Failure risk assessment method, device and equipment for disassembly process of retired electromechanical products

Abstract

The present invention discloses a failure risk assessment method, device and equipment for the disassembly process of retired electromechanical products, the method comprising: starting from the failure level of the components of retired electromechanical products, determining the failure behavior performance of the disassembly process of retired electromechanical products, and by introducing a fuzzy reasoning method, processing multidimensional failure variables to accurately describe the failure state of a single disassembly process of retired electromechanical products; using an improved discrete firefly algorithm to optimize the Bayesian network topology structure used to reflect the causal relationship between the failure disassembly processes of retired electromechanical products; using the expected maximum parameter estimation method to estimate the Bayesian network parameters of each node of the Bayesian network topology structure, so as to estimate the failure condition probability of each disassembly process of retired electromechanical products. The present invention can accurately identify the associated failure risks of potential disassembly processes in the disassembly process of retired electromechanical products.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a failure risk assessment method, device and equipment for a disassembly process of retired electromechanical products. Background Art

[0002] With the rapid development and upgrading of technology, a large number of electromechanical products have been eliminated and retired. At the same time, people's attention to environmental protection has been increasing, and more and more people have realized the impact of discarded electromechanical products on the environment, which has made the recycling of retired electromechanical products a hotly discussed topic. Recycling of retired electromechanical products refers to the process of collecting, disassembling and reusing discarded or scrapped electronic equipment, household appliances and other mechanical equipment. Electronic equipment and mechanical equipment contain many valuable materials and components, such as metals, plastics, rare earths, etc. By recycling and reusing these materials, we can reduce the exploitation of natural resources, reduce the pressure of resource consumption, and achieve sustainable use of resources.

[0003] Disassembly is a core link in recycling, which refers to the process of breaking down a product or subassembly into multiple parts. The disassembly process is a key step in the disassembly process. Through reasonable planning and operation, it ensures the smooth progress, efficient completion and accurate implementation of the disassembly process. It can be seen that the success of the disassembly process directly affects the effect and efficiency of the disassembly. However, retired electromechanical products have undergone long-term use and changes in various environmental conditions, and the quality of their parts is very different. At present, there is no reliable testing equipment and technical methods to accurately identify the quality status of the internal parts of retired electromechanical products in advance, which makes random failure behaviors in the disassembly process of retired electromechanical products occur from time to time. In-depth understanding of the failure behavior mechanism of the disassembly process of retired electromechanical products is of great significance to improving disassembly efficiency, resource recycling and environmental sustainable development. The study of the failure behavior mechanism of the disassembly process faces two challenges: First, the randomness of the failure behavior of a single disassembly process makes it difficult to capture its rules. The disassembly process involves multiple factors, including product structure, working environment, operating technology, etc. The combination and interaction of these factors lead to the diversity and randomness of the failure behavior of the disassembly process; second, there is a certain causal correlation between the failure behaviors of multiple disassembly processes. The failure of the preceding process may cause cascading failures in the subsequent process, which increases the complexity of analyzing and solving the problem. Therefore, in the process of studying the failure behavior mechanism of the disassembly process, it is necessary to accurately characterize and deeply analyze the generation mechanism of random failure of a single disassembly process and the causal relationship mechanism of failure between multiple disassembly processes.

[0004] Therefore, the process failure behavior during the disassembly of retired electromechanical products poses a severe challenge to the overall disassembly efficiency and quality. Considering the ambiguity of disassembly process failure, its state characterization becomes a key problem. At the same time, the causal relationship between disassembly processes has a significant impact on the failure probability of subsequent processes, further increasing the complexity of risk assessment. It can be seen that the existing technology cannot well identify the potential risk of associated failure of disassembly processes during the disassembly of retired electromechanical products. Summary of the invention

[0005] The embodiments of the present invention provide a method, device and equipment for evaluating failure risks in the disassembly process of retired electromechanical products, which can accurately identify the associated failure risks of potential disassembly processes during the disassembly of retired electromechanical products.

[0006] An embodiment of the present invention provides a failure risk assessment method for a disassembly process of retired electromechanical products under double-layer reasoning, comprising:

[0007] Starting from the failure level of retired electromechanical products, the failure behavior of the disassembly process of retired electromechanical products is determined, and by introducing fuzzy reasoning methods, multi-dimensional failure variables are processed to accurately describe the failure state of a single disassembly process of retired electromechanical products.

[0008] The Bayesian network topology structure used to reflect the causal relationship between failure disassembly processes of retired electromechanical products is optimized using the improved discrete firefly algorithm.

[0009] The expected maximum parameter estimation method is used to estimate the Bayesian network parameters of each node of the Bayesian network topology structure to estimate the failure conditional probability of each disassembly process of retired electromechanical products.

[0010] As an improvement of the above scheme, starting from the failure level of components of retired electromechanical products, the failure behavior of the disassembly process of retired electromechanical products is determined, and by introducing fuzzy reasoning methods, multi-dimensional failure variables are processed to accurately describe the failure state of a single disassembly process of retired electromechanical products, including:

[0011] Conduct in-depth analysis of retired electromechanical products to identify multiple dimensional factors that affect the failure of the disassembly process; dimensional factors include: reliability failure, structural failure and process failure;

[0012] Collect historical data of retired electromechanical products during the dismantling process, including service time records of parts, deformation measurements and dismantling operation parameters;

[0013] Based on professional knowledge and practical experience, a fuzzy rule base is constructed; the rule base defines the fuzzy relationship between the failure dimension variables and the failure status of the disassembly process;

[0014] Applying the fuzzy reasoning method, the collected multi-dimensional failure variables are input into the fuzzy reasoning engine. By matching with the fuzzy rule base, the membership function of each dimension is calculated to reflect the fuzziness degree of the parts in different failure dimensions.

[0015] The maximum membership principle is used to defuzzify the membership of multiple dimensions obtained by fuzzy reasoning, so as to clarify the failure status of a single disassembly process.

[0016] As an improvement of the above scheme, the Bayesian network topology structure for reflecting the causal relationship between failure disassembly processes of retired electromechanical products by using the improved discrete firefly algorithm is optimized, including:

[0017] Constructing a disassembly process dataset: Based on the disassembly experience and historical data of retired electromechanical products, a disassembly process dataset is established. This dataset contains the status information of different disassembly processes and the actual causal relationship between them. The dataset records the execution results, failure conditions and impact of each process on subsequent processes during the disassembly process.

[0018] Define individual encoding and decoding rules: In the discrete firefly algorithm, each individual represents a possible Bayesian network topology; individuals are encoded in the form of an adjacency matrix, where the elements in the matrix represent the dependencies between processes;

[0019] Initialize the population: Generate the initial population based on the priority and order of the actual disassembly process using Kahn's topological sorting method; each individual is a directed acyclic graph (DAG) that satisfies the logical relationship of the disassembly process;

[0020] Fitness function design: Design a fitness function to measure the quality of individuals for the Bayesian network topology;

[0021] Improved discrete firefly algorithm optimization: including three main steps: search, update, and evaluation; in the search phase, the firefly swarm randomly explores the solution space to find a better network structure; in the update phase, the firefly position is adjusted according to the fitness value and the attraction and repulsion mechanism of the algorithm; in the evaluation phase, the fitness function is used to evaluate each individual and retain or update a better topological structure;

[0022] Iterative optimization and convergence judgment: Through multiple iterations, the position of the firefly colony is continuously optimized, so as to continuously find a better Bayesian network topology structure; after each iteration, check whether the algorithm has reached the preset stop condition, and end the optimization process after reaching the stop condition; the stop condition includes: the upper limit of the number of iterations or the threshold of fitness improvement;

[0023] Output the optimal structure: When the optimization process ends, the individual with the highest fitness is selected as the optimal solution, thereby obtaining a Bayesian network topology that best reflects the failure causal relationship between the disassembly processes of retired electromechanical products.

[0024] As an improvement of the above scheme, the method of using the expected maximum parameter estimation method to estimate the Bayesian network parameters of each node of the Bayesian network topology structure to estimate the failure condition probability of each disassembly process of retired electromechanical products includes:

[0025] Prepare training data: Based on the historical records of disassembly of retired electromechanical products, collect failure data of relevant disassembly processes;

[0026] Initialize parameter estimation: For each node representing the disassembly process in the Bayesian network, a parameter estimation value is preliminarily set based on the training data, including the initial value of the conditional probability table, which is obtained based on prior knowledge or simple statistical analysis;

[0027] Expected step: Under the current parameter estimation, calculate the probability of each sample being in various states when the state of its parent node is given, that is, calculate the expected count of each node;

[0028] Maximization step: Based on the expected counts calculated in the expectation step, re-estimate the conditional probability distribution of each node so that these estimates maximize the likelihood of the data;

[0029] Iterative optimization: Repeat the expected step and the maximization step until the change in network parameters is less than the preset threshold or reaches the predetermined maximum number of iterations;

[0030] Calculate the conditional probability of failure: Through the above iterative process, the conditional probability table finally obtained reflects the conditional probability of failure of each disassembly process given the state of its parent node; these probabilities reflect the uncertainty of failure of the disassembly process and the mutual influence between processes.

[0031] Another embodiment of the present invention provides a failure risk assessment device for a disassembly process of retired electromechanical products under double-layer reasoning, including:

[0032] The fuzzy reasoning module is used to determine the failure behavior of the disassembly process of retired electromechanical products based on the failure level of the components of retired electromechanical products, and to accurately describe the failure state of a single disassembly process of retired electromechanical products by introducing fuzzy reasoning methods to process multidimensional failure variables;

[0033] An optimization module is used to optimize the Bayesian network topology structure used to reflect the causal relationship between failure disassembly processes of retired electromechanical products using an improved discrete firefly algorithm;

[0034] The estimation module is used to estimate the Bayesian network parameters of each node of the Bayesian network topology structure by using the expected maximum parameter estimation method, so as to estimate the failure condition probability of each disassembly process of retired electromechanical products.

[0035] As an improvement of the above solution, the fuzzy reasoning module is specifically used for:

[0036] Conduct in-depth analysis of retired electromechanical products to identify multiple dimensional factors that affect the failure of the disassembly process; dimensional factors include: reliability failure, structural failure and process failure;

[0037] Collect historical data of retired electromechanical products during the dismantling process, including service time records of parts, deformation measurements and dismantling operation parameters;

[0038] Based on professional knowledge and practical experience, a fuzzy rule base is constructed; the rule base defines the fuzzy relationship between the failure dimension variables and the failure status of the disassembly process;

[0039] Applying the fuzzy reasoning method, the collected multi-dimensional failure variables are input into the fuzzy reasoning engine. By matching with the fuzzy rule base, the membership function of each dimension is calculated to reflect the fuzziness degree of the parts in different failure dimensions.

[0040] The maximum membership principle is used to defuzzify the membership of multiple dimensions obtained by fuzzy reasoning, so as to clarify the failure status of a single disassembly process.

[0041] As an improvement of the above solution, the optimization module is specifically used for:

[0042] Constructing a disassembly process dataset: Based on the disassembly experience and historical data of retired electromechanical products, a disassembly process dataset is established. This dataset contains the status information of different disassembly processes and the actual causal relationship between them. The dataset records the execution results, failure conditions and impact of each process on subsequent processes during the disassembly process.

[0043] Define individual encoding and decoding rules: In the discrete firefly algorithm, each individual represents a possible Bayesian network topology; individuals are encoded in the form of an adjacency matrix, where the elements in the matrix represent the dependencies between processes;

[0044] Initialize the population: Generate the initial population based on the priority and order of the actual disassembly process using Kahn's topological sorting method; each individual is a directed acyclic graph (DAG) that satisfies the logical relationship of the disassembly process;

[0045] Fitness function design: Design a fitness function to measure the quality of individuals for the Bayesian network topology;

[0046] Improved discrete firefly algorithm optimization: including three main steps: search, update, and evaluation; in the search phase, the firefly swarm randomly explores the solution space to find a better network structure; in the update phase, the firefly position is adjusted according to the fitness value and the attraction and repulsion mechanism of the algorithm; in the evaluation phase, the fitness function is used to evaluate each individual and retain or update a better topological structure;

[0047] Iterative optimization and convergence judgment: Through multiple iterations, the position of the firefly colony is continuously optimized, so as to continuously find a better Bayesian network topology structure; after each iteration, check whether the algorithm has reached the preset stop condition, and end the optimization process after reaching the stop condition; the stop condition includes: the upper limit of the number of iterations or the threshold of fitness improvement;

[0048] Output the optimal structure: When the optimization process ends, the individual with the highest fitness is selected as the optimal solution, thereby obtaining a Bayesian network topology that best reflects the failure causal relationship between the disassembly processes of retired electromechanical products.

[0049] As an improvement of the above solution, the estimation module is specifically used for:

[0050] Prepare training data: Based on the historical records of disassembly of retired electromechanical products, collect failure data of relevant disassembly processes;

[0051] Initialize parameter estimation: For each node representing the disassembly process in the Bayesian network, a parameter estimation value is preliminarily set based on the training data, including the initial value of the conditional probability table, which is obtained based on prior knowledge or simple statistical analysis;

[0052] Expected step: Under the current parameter estimation, calculate the probability of each sample being in various states when the state of its parent node is given, that is, calculate the expected count of each node;

[0053] Maximization step: Based on the expected counts calculated in the expectation step, re-estimate the conditional probability distribution of each node so that these estimates maximize the likelihood of the data;

[0054] Iterative optimization: Repeat the expected step and the maximization step until the change in network parameters is less than the preset threshold or reaches the predetermined maximum number of iterations;

[0055] Calculate the conditional probability of failure: Through the above iterative process, the conditional probability table finally obtained reflects the conditional probability of failure of each disassembly process given the state of its parent node; these probabilities reflect the uncertainty of failure of the disassembly process and the mutual influence between processes.

[0056] Another embodiment of the present invention provides a failure risk assessment device for the disassembly process of retired electromechanical products under two-layer reasoning, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, the failure risk assessment method for the disassembly process of retired electromechanical products under two-layer reasoning described in the above-mentioned embodiment of the invention is implemented.

[0057] Another embodiment of the present invention provides a storage medium, wherein the computer-readable storage medium includes a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the failure risk assessment method for the disassembly process of retired electromechanical products under the two-layer reasoning described in the above-mentioned embodiment of the invention.

[0058] Compared with the prior art, the embodiments of the present invention have the following beneficial effects:

[0059] The embodiment of the present invention creatively integrates fuzzy reasoning and Bayesian network technology to form a two-layer reasoning framework for the randomness of process failure and the complexity of causal association faced in the disassembly process of retired electromechanical products. First, starting from the specific manifestations of component failure, fuzzy reasoning is used to deal with the multi-dimensional uncertainty in the disassembly process, and the failure state of a single process is accurately characterized. Subsequently, the Bayesian network structure is optimized by the optimized discrete firefly algorithm to better map the failure causal chain between disassembly processes. Finally, the expected maximum algorithm is used to estimate the network parameters, so as to obtain the failure condition probability of each disassembly process, and realize the refined assessment of the disassembly risk. Therefore, the embodiment of the present invention effectively solves the mechanism analysis problem of the failure behavior of the disassembly process of retired electromechanical products through a two-layer reasoning strategy, improves the accuracy and practicality of risk assessment, and provides strong technical support for the green recycling of retired products. From the above analysis, it can be seen that the embodiment of the present invention significantly enhances the ability to identify and evaluate the process failure associated risks in the disassembly process of retired electromechanical products, thereby ultimately improving the efficiency and quality of disassembly operations. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 It is a flow chart of a failure risk assessment method for a disassembly process of retired electromechanical products under double-layer reasoning provided by an embodiment of the present invention;

[0061] Figure 2 It is a process diagram of a failure risk assessment method for a disassembly process of retired electromechanical products under double-layer reasoning provided by an embodiment of the present invention;

[0062] Figure 3 is a schematic diagram of an example of a disassembly priority diagram provided by an embodiment of the present invention;

[0063] Figure 4is a schematic diagram of a reliability failure triangular membership function provided by an embodiment of the present invention;

[0064] Figure 5 is a schematic diagram of an encoding method provided by an embodiment of the present invention;

[0065] Figure 6 is a schematic diagram of iteratively generating an initial population provided by an embodiment of the present invention;

[0066] Figure 7 is a schematic diagram of individual location update provided by an embodiment of the present invention;

[0067] Figure 8 This is a schematic diagram of an illegal network topology structure provided by an embodiment of the present invention;

[0068] Fig. 9 It is a structural schematic diagram of a failure risk assessment device for a disassembly process of retired electromechanical products under double-layer reasoning provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0069] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0070] See also Figure 1 , is a flow chart of a failure risk assessment method for a disassembly process of retired electromechanical products under double-layer reasoning provided by an embodiment of the present invention. The failure risk assessment method for a disassembly process of retired electromechanical products under double-layer reasoning includes steps S10 to S12:

[0071] S10, starting from the failure level of components of retired electromechanical products, determine the failure behavior of the disassembly process of retired electromechanical products, and introduce fuzzy reasoning methods to process multidimensional failure variables and accurately describe the failure state of a single disassembly process of retired electromechanical products;

[0072] S11, using the improved discrete firefly algorithm to optimize the Bayesian network topology structure used to reflect the causal relationship between the failure disassembly processes of retired electromechanical products;

[0073] S12, using the expected maximum parameter estimation method to estimate the Bayesian network parameters of each node of the Bayesian network topology structure, so as to estimate the failure conditional probability of each disassembly process of retired electromechanical products.

[0074] like Figure 2As shown, the embodiment of the present invention creatively integrates fuzzy reasoning and Bayesian network technology to form a two-layer reasoning framework for the randomness of process failure and the complexity of causal association faced in the disassembly process of retired electromechanical products. First, starting from the specific manifestations of component failure, fuzzy reasoning is used to deal with the multi-dimensional uncertainty in the disassembly process, and the failure state of a single process is accurately characterized. Subsequently, the Bayesian network structure is optimized by the optimized discrete firefly algorithm to better map the failure causal chain between disassembly processes. Finally, the expected maximum algorithm is used to estimate the network parameters, so as to obtain the failure condition probability of each disassembly process, and realize the refined assessment of the disassembly risk. Therefore, the embodiment of the present invention effectively solves the mechanism analysis problem of the failure behavior of the disassembly process of retired electromechanical products through a two-layer reasoning strategy, improves the accuracy and practicality of risk assessment, and provides strong technical support for the green recycling of retired products. From the above analysis, it can be seen that the embodiment of the present invention greatly enhances the ability to identify and evaluate the process failure association risk in the disassembly process of retired electromechanical products, thereby ultimately improving the efficiency and quality of the disassembly operation.

[0075] As an improvement of the above scheme, starting from the failure level of components of retired electromechanical products, the failure behavior of the disassembly process of retired electromechanical products is determined, and by introducing fuzzy reasoning methods, multi-dimensional failure variables are processed to accurately describe the failure state of a single disassembly process of retired electromechanical products, including:

[0076] Conduct in-depth analysis of retired electromechanical products to identify multiple dimensional factors that affect the failure of the disassembly process; dimensional factors include: reliability failure, structural failure and process failure;

[0077] Collect historical data of retired electromechanical products during the dismantling process, including service time records of parts, deformation measurements and dismantling operation parameters;

[0078] Based on professional knowledge and practical experience, a fuzzy rule base is constructed; the rule base defines the fuzzy relationship between the failure dimension variables and the failure status of the disassembly process;

[0079] Applying the fuzzy reasoning method, the collected multi-dimensional failure variables are input into the fuzzy reasoning engine. By matching with the fuzzy rule base, the membership function of each dimension is calculated to reflect the fuzziness degree of the parts in different failure dimensions.

[0080] The maximum membership principle is used to defuzzify the membership of multiple dimensions obtained by fuzzy reasoning, so as to clarify the failure status of a single disassembly process.

[0081] In this embodiment, specifically:

[0082] Disassembly of retired electromechanical products is an operation characterized by complexity and order, which includes multiple disassembly processes. There are interdependencies and connections between these processes. Disassembly priority diagrams have been widely used to show the relative importance of priorities between disassembly processes. Figure 3 As shown in the disassembly priority diagram, the arrow indicates the direction of disassembly. If the previous disassembly process fails, some parts or components may not be properly disassembled or damaged, which will directly affect the subsequent disassembly process. It can be seen that the priority constraint between disassembly processes is an important factor leading to the causal relationship between disassembly process failure behaviors.

[0083] The disassembly priority graph cannot provide more detailed information for the failure behavior of the disassembly process. In order to more accurately characterize the failure state of the disassembly process and further reveal the intrinsic causal relationship between the previous and subsequent disassembly processes, fuzzy reasoning and Bayesian networks provide a deeper reasoning method that can better handle uncertainty and probability calculation, thereby more comprehensively understanding the failure behavior of the disassembly process. Each disassembly process of retired electromechanical products can be regarded as a node in a Bayesian network. The directed edges between nodes represent the dependency relationship between the disassembly processes, from the parent process to the child process. By calculating the conditional probability table of each node, a Bayesian network model that characterizes the causal relationship between the disassembly processes can be constructed. For a single disassembly process, it is first necessary to describe its failure performance and use fuzzy reasoning to describe the failure state of the disassembly process. Secondly, given that there are many possible combinations of connection relationships between nodes, the process of identifying the optimal Bayesian network topology involves the exploration and evaluation of the potential solution space, so it is necessary to use a combinatorial optimization algorithm to determine the optimal network topology. Once the network topology is determined, the intrinsic causal relationship between the failure behaviors of the disassembly process can be accurately characterized. Then, by learning the parameters of historical data, we can obtain the probability table of random failure conditions for each disassembly process. Combining forward reasoning and backward reasoning, the Bayesian network can make effective inferences and predictions under incomplete or uncertain information conditions, evaluate the risk and reliability of each disassembly process, and provide guidance and decision-making basis for disassembly operations.

[0084] In the fuzzy reasoning part, this embodiment uses fuzzy logic methods with the help of expert experience and knowledge to comprehensively process multi-dimensional failure information, so as to accurately determine the failure status of parts. In the Bayesian network reasoning part, it is proposed to improve the discrete firefly algorithm to determine the optimal network topology through iterative optimization to represent the failure causal relationship between disassembly processes, and use the expectation maximization algorithm to calculate the failure probability of each node under a given parent node state based on historical data, that is, the conditional probability of random failure, to reflect the correlation and uncertainty of failures in different disassembly processes. This model can describe and analyze the failure behaviors that may occur in the disassembly process, thereby providing a new way to improve the efficiency and reliability of the disassembly process.

[0085] Each dimensional variable is first fuzzified, and a fuzzy set is defined for each variable, such as "low", "medium", and "high" to represent the level of reliability failure, "weak", "medium", and "strong" to represent the degree of structural failure, and "small", "medium", and "large" to represent the impact of process failure. A membership function is defined for each fuzzy set to describe its degree of uncertainty. The triangular membership function is simple and easy to explain. This embodiment uses a triangular membership function to convert the values ​​of each dimensional variable into corresponding membership values. If the disassembly process, that is, the value of the component in the reliability failure dimension is Figure 4 .

[0086] Based on expert knowledge and experience, a set of fuzzy rules is established to establish a connection between the failure dimension variables and the fuzzy set of the failure state of the parts. Using the established fuzzy rules, fuzzy reasoning is performed, and the fuzzy value of the failure state of the disassembly process is obtained by calculating the matching degree between the fuzzy set of the input variables and the rules. Then, the maximum membership rule is used to defuzzify and obtain the final output representing the failure state of the parts.

[0087] From the above analysis, it can be seen that this embodiment deeply analyzes the disassembly characteristics of retired electromechanical products, identifies the key failure dimensions that affect disassembly - reliability, structure and processability, and collects relevant historical data; builds a fuzzy rule base based on professional experience and actual data to associate failure dimensions with the status of the disassembly process; uses fuzzy reasoning to convert multi-dimensional data into fuzzy sets, and calculates the membership by matching with the rule base to reflect the degree of failure of each dimension; finally, the maximum membership principle is used to determine the specific failure state, and an accurate description of the failure behavior of the disassembly process is achieved. Therefore, this embodiment effectively improves the accuracy of failure state identification in the disassembly process of retired electromechanical products through fuzzy reasoning and multi-dimensional data analysis, and provides a solid foundation for subsequent risk assessment and optimization of the disassembly process.

[0088] As an improvement of the above scheme, the Bayesian network topology structure for reflecting the causal relationship between failure disassembly processes of retired electromechanical products by using the improved discrete firefly algorithm is optimized, including:

[0089] Constructing a disassembly process dataset: Based on the disassembly experience and historical data of retired electromechanical products, a disassembly process dataset is established. This dataset contains the status information of different disassembly processes and the actual causal relationship between them. The dataset records the execution results, failure conditions and impact of each process on subsequent processes during the disassembly process.

[0090] Define individual encoding and decoding rules: In the discrete firefly algorithm, each individual represents a possible Bayesian network topology; individuals are encoded in the form of an adjacency matrix, where the elements in the matrix represent the dependencies between processes;

[0091] Initialize the population: Generate the initial population based on the priority and order of the actual disassembly process using Kahn's topological sorting method; each individual is a directed acyclic graph (DAG) that satisfies the logical relationship of the disassembly process;

[0092] Fitness function design: Design a fitness function to measure the quality of individuals for the Bayesian network topology;

[0093] Improved discrete firefly algorithm optimization: including three main steps: search, update, and evaluation; in the search phase, the firefly swarm randomly explores the solution space to find a better network structure; in the update phase, the firefly position is adjusted according to the fitness value and the attraction and repulsion mechanism of the algorithm; in the evaluation phase, the fitness function is used to evaluate each individual and retain or update a better topological structure;

[0094] Iterative optimization and convergence judgment: Through multiple iterations, the position of the firefly colony is continuously optimized, so as to continuously find a better Bayesian network topology structure; after each iteration, check whether the algorithm has reached the preset stop condition, and end the optimization process after reaching the stop condition; the stop condition includes: the upper limit of the number of iterations or the threshold of fitness improvement;

[0095] Output the optimal structure: When the optimization process ends, the individual with the highest fitness is selected as the optimal solution, thereby obtaining a Bayesian network topology that best reflects the failure causal relationship between the disassembly processes of retired electromechanical products.

[0096] In this embodiment, specifically:

[0097] The arrangement and connection of nodes in the Bayesian network topology structure not only reflects the sequence of disassembly processes of retired electromechanical products, but also reveals the propagation path and mutual influence relationship between failure behaviors in the disassembly processes. This embodiment improves the discrete firefly algorithm to efficiently learn the network topology structure, and selects the Bayesian information criterion scoring function f BIC (G, D) quantitatively evaluates the matching degree between the network topology and the disassembled sample data set, and its expression is as follows:

[0098]

[0099] Among them, n represents the number of processes, q i Representative process X i The parent process set pa(x i ) of the node states, r i Representative process Xi The number of node states, m ijk It is the disassembly sample data set D that meets the disassembly process X i The number of samples whose node state is k and whose parent process node state is j, m is the total number of disassembled samples, the first term represents the log-likelihood of the disassembled sample data and the BN topological structure, and the second term is used to measure the parameter complexity and the storage space of the disassembled sample data. BIC The larger the (G,D) value is, the more accurately the current BN topology can describe the failure causal relationship between the disassembly processes of retired electromechanical products.

[0100] BN node definition:

[0101] Each disassembly process of retired electromechanical products is abstracted as a node in the Bayesian network, where the node variable is defined as the failure state of the corresponding disassembly process. In view of the diversity of causes of failure behavior in the disassembly process, a single state variable is difficult to fully describe its complex characteristics. From the perspective of ensuring the smooth implementation of the disassembly operation, this embodiment divides the relevant influencing factors into three dimensions, each of which contains two failure factors, namely ① reliability failure, the relevant failure factors are service time and degradation degree; ② structural failure, the relevant failure factors are deformation and crack expansion rate; ③ process failure, the relevant failure factors are disassembly speed and disassembly intensity, the value range of the failure dimension variable is defined as [0,1], and the failure factor variables are all normalized.

[0102] At the reliability failure level, the two key factors of service time (t) and degradation degree (d) are comprehensively considered through nonlinear combination, and the formula based on exponential function RF(t,d)=1-exp(-td 2 ), describing the basic trend of reliability failure. At the structural failure level, considering the interaction between deformation and crack growth rate, the formula is defined To measure the structural failure of components. At the process failure level, establish the formula It reflects the comprehensive impact of disassembly speed and disassembly intensity on process failure. Fuzzy logic methods are further used to comprehensively process and analyze the information of these three dimensions to provide a more accurate and comprehensive disassembly process failure state analysis framework. In the past, when dealing with the failure of the disassembly process, researchers generally roughly divided it into two states: completely normal and completely failed. However, in the actual disassembly project of retired electromechanical products, there are still fuzzy intermediate states from normal operation to complete failure. Therefore, this embodiment sets the failure state to three value conditions, including normal, semi-failure and complete failure, represented by the letters N, P and F respectively.

[0103] Table 1 Variables and their descriptions

[0104]

[0105] Under the condition that the Bayesian network topology structure for the causal relationship between disassembly processes is known, the random failure conditional probability of each disassembly process is evaluated with the help of statistical information in the disassembly sample data set. ijk Denoted as the node state parameter of the Bayesian network, representing the disassembly process X i In its parent process set pa(x i ) is in node state j, disassembly process X i The probability that the node state is k (i.e., the disassembly process X i The conditional probability of random failure), θ ijk The expression is as follows:

[0106]

[0107] Each disassembly process node corresponds to a conditional probability table (CPT), which is the conditional probability set {θ ijk}.

[0108] In the Bayesian network reasoning part, an improved discrete firefly algorithm is proposed to determine the optimal network topology through iterative optimization to represent the failure causal relationship between disassembly processes. The expectation maximization algorithm is used to calculate the failure probability of each node under a given parent node state based on historical data, that is, the conditional probability of random failure, to reflect the correlation and uncertainty of failures in different disassembly processes. This model can describe and analyze the failure behaviors that may occur in the disassembly process, thus providing a new way to improve the efficiency and reliability of the disassembly process.

[0109] The process of improving the discrete firefly algorithm to solve the topological structure is as follows:

[0110] 1. Individual code:

[0111] Individual coding plays an important role in optimization problems. In the problem of solving the optimal network structure, individual coding is a way to describe a possible network structure. The adjacency matrix is ​​a commonly used method to represent the node connection relationship in the network structure. For a network with N nodes, its adjacency matrix is

[0112]

[0113] In the adjacency matrix, element a ij Indicates whether there is a directed edge from node i to node j. If there is a directed edge, then a ij is 1; if there is no directed edge, then a ijis 0. Figure 5 As shown, in order to more conveniently process, compare, and optimize the network structures of different individuals in the algorithm and improve the computational efficiency, the adjacency matrix is ​​expanded row by row to obtain the vector X = [a 11 , …a 1n , …, a n1 , …, a nn ], and use this vector to represent an individual.

[0114] 2. Initial population generation:

[0115] The population initialization is the process of randomly generating multiple directed acyclic graphs, using the concept of topological sorting proposed by Kahn

[23] can help achieve this goal. According to the properties of topological sorting, there is at least one topological sorting in a directed acyclic graph. Therefore, when a topologically sorted node list is given, a directed acyclic graph can be constructed based on this list, and all directed edges must comply with the rules of topological sorting, that is, from the nodes in front to the nodes in the back, ensuring that the generated directed graph does not have loops.

[0116] The disassembly and recycling of retired electromechanical products is often carried out according to a given disassembly sequence. The disassembly sequence reflects the disassembly priority relationship between parts to a certain extent. This order is closely related to the failure behavior of the process. Therefore, when initializing the population, the disassembly sequence executed in the actual disassembly task is used as the topological sequence list, such as Figure 6 As shown. Based on this list, multiple initial network structures are randomly generated so that the algorithm can adapt to the characteristics of the actual disassembly task as quickly as possible and obtain a Bayesian network topology that is highly matched with the disassembly sample data set as quickly as possible. The specific steps for generating the initial population are as follows:

[0117] (1) Use the disassembly sequence as the node list for topological sorting.

[0118] (2) Create an empty population according to the preset population size and number of processes.

[0119] (3) Create an empty adjacency matrix based on the number of nodes.

[0120] (4) For each node in the topological sort, traverse the topological sort list backwards starting from the position of the current node.

[0121] (5) Randomly select a node after the current node and set the value of the corresponding element in the adjacency matrix to 1. (6) Expand the adjacency matrix row by row, and the resulting vector is represented as an individual.

[0122] (7) Repeat steps (4), (5), and (6) until the population size is reached.

[0123] 3. Aggregation strategy:

[0124] In the process of iterative optimization of the population, the intensity of each firefly's light emission is determined by the value of the objective function. Fireflies with strong light emission will attract fireflies with weak light emission to move towards them. The mutual attraction β is related to the distance between the two. Individuals with closer distance have greater attraction. Since the network structure represented by the individual is represented in the form of a vector, the solution space is discrete, so the Manhattan distance is used to measure the distance between individuals.

[0125]

[0126] Then the calculation formula of attraction is as follows

[0127]

[0128] In the formula, β0 is the attraction between two fireflies when the distance is 0, that is, the initial value of the attraction. In order to simulate the attenuation of firefly light, a light absorption coefficient γ is introduced. This factor determines that the light intensity gradually weakens with the increase of distance, making individuals farther away less attracted.

[0129] When Firefly X i The objective function value is lower than that of Firefly X j When the random vector R is set to (r1, r2, ..., r n2 ), each element in the vector is a random number between 0 and 1. Firefly X i X i The position update formula is defined as follows

[0130]

[0131] The calculation process of I is:

[0132]

[0133] Vector I combines the position information of two individuals and is the sum of X i and X j A new binary vector is obtained by performing bitwise XOR operation on the elements at the corresponding positions in . When the value of the element in the vector is 1, it indicates that at the same position k, individual X i and individual X j The connection of the edges is different; when the value is 0, it means that at the same position, individual X i and individual X j There are edges or there are no edges. From the aggregation strategy, we can know that individuals have a certain ability to move autonomously. Only when the value of the element in the random vector R is greater than the attraction, the individual X iOnly then will the XOR operation be performed with vector I at the corresponding position to update the position, such as Figure 7 shown.

[0134] 4. Mutation Mechanism

[0135] In order to avoid the algorithm falling into the local optimal solution, the updated individual X i When performing mutation operations, edge addition, edge deletion, and edge reversal operations can be performed to introduce randomness and diversity, thereby preventing the algorithm from falling into a local optimal solution. Specifically, when performing an edge addition operation, two nodes can be randomly selected and an edge connecting them can be added. When performing an edge deletion operation, an existing edge can be randomly selected and removed to simplify the network structure. Finally, when performing an edge reversal operation, a directed edge can be randomly selected and its direction reversed. These mutation operations can introduce new relationships and dependencies, thereby affecting the conditional probability distribution and causal relationships.

[0136] During the iteration of the algorithm, due to the interaction of factors such as the vast search space and individual update and mutation strategies, illegal network topologies may appear, mainly including ring structures, bidirectional edges, and isolated nodes. The following is an illustration of a simple network structure that contains 4 nodes and 3 edges.

[0137] A Bayesian network is a directed acyclic graph, so any structure that forms a cycle is illegal. Figure 8 (b) shows a directed cycle consisting of three nodes, which can be broken using the depth-first search algorithm (DFS); there may be multiple directed edges between the same pair of nodes, which will lead to repeated dependencies, such as Figure 8 As shown in (c), there is a bidirectional edge between node 1 and node 2. At this time, the values ​​of the elements at the diagonal symmetric positions in the adjacency matrix are equal. Therefore, to detect the existence of a bidirectional edge, we can compare a ij X ij and a ji (i≠j)X ji (i≠j) Whether the element value is all 1. When there are bidirectional edges in the individual network structure, randomly delete one of the two directed edges; Figure 8 There is an isolated node in (d). This node has no parent node or child node. The values ​​of the row and column elements of the node in the adjacency matrix are all zero. Since isolated nodes are not connected to other nodes, they are meaningless in the Bayesian network structure. Consider adding necessary edges to maintain the connectivity and integrity of the network by reconnecting isolated nodes with other related nodes.

[0138] 5. Algorithm description:

[0139] The steps of the optimal Bayesian network structure solution algorithm to characterize the causal relationship between failures in the disassembly process are as follows:

[0140] (1) Initialization parameters: initial attraction β0, light absorption coefficient γ, maximum number of iterations MaxIt, population size NP.

[0141] (2) Based on the population size and encoding scheme, use topological sorting to initialize the population Pop and calculate the objective function value f(X i ).

[0142] (3) Set the initial value of the iteration counter t to 1.

[0143] (4) Create a temporary population Pop temp = Pop, and set the target degree values ​​of all its individuals to negative infinity.

[0144] (5) Compare the individuals in the population pairwise. If there is f(X i )<f(X j ), update individual X i The position is X itemp , the objective function value is recorded as f(X itemp ).

[0145] (6) If The objective function value is less than X itemp , update the temporary population For X itemp , and update its objective function value at the same time.

[0146] (7) Merge Pop and Pop temp , and select the individuals with the top NP fitness values ​​as the population for the next iteration.

[0147] (8) Update the value of iteration counter t.

[0148] (9) Determine whether the maximum number of iterations has been reached. If not, return to step (4); otherwise, proceed to the next step.

[0149] Sort the individuals in the population Pop from high to low according to the objective function value, and find the current global optimal solution X best .

[0150] From the above analysis, it can be seen that this embodiment innovatively uses the improved discrete firefly algorithm to optimize the topological structure of the Bayesian network to accurately map the failure causal chain between the disassembly processes of retired electromechanical products. By constructing a detailed disassembly process data set, defining detailed individual coding rules, and initializing the population based on Kahn topological sorting, a rigorous fitness function is designed to guide the algorithm's search and optimization process. The algorithm continuously adjusts the position of fireflies (representing different network structures) through iterative exploration until it finds the Bayesian network structure that best describes the causal relationship of process failure. During this period, the algorithm integrates efficient search strategies, dynamic individual position update rules, and intelligent mutation mechanisms to ensure that the algorithm can effectively navigate in complex solution spaces, avoid local optimality, and promote the discovery of global optimal solutions. Finally, the optimal network structure is output according to the preset stop conditions, achieving high-precision modeling of the failure mode of the disassembly process. In summary, this embodiment significantly enhances the accuracy and efficiency of the Bayesian network in parsing the failure causal relationship of the disassembly process of retired electromechanical products through intelligent optimization strategies.

[0151] As an improvement of the above scheme, the method of using the expected maximum parameter estimation method to estimate the Bayesian network parameters of each node of the Bayesian network topology structure to estimate the failure condition probability of each disassembly process of retired electromechanical products includes:

[0152] Prepare training data: Based on the historical records of disassembly of retired electromechanical products, collect failure data of relevant disassembly processes;

[0153] Initialize parameter estimation: For each node representing the disassembly process in the Bayesian network, a parameter estimation value is preliminarily set based on the training data, including the initial value of the conditional probability table, which is obtained based on prior knowledge or simple statistical analysis;

[0154] Expected step: Under the current parameter estimation, calculate the probability of each sample being in various states when the state of its parent node is given, that is, calculate the expected count of each node;

[0155] Maximization step: Based on the expected counts calculated in the expectation step, re-estimate the conditional probability distribution of each node so that these estimates maximize the likelihood of the data;

[0156] Iterative optimization: Repeat the expected step and the maximization step until the change in network parameters is less than the preset threshold or reaches the predetermined maximum number of iterations;

[0157] Calculate the conditional probability of failure: Through the above iterative process, the conditional probability table finally obtained reflects the conditional probability of failure of each disassembly process given the state of its parent node; these probabilities reflect the uncertainty of failure of the disassembly process and the mutual influence between processes.

[0158] In this embodiment, specifically:

[0159] After determining the structure of the Bayesian network, its parameters need to be determined. Parameter estimation is to estimate the conditional probability table of each node in the network based on known data samples. Since failure data is often a small sample, when data is insufficient, introducing expert knowledge can improve the accuracy of Bayesian network parameter learning. Assuming the sample set size is M, for the root node X i In a disassembly example, the corresponding fuzzy set obtained by fuzzy reasoning is P i k =[a1,a2,a3](1≤k≤M,0≤a≤1), average the fuzzy set results, and get the root node fuzzy prior probability as

[0160]

[0161] EM algorithm (Expectation-Maximization) is a commonly used parameter estimation method

[24] , which can use prior knowledge to improve the generalization ability of the model, and has the characteristics of robustness and high efficiency. The process of using the EM algorithm to complete the Bayesian network parameter estimation is as follows:

[0162] E-step (Expectation): In the E-step, the posterior probability corresponding to each sample is calculated based on the current parameter estimates and observed data. Specifically, based on the topological structure of the Bayesian network and the current parameter estimates, the posterior probability of each node in each sample is calculated, expressed as P(X i |X {-i} ), where X i represents the value of node i, X {-i} Indicates the values ​​of nodes other than node i.

[0163] M-step (Maximization): In the M-step, the conditional probability table of each node is updated according to the posterior probability calculated in the E-step. Specifically, the conditional probability table of each node is updated using the maximum a posteriori estimation method based on the posterior probability and the observed data.

[0164] Repeat steps E and M: Repeat the above two steps until the parameter estimation converges or the preset number of iterations is reached. In each iteration, the accuracy and stability of the parameter estimation are gradually improved by calculating the posterior probability and updating the conditional probability table.

[0165] From the above analysis, it can be seen that this embodiment integrates historical failure data with the expectation maximization (EM) algorithm. First, training data is prepared based on the disassembly records of retired electromechanical products, and the conditional probability of each disassembly process node in the Bayesian network is preliminarily estimated based on this. Subsequently, by alternating the expectation (E) step and the maximization (M) step, the parameter estimation is continuously refined until convergence, so as to accurately quantify the conditional probability of failure of each disassembly process in a specific context, and fully reveal the uncertainty of the failure of the disassembly process and the mechanism of their mutual influence. Therefore, this embodiment significantly improves the accuracy and practicality of the Bayesian network model in the risk assessment of disassembly failure of retired electromechanical products, and achieves accurate estimation of the failure probability of the disassembly process through iterative optimization.

[0166] In summary, the embodiment of the present invention innovatively combines fuzzy reasoning methods with Bayesian network theory to address the random failure behaviors that occur during the disassembly and recycling process of retired electromechanical products, and constructs a risk assessment model for random failure behaviors of disassembly processes under double-layer reasoning. This model not only conducts an in-depth analysis of the causal relationship between failures of multiple disassembly processes, but also accurately estimates the conditional probability of random failure of a single disassembly process. This paper takes the failure behavior analysis of a single disassembly process as the starting point, and uses fuzzy reasoning methods to identify the failure state of nodes in the Bayesian network by comprehensively considering multiple failure dimension information. In order to reflect the complex causal relationship between failed disassembly processes, an improved discrete firefly algorithm is proposed to solve the topological structure. While retaining the optimization performance of the firefly algorithm, the algorithm is adaptively improved for the characteristics of the failure problem of the disassembly process, so that the complex relationship between the disassembly processes can be more effectively handled. After determining the topological structure of the Bayesian network, combined with the EM parameter estimation method, the conditional probability of random failure of each disassembly process is accurately estimated. The risk assessment model proposed in the embodiment of the present invention helps to better identify potential disassembly process-related failure risks, which is of great significance for effectively improving the disassembly efficiency and safety of retired electromechanical products, and provides a new and reliable failure analysis method for the disassembly industry of retired electromechanical products. Further research can improve the quality and reliability of failure information through more data collection and analysis, improve the characterization of disassembly process failure behavior, improve the accuracy and credibility of the model, and apply it to a wider range of fields and practical problems.

[0167] See also Fig. 9 , is a schematic diagram of a failure risk assessment device for a disassembly process of retired electromechanical products under double-layer reasoning provided by an embodiment of the present invention. The failure risk assessment device for a disassembly process of retired electromechanical products under double-layer reasoning comprises:

[0168] The fuzzy reasoning module 10 is used to determine the failure behavior of the disassembly process of the retired electromechanical products from the failure level of the components of the retired electromechanical products, and to accurately describe the failure state of a single disassembly process of the retired electromechanical products by introducing a fuzzy reasoning method to process multidimensional failure variables;

[0169] An optimization module 11 is used to optimize the Bayesian network topology structure used to reflect the causal relationship between failure disassembly processes of retired electromechanical products by using an improved discrete firefly algorithm;

[0170] The estimation module 12 is used to estimate the Bayesian network parameters of each node of the Bayesian network topology structure by using the expected maximum parameter estimation method, so as to estimate the failure condition probability of each disassembly process of the retired electromechanical product.

[0171] The embodiment of the present invention creatively integrates fuzzy reasoning and Bayesian network technology to form a two-layer reasoning framework for the randomness of process failure and the complexity of causal association faced in the disassembly process of retired electromechanical products. First, starting from the specific manifestations of component failure, fuzzy reasoning is used to deal with the multi-dimensional uncertainty in the disassembly process, and the failure state of a single process is accurately characterized. Subsequently, the Bayesian network structure is optimized by the optimized discrete firefly algorithm to better map the failure causal chain between disassembly processes. Finally, the expected maximum algorithm is used to estimate the network parameters, so as to obtain the failure condition probability of each disassembly process, and realize the refined assessment of the disassembly risk. Therefore, the embodiment of the present invention effectively solves the mechanism analysis problem of the failure behavior of the disassembly process of retired electromechanical products through a two-layer reasoning strategy, improves the accuracy and practicality of risk assessment, and provides strong technical support for the green recycling of retired products. From the above analysis, it can be seen that the embodiment of the present invention significantly enhances the ability to identify and evaluate the process failure associated risks in the disassembly process of retired electromechanical products, thereby ultimately improving the efficiency and quality of disassembly operations.

[0172] As an improvement of the above solution, the fuzzy reasoning module is specifically used for:

[0173] Conduct in-depth analysis of retired electromechanical products to identify multiple dimensional factors that affect the failure of the disassembly process; dimensional factors include: reliability failure, structural failure and process failure;

[0174] Collect historical data of retired electromechanical products during the dismantling process, including service time records of parts, deformation measurements and dismantling operation parameters;

[0175] Based on professional knowledge and practical experience, a fuzzy rule base is constructed; the rule base defines the fuzzy relationship between the failure dimension variables and the failure status of the disassembly process;

[0176] Applying the fuzzy reasoning method, the collected multi-dimensional failure variables are input into the fuzzy reasoning engine. By matching with the fuzzy rule base, the membership function of each dimension is calculated to reflect the fuzziness degree of the parts in different failure dimensions.

[0177] The maximum membership principle is used to defuzzify the membership of multiple dimensions obtained by fuzzy reasoning, so as to clarify the failure status of a single disassembly process.

[0178] As an improvement of the above solution, the optimization module is specifically used for:

[0179] Constructing a disassembly process dataset: Based on the disassembly experience and historical data of retired electromechanical products, a disassembly process dataset is established. This dataset contains the status information of different disassembly processes and the actual causal relationship between them. The dataset records the execution results, failure conditions and impact of each process on subsequent processes during the disassembly process.

[0180] Define individual encoding and decoding rules: In the discrete firefly algorithm, each individual represents a possible Bayesian network topology; individuals are encoded in the form of an adjacency matrix, where the elements in the matrix represent the dependencies between processes;

[0181] Initialize the population: Generate the initial population based on the priority and order of the actual disassembly process using Kahn's topological sorting method; each individual is a directed acyclic graph (DAG) that satisfies the logical relationship of the disassembly process;

[0182] Fitness function design: Design a fitness function to measure the quality of individuals for the Bayesian network topology;

[0183] Improved discrete firefly algorithm optimization: including three main steps: search, update, and evaluation; in the search phase, the firefly swarm randomly explores the solution space to find a better network structure; in the update phase, the firefly position is adjusted according to the fitness value and the attraction and repulsion mechanism of the algorithm; in the evaluation phase, the fitness function is used to evaluate each individual and retain or update a better topological structure;

[0184] Iterative optimization and convergence judgment: Through multiple iterations, the position of the firefly colony is continuously optimized, so as to continuously find a better Bayesian network topology structure; after each iteration, check whether the algorithm has reached the preset stop condition, and end the optimization process after reaching the stop condition; the stop condition includes: the upper limit of the number of iterations or the threshold of fitness improvement;

[0185] Output the optimal structure: When the optimization process ends, the individual with the highest fitness is selected as the optimal solution, thereby obtaining a Bayesian network topology that best reflects the failure causal relationship between the disassembly processes of retired electromechanical products.

[0186] As an improvement of the above solution, the estimation module is specifically used for:

[0187] Prepare training data: Based on the historical records of disassembly of retired electromechanical products, collect failure data of relevant disassembly processes;

[0188] Initialize parameter estimation: For each node representing the disassembly process in the Bayesian network, a parameter estimation value is preliminarily set based on the training data, including the initial value of the conditional probability table, which is obtained based on prior knowledge or simple statistical analysis;

[0189] Expected step: Under the current parameter estimation, calculate the probability of each sample being in various states when the state of its parent node is given, that is, calculate the expected count of each node;

[0190] Maximization step: Based on the expected counts calculated in the expectation step, re-estimate the conditional probability distribution of each node so that these estimates maximize the likelihood of the data;

[0191] Iterative optimization: Repeat the expected step and the maximization step until the change in network parameters is less than the preset threshold or reaches the predetermined maximum number of iterations;

[0192] Calculate the conditional probability of failure: Through the above iterative process, the conditional probability table finally obtained reflects the conditional probability of failure of each disassembly process given the state of its parent node; these probabilities reflect the uncertainty of failure of the disassembly process and the mutual influence between processes.

[0193] An embodiment of the present invention provides a failure risk assessment device for the disassembly process of retired electromechanical products under two-layer reasoning. The failure risk assessment device for the disassembly process of retired electromechanical products under two-layer reasoning of this embodiment includes: a processor, a memory, and a computer program stored in the memory and executable on the processor, such as a failure risk assessment program for the disassembly process of retired electromechanical products under two-layer reasoning. When the processor executes the computer program, the steps in the above-mentioned various embodiments of the method for failure risk assessment for the disassembly process of retired electromechanical products under two-layer reasoning are implemented. Alternatively, when the processor executes the computer program, the functions of each module / unit in the above-mentioned various device embodiments are implemented.

[0194] Exemplarily, the computer program may be divided into one or more modules / units, which are stored in the memory and executed by the processor to complete the present invention. The one or more modules / units may be a series of computer program instruction segments capable of completing specific functions, which are used to describe the execution process of the computer program in the failure risk assessment device for the disassembly process of retired electromechanical products under the two-layer reasoning.

[0195] The failure risk assessment device for the disassembly process of retired electromechanical products under the two-layer reasoning can be a computing device such as a desktop computer, a notebook, a PDA, and a cloud server. The failure risk assessment device for the disassembly process of retired electromechanical products under the two-layer reasoning may include, but is not limited to, a processor and a memory. Those skilled in the art can understand that the schematic diagram is only an example of the failure risk assessment device for the disassembly process of retired electromechanical products under the two-layer reasoning, and does not constitute a limitation on the failure risk assessment device for the disassembly process of retired electromechanical products under the two-layer reasoning, and may include more or fewer components than shown in the figure, or a combination of certain components, or different components. For example, the failure risk assessment device for the disassembly process of retired electromechanical products under the two-layer reasoning may also include input and output devices, network access devices, buses, etc.

[0196] The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the failure risk assessment device for the disassembly process of retired electromechanical products under the double-layer reasoning, and uses various interfaces and lines to connect various parts of the failure risk assessment device for the disassembly process of retired electromechanical products under the double-layer reasoning.

[0197] The memory can be used to store the computer program and / or module, and the processor realizes various functions of the failure risk assessment device for the disassembly process of retired electromechanical products under the dual-layer reasoning by running or executing the computer program and / or module stored in the memory, and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, an application required for at least one function (such as a sound playback function, an image playback function, etc.), etc.; the data storage area can store data created according to the use of the mobile phone (such as audio data, a phone book, etc.), etc. In addition, the memory can include a high-speed random access memory, and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0198] Wherein, if the module / unit integrated in the failure risk assessment device for the disassembly process of retired electromechanical products under the two-layer reasoning is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when executed by the processor. Wherein, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form, etc. The computer-readable medium can include: any entity or device that can carry the computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electrical carrier signal, telecommunication signal and software distribution medium, etc. It should be noted that the content contained in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media does not include electrical carrier signals and telecommunication signals.

[0199] It should be noted that the device embodiments described above are merely schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. In addition, in the accompanying drawings of the device embodiments provided by the present invention, the connection relationship between the modules indicates that there is a communication connection between them, which may be specifically implemented as one or more communication buses or signal lines. A person of ordinary skill in the art may understand and implement it without paying any creative effort.

[0200] The above is a preferred embodiment of the present invention. It should be pointed out that a person skilled in the art can make several improvements and modifications without departing from the principle of the present invention. These improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A failure risk assessment method for the disassembly process of retired electromechanical products under two-level reasoning, characterized in that: include: Starting from the failure level of components of retired electromechanical products, the failure behavior of the disassembly process of retired electromechanical products is determined, and by introducing fuzzy reasoning methods, multi-dimensional failure variables are processed to accurately describe the failure state of a single disassembly process of retired electromechanical products. The Bayesian network topology structure used to reflect the causal relationship between failure disassembly processes of retired electromechanical products is optimized using the improved discrete firefly algorithm. The expected maximum parameter estimation method is used to estimate the Bayesian network parameters of each node of the Bayesian network topology structure to estimate the failure conditional probability of each disassembly process of retired electromechanical products; The Bayesian network topology structure optimized by using the improved discrete firefly algorithm to reflect the causal relationship between the failure disassembly processes of retired electromechanical products includes: Constructing a disassembly process dataset: Based on the disassembly experience and historical data of retired electromechanical products, a disassembly process dataset is established. This dataset contains the status information of different disassembly processes and the actual causal relationship between them. The dataset records the execution results, failure conditions and impact of each process on subsequent processes during the disassembly process. Define individual encoding and decoding rules: In the discrete firefly algorithm, each individual represents a possible Bayesian network topology; individuals are encoded in the form of an adjacency matrix, where the elements in the matrix represent the dependencies between processes; Initialize the population: Generate the initial population based on the priority and order of the actual disassembly process using Kahn's topological sorting method; each individual is a directed acyclic graph (DAG) that satisfies the logical relationship of the disassembly process; Fitness function design: Design a fitness function to measure the quality of individuals for the Bayesian network topology; Improved discrete firefly algorithm optimization: including three main steps: search, update, and evaluation; in the search phase, the firefly swarm randomly explores the solution space to find a better network structure; in the update phase, the firefly position is adjusted according to the fitness value and the attraction and repulsion mechanism of the algorithm; in the evaluation phase, the fitness function is used to evaluate each individual and retain or update a better topological structure; Iterative optimization and convergence judgment: Through multiple iterations, the position of the firefly colony is continuously optimized, so as to continuously find a better Bayesian network topology structure; after each iteration, check whether the algorithm has reached the preset stop condition, and end the optimization process after reaching the stop condition; the stop condition includes: the upper limit of the number of iterations or the threshold of fitness improvement; Output the optimal structure: When the optimization process ends, the individual with the highest fitness is selected as the optimal solution, thereby obtaining a Bayesian network topology that best reflects the failure causal relationship between the disassembly processes of retired electromechanical products.

2. The failure risk assessment method for the disassembly process of retired electromechanical products under double-layer reasoning as claimed in claim 1 is characterized in that: Starting from the failure level of components of retired electromechanical products, the failure behavior of the disassembly process of retired electromechanical products is determined, and by introducing fuzzy reasoning methods, multi-dimensional failure variables are processed to accurately describe the failure state of a single disassembly process of retired electromechanical products, including: Conduct in-depth analysis of retired electromechanical products to identify multiple dimensional factors that affect the failure of the disassembly process; dimensional factors include: reliability failure, structural failure and process failure; Collect historical data of retired electromechanical products during the dismantling process, including service time records of parts, deformation measurements and dismantling operation parameters; Based on professional knowledge and practical experience, a fuzzy rule base is constructed; the rule base defines the fuzzy relationship between the failure dimension variables and the failure status of the disassembly process; Applying the fuzzy reasoning method, the collected multi-dimensional failure variables are input into the fuzzy reasoning engine. By matching with the fuzzy rule base, the membership function of each dimension is calculated to reflect the fuzziness degree of the parts in different failure dimensions. The maximum membership principle is used to defuzzify the membership of multiple dimensions obtained by fuzzy reasoning, so as to clarify the failure status of a single disassembly process.

3. The failure risk assessment method for the disassembly process of retired electromechanical products under double-layer reasoning as claimed in claim 1 is characterized in that: The method of using the expected maximum parameter estimation method to estimate the Bayesian network parameters of each node of the Bayesian network topology structure to estimate the failure condition probability of each disassembly process of retired electromechanical products includes: Prepare training data: Based on the historical records of disassembly of retired electromechanical products, collect failure data of relevant disassembly processes; Initialize parameter estimation: For each node representing the disassembly process in the Bayesian network, a parameter estimation value is preliminarily set based on the training data, including the initial value of the conditional probability table, which is obtained based on prior knowledge or simple statistical analysis; Expected step: Under the current parameter estimation, calculate the probability of each sample being in various states when the state of its parent node is given, that is, calculate the expected count of each node; Maximization step: Based on the expected counts calculated in the expectation step, re-estimate the conditional probability distribution of each node so that these estimates maximize the likelihood of the data; Iterative optimization: Repeat the expected step and the maximization step until the change in network parameters is less than the preset threshold or reaches the predetermined maximum number of iterations; Calculate the conditional probability of failure: Through the above iterative process, the conditional probability table finally obtained reflects the conditional probability of failure of each disassembly process given the state of its parent node; these probabilities reflect the uncertainty of failure of the disassembly process and the mutual influence between processes.

4. A failure risk assessment device for the disassembly process of retired electromechanical products under double-layer reasoning, characterized in that: include: The fuzzy reasoning module is used to determine the failure behavior of the disassembly process of retired electromechanical products based on the failure level of the components of retired electromechanical products, and to accurately describe the failure state of a single disassembly process of retired electromechanical products by introducing fuzzy reasoning methods to process multidimensional failure variables; An optimization module is used to optimize the Bayesian network topology structure used to reflect the causal relationship between failure disassembly processes of retired electromechanical products using an improved discrete firefly algorithm; An estimation module, used for estimating the Bayesian network parameters of each node of the Bayesian network topology structure by using the expected maximum parameter estimation method, so as to estimate the failure condition probability of each disassembly process of the retired electromechanical product; Wherein, the optimization module is specifically used for: Constructing a disassembly process dataset: Based on the disassembly experience and historical data of retired electromechanical products, a disassembly process dataset is established. This dataset contains the status information of different disassembly processes and the actual causal relationship between them. The dataset records the execution results, failure conditions and impact of each process on subsequent processes during the disassembly process. Define individual encoding and decoding rules: In the discrete firefly algorithm, each individual represents a possible Bayesian network topology; individuals are encoded in the form of an adjacency matrix, where the elements in the matrix represent the dependencies between processes; Initialize the population: Generate the initial population based on the priority and order of the actual disassembly process using Kahn's topological sorting method; each individual is a directed acyclic graph (DAG) that satisfies the logical relationship of the disassembly process; Fitness function design: Design a fitness function to measure the quality of individuals for the Bayesian network topology; Improved discrete firefly algorithm optimization: including three main steps: search, update, and evaluation; in the search phase, the firefly swarm randomly explores the solution space to find a better network structure; in the update phase, the firefly position is adjusted according to the fitness value and the attraction and repulsion mechanism of the algorithm; in the evaluation phase, the fitness function is used to evaluate each individual and retain or update a better topological structure; Iterative optimization and convergence judgment: Through multiple iterations, the position of the firefly colony is continuously optimized, so as to continuously find a better Bayesian network topology structure; after each iteration, check whether the algorithm has reached the preset stop condition, and end the optimization process after reaching the stop condition; the stop condition includes: the upper limit of the number of iterations or the threshold of fitness improvement; Output the optimal structure: When the optimization process ends, the individual with the highest fitness is selected as the optimal solution, thereby obtaining a Bayesian network topology that best reflects the failure causal relationship between the disassembly processes of retired electromechanical products.

5. The failure risk assessment device for the disassembly process of retired electromechanical products under double-layer reasoning as claimed in claim 4 is characterized in that: The fuzzy reasoning module is specifically used for: Conduct in-depth analysis of retired electromechanical products to identify multiple dimensional factors that affect the failure of the disassembly process; dimensional factors include: reliability failure, structural failure and process failure; Collect historical data of retired electromechanical products during the dismantling process, including service time records of parts, deformation measurements and dismantling operation parameters; Based on professional knowledge and practical experience, a fuzzy rule base is constructed; the rule base defines the fuzzy relationship between the failure dimension variables and the failure status of the disassembly process; Applying the fuzzy reasoning method, the collected multi-dimensional failure variables are input into the fuzzy reasoning engine. By matching with the fuzzy rule base, the membership function of each dimension is calculated to reflect the fuzziness degree of the parts in different failure dimensions. The maximum membership principle is used to defuzzify the membership of multiple dimensions obtained by fuzzy reasoning, so as to clarify the failure status of a single disassembly process.

6. The failure risk assessment device for the disassembly process of retired electromechanical products under double-layer reasoning as claimed in claim 4 is characterized in that: The estimation module is specifically used for: Prepare training data: Based on the historical records of disassembly of retired electromechanical products, collect failure data of relevant disassembly processes; Initialize parameter estimation: For each node representing the disassembly process in the Bayesian network, a parameter estimation value is preliminarily set based on the training data, including the initial value of the conditional probability table, which is obtained based on prior knowledge or simple statistical analysis; Expected step: Under the current parameter estimation, calculate the probability of each sample being in various states when the state of its parent node is given, that is, calculate the expected count of each node; Maximization step: Based on the expected counts calculated in the expectation step, re-estimate the conditional probability distribution of each node so that these estimates maximize the likelihood of the data; Iterative optimization: Repeat the expected step and the maximization step until the change in network parameters is less than the preset threshold or reaches the predetermined maximum number of iterations; Calculate the failure conditional probability: Through the above iterative process, the final conditional probability table reflects the failure conditional probability of each disassembly process given the state of its parent node; These probabilities reflect the uncertainty of failure in the disassembly process and the mutual influence between processes.

7. A failure risk assessment device for the disassembly process of retired electromechanical products under double-layer reasoning, characterized in that: It includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, and when the processor executes the computer program, it implements the failure risk assessment method for the disassembly process of retired electromechanical products under the two-layer reasoning as described in any one of claims 1 to 3.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored computer program, wherein, when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the failure risk assessment method for the disassembly process of retired electromechanical products under the two-layer reasoning as described in any one of claims 1 to 3.

Images