x1,…,x k )=p(x k |x1,…,x k-1 )…p(x2|x1)p(x1) (4)
[0038] Based on the abnormal list of the physical production workshop, the real-time production abnormal events are identified as top events, and the abnormal factors that cause the abnormal events are decomposed into intermediate events and bottom events step by step, so as to establish a production abnormality tracing Bayesian network, such as Figure 4 .
[0039] Then, according to the known conditions and logical relationships in the Bayesian network, each decision node, state node and result node in the decision tree corresponds to the decision node, state node and benefit node of the Bayesian network respectively, and the nodes in the Bayesian network correspond to the decision nodes in the decision tree. The top event corresponds to the root node, and the middle event and the bottom event correspond to the leaf node. The same event can be merged into one node. Recursively select the extended attributes for each subset of the divided samples for division until the stop condition is met, and finally generate the abnormal cause tracing decision tree.
[0040] ①Select extended attributes (select the root node of the tree)
[0041] The sample set of production abnormal events is X = {e1, e2, ..., e n}. For each attribute A (k) , 1≤k≤n. for A (k) Each attribute value T i (k) , i=1,2,…,m k , calculate its relative to category T j (n+1) , j = 1, 2, ..., m relative frequency P ij (k) , as follows:
[0042] P ij (k) =M(T i (k) ∩T j (n+1) ) / M(T i (k) ) (5)
[0043] Where i, j are attribute value indexes, m is the number of attribute values of the n+1th attribute, and m k is the number of attribute values of the kth attribute, and the symbol M(A) represents the sum of all membership degrees of set A.
[0044] A (k) Each attribute value T i (k) , i=1,2,…,m k , calculate its classification entropy:
[0045]
[0046] For each attribute, calculate the average classification entropy:
[0047]
[0048] Select E k Take the minimum attribute as the extended attribute, that is, select k0 so that
[0049]
[0050] ② Split sample set
[0051] β A (c) represents the possibility that A belongs to C, u(A)e and u(C)e represent the importance of sample e to nodes A and C respectively, and the calculation formula is:
[0052]
[0053] The truth of a node is the maximum possibility that the sample in the node belongs to each class. Recursively calculate their average classification entropy, select the expansion node, and finally generate a decision tree.
[0054] ③ Obtain the abnormal cause tracing rule library
[0055] Each path from the root node to the leaf node in the decision tree can be converted into a cause tracing rule:
[0056] Rule 1, IF Attri1=A, Then AL=…;…; Rule n, IF Attri1=H, Then AL=….
[0057] Cause tracing rules are generated from the decision tree, and the number of rule base rules is obtained according to the number of leaf nodes of the decision tree.
[0058] experiment
[0059] The computational efficiency and accuracy of abnormal cause tracing using the traditional decision tree (hereinafter referred to as the decision tree model) and the Bayesian network for production abnormality tracing and the decision tree for abnormal cause tracing (hereinafter referred to as the joint model) in this invention are compared. The data in the data set are trained using SPSS software, and the instance data is used as the test group to obtain the abnormal event cause tracing of the key workstations of the two cause tracing models. In order to more clearly reflect the difference between the cause tracing results of the decision tree model and the joint model, the actual results, the cause tracing results of the decision tree model, and the cause tracing results of the joint model are respectively mapped to Figure 5 In. By Figure 5 It can be seen that in the 5th group, the actual abnormal cause of the instance is M2, and the corresponding position in the first column of the fifth group is orange; the cause tracing result of the decision tree model is M1, and the corresponding position in the second column is green; the cause tracing result of the joint model is M1 or M2, and the corresponding position in the third column is sky blue. Comparing the algorithm execution time and cause tracing accuracy data, it can be seen that the test set of the decision tree model has a relatively low accuracy of cause tracing results, while the test set of the joint model has a better effect. Therefore, the abnormal cause tracing model for the production workshop built based on the joint algorithm improves the cause tracing accuracy and also has a certain improvement in execution speed.
[0060] The above is only a preferred implementation of a method for tracing the cause of abnormality in a production workshop based on digital twins disclosed in the present invention, and is not intended to limit the scope of protection of the embodiments of this specification. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the embodiments of this specification shall be included in the scope of protection of the embodiments of this specification.
Claims
1. A method for tracing abnormal causes of production workshops based on digital twins, characterized in that: The steps include: Construct a digital twin production workshop abnormality cause tracing model, the production workshop abnormality cause tracing model includes a physical production workshop, a virtual production workshop, a full-factor twin data platform and a production abnormal event cause tracing service system, wherein the physical production workshop and the virtual production workshop interact, and the full-factor twin data platform is used to drive the virtual model in the virtual production workshop to perform iterative optimization, and the production abnormal event cause tracing service system is used to trace the abnormal cause based on the data-driven production abnormality tracing Bayesian network and the abnormal cause tracing decision tree and display the tracing results; Create a production abnormality cause tracing mechanism: the production abnormal event cause tracing service system calls the full-factor twin data platform to trace the abnormal events in the physical production workshop to obtain the tracing results; the virtual production workshop simulates the data of the physical production workshop to obtain the simulation results, and compares the tracing results with the simulation results. If the comparison results are consistent, an abnormality sorting solution is provided to the physical production workshop based on the tracing results; if the comparison results are inconsistent, the full-factor twin data platform is optimized.
2. According to the method for tracing abnormal causes of production workshops based on digital twins in claim 1, it is characterized in that: The virtual model in the virtual production workshop includes a personnel model, a material model, a method model, an equipment model and an environment model.
3. According to the method for tracing abnormal causes of production workshops based on digital twins in claim 1, it is characterized in that: The full-factor twin data platform includes physical data, virtual data, knowledge base, intelligent algorithm library, production database and model library.
4. According to the method for tracing abnormal causes of production workshops based on digital twins in claim 1, it is characterized in that: The production abnormal event cause tracing service system includes a production abnormality tracing Bayesian network, an abnormality cause tracing decision tree and a visualization tool, wherein the production abnormality tracing Bayesian network is used to reflect the probabilistic relationship between abnormal events and abnormal influencing factors; the abnormality cause tracing decision tree is constructed based on the historical abnormal data statistically collected in the physical production workshop and the constructed Bayesian network, each decision node, state node and result node in the abnormality cause tracing decision tree corresponds to the decision node, state node and benefit node of the Bayesian network respectively, the same event is merged into one node, each path in the abnormality cause tracing decision tree is converted into a cause tracing rule, and an abnormality tracing rule base is obtained, and the abnormality tracing rule base is stored in a knowledge base; the visualization tool is used to display the tracing results.
Citation Information
Cited By
Product efficiency early warning abnormal work order processing method
CN120806581A
Product performance warning abnormal work order handling method
CN120806581B
Production management method and system for water reducing agent cooling equipment
CN121365325A
