Intelligent control method and system for stamping robot

Abstract

The invention relates to the technical field of intelligent manufacturing, and discloses an intelligent control method and system for a stamping robot, and the method comprises the steps: constructing a knowledge graph containing the multi-dimensional relation of the quality of material technology equipment; the process knowledge is converted into a formalized logic rule, and the formalized logic rule is fused with a deep neural network to construct a neural symbol inference model; modeling the stamping process into a discrete event system, and verifying the correctness of a process control rule by using a linear temporal logic protocol; detecting and solving parameter constraint conflicts in the process control rule by using a satisfiability model theory solver; generating and executing a self-adaptive stamping control strategy based on the verified rule base and production process feedback; according to the method, symbol reasoning and deep learning are fused, the technical characteristics of high interpretability, high safety, high adaptive capacity and the like are achieved, and the quality consistency and production efficiency of stamping production can be effectively improved.

Description

Technical Field

[0001] The present invention relates to the field of intelligent manufacturing technology, and more specifically, to an intelligent control method and system for a stamping robot. Background Art

[0002] With the rapid development of industry and intelligent manufacturing, stamping automation, as an important part of the manufacturing industry, has put forward higher requirements for product quality stability and production efficiency. The complexity of stamping process parameter selection, the empirical dependence of process knowledge, and various uncertainties in the production process have made the intelligent control of stamping robots face many challenges.

[0003] At present, the control of stamping robots mainly adopts traditional PID control or simple rule-based control methods. These methods mainly rely on manual experience for parameter adjustment, lack systematic knowledge management and reasoning capabilities, and are difficult to cope with changes in process parameter requirements caused by factors such as fluctuations in material properties and changes in environmental conditions. In recent years, although the control method based on deep learning has powerful data fitting capabilities, its black box characteristics lead to opaque decision-making processes and lack of explainability, making it difficult to gain the trust and acceptance of operators. In addition, the existing stamping control methods generally lack a strict verification mechanism for control rules, and cannot guarantee safety and consistency under all possible operating conditions. The complex constraint relationship between process parameters easily leads to conflicts between control rules, and the lack of effective conflict detection and resolution methods further limits the reliability and stability of the control system.

[0004] Therefore, the main technical problem currently faced in the field of intelligent control of stamping robots is how to build a control system that is both explainable, safe and adaptable, and can systematically manage process knowledge, strictly verify control rules, intelligently resolve parameter conflicts, and continuously optimize control strategies based on production feedback. Summary of the invention

[0005] The present invention provides an intelligent control method and system for a stamping robot, which solve the technical problems in the related art of the stamping robot control, namely, lack of explainability, insufficient safety verification mechanism and limited adaptive capability.

[0006] The present invention provides an intelligent control method for a stamping robot, comprising the following steps: Construct a knowledge graph containing the multi-dimensional relationship between material, process, equipment and quality, identify the causal relationship between parameters and form a process knowledge base; Convert process knowledge into formal logic rules and integrate them with deep neural networks to build a neural symbolic reasoning model; The stamping process is modeled as a discrete event system, and the correctness and consistency of the process control rules are verified using linear temporal logic specifications; Detect and resolve parameter constraint conflicts in process control rules using satisfiability modulo theory solvers; Generate and execute adaptive stamping control strategies based on a proven rule base and production process feedback, and continuously optimize the control system through a closed-loop feedback mechanism.

[0007] In a preferred embodiment, the step of constructing a knowledge graph containing multidimensional relationships of material, process, equipment and quality comprises: Collect multi-source heterogeneous data from stamping process database, equipment operation log and expert experience documents; Use named entity recognition and relation extraction algorithms to identify key entities and dependency relationships from preprocessed data; Building a knowledge graph: ; in, represents the knowledge graph, Represents a collection of entities, Represents a set of relations; Causal inference algorithms are applied to identify causal relationships between parameters, and the strength and direction of causal relationships are determined through conditional independence test statistics: ; in Represents the conditional independence test statistic, which is used to measure the Under the condition of and independence, Represents a variable , , Sum all the values ​​of , Indicates the calculation of the log-likelihood ratio of conditional independence. and In a given When the condition is independent, the value is 0; otherwise, the larger the value, the and The stronger the correlation, Indicator , , The joint probability distribution of and Respectively indicate that in the given Under the conditions and The conditional probability distribution of .

[0008] In a preferred embodiment, the joint optimization function of the neural symbolic reasoning model is: ; in, represents the total loss function of the neural symbolic reasoning model, Represents the neural network loss The weight coefficient of Indicates logical consistency loss The weight coefficient of is the neural network loss function, It is the loss of logical consistency; ; ; in, is the sample size, For neural networks, For the input samples, For the The true output of samples is is a set of logical rules, Representation Rules In neural networks Satisfaction under output.

[0009] In a preferred embodiment, the step of modeling the stamping process as a discrete event system comprises: Define a discrete event system: ; in, is a discrete event system, is a combination of process parameters and equipment status. is the initial state, To control the command set, is the state transfer function; Create a time automaton: ; in, represents a timed automaton, is the location set, is the initial position, is a set of clock variables, is a set of transition edges with clock constraints, is a position-invariant function; The process safety and activity conditions are transformed into linear temporal logic formulas for specification definition.

[0010] In a preferred embodiment, the step of detecting and resolving parameter constraint conflicts using a satisfiability modulo theory solver comprises: Convert logical rules into constraint formulas for SMT problems ; Apply SMT solver to check the satisfiability of constraint formula set; Constructing a conflict graph for analysis of detected conflicts; Generate resolution strategies for different types of conflicts , including parameter range redivision, rule priority assignment, and supplementary rule generation; Update the rule base and confirm conflict elimination through formal verification.

[0011] In a preferred embodiment, the step of generating and executing the adaptive stamping control strategy includes: Real-time collection of process parameter data through sensor networks ; Apply Kalman filter to estimate the current state of the system and predict the future state; Calling the neural symbolic reasoning model to generate control decisions, where the control decisions are obtained by weighted combination of rule reasoning results and neural network prediction results, and the weighting coefficients are dynamically adjusted according to the accuracy of historical decisions; Convert control decisions into robot control instructions and establish a closed-loop feedback mechanism; Neural network parameter update, rule optimization and control parameter adaptive adjustment are achieved based on execution feedback data.

[0012] In a preferred embodiment, the formula for estimating the current state of the system using the Kalman filter is: ; in, express The estimated value of the system state at time express The estimated value of the system state at time is the state transfer matrix, is the control input matrix, is the control input, is the observed value, is the observation matrix, is the Kalman gain; The formula for the weighted combination rule reasoning result and the neural network prediction result is: ; in, express The optimal control decision at the moment, is the rule reasoning result, For the neural network prediction results, is the adaptive weight coefficient.

[0013] In a preferred embodiment, the rule optimization is performed based on a rule importance metric: ; in, Representation Rules The importance measure, represents the number of time steps in a control cycle, Representation Rules Decision-making The contribution of Indicates Logical rules, Indicates The optimal control decision for each time step; The neural network parameters are updated using stochastic gradient descent: ; in, represents the parameter vector of the neural network, is the learning rate, is the actual observed value, is the loss function, Represents the loss function About parameters The gradient of Indicates in status Below is the predicted value output by the neural network.

[0014] In a preferred embodiment, during the reasoning process of the neural symbolic reasoning model, the importance score of the input feature is calculated by the reverse attribution technique: ; in, Represents input features The importance score of Represents the neural network output About input features The partial derivative of Indicates that the input sample Calculate partial derivatives at ; Combined with the activated logic rules, a decision explainability report is generated, allowing operators to understand the basis and process of system decisions.

[0015] In a preferred embodiment, an intelligent control system for a stamping robot includes: The knowledge graph construction module is used to construct a knowledge graph containing the multi-dimensional relationship between materials, processes, equipment and quality, and to identify the causal relationship between parameters to form a process knowledge base; The neural symbolic reasoning module is used to convert process knowledge into formal logic rules and integrate it with the deep neural network to build a neural symbolic reasoning model; Formal verification module, which is used to model the stamping process as a discrete event system and verify the correctness and consistency of the process control rules using linear temporal logic specifications; A conflict detection and resolution module is used to detect and resolve parameter constraint conflicts in process control rules using a satisfiability modulus theory solver; The adaptive control module is used to generate and execute adaptive stamping control strategies based on a proven rule base and production process feedback, and to achieve continuous optimization of the control system through a closed-loop feedback mechanism.

[0016] The beneficial effects of the present invention are: Improve system safety and reliability: Through the formal verification mechanism, the correctness and consistency of process control rules can be strictly verified to ensure that the control system will not produce operations that violate safety constraints in any state.

[0017] Enhanced explainability of decision-making process: The neural symbolic fusion model combines the predictive power of deep learning with the explainability of symbolic logic to make the decision-making process of the control system transparent. The system is able to generate detailed decision explanation reports, including the key factors affecting the decision and their weights, so that operators can understand and trust the system decision.

[0018] Improve parameter optimization efficiency: The parameter constraint conflict detection and resolution mechanism based on the SMT solver can quickly identify and resolve conflicting constraints between process parameters, greatly reducing parameter tuning time.

[0019] Achieve precise control and adaptive optimization: The adaptive control strategy combines rule reasoning and deep learning to dynamically adjust control parameters according to changes in operating conditions.

[0020] Reduce system development and maintenance costs: Through automated knowledge acquisition, rule verification, and conflict resolution mechanisms, the labor costs of system development and debugging are greatly reduced. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 It is a flow chart of an intelligent control method for a stamping robot of the present invention; Figure 2 It is a detailed flow chart of the present invention for forming a process knowledge base for identifying the causal relationship between parameters; Figure 3 is a detailed flow chart of constructing a neural symbolic reasoning model of the present invention; Figure 4 It is a detailed flow chart for verifying the correctness and consistency of the process control rules of the present invention; Figure 5is a detailed flow chart of parameter constraint conflicts in the control rules of the present invention; Figure 6 It is a detailed flow chart of the present invention for realizing continuous optimization of the control system through a closed-loop feedback mechanism. DETAILED DESCRIPTION

[0022] The subject matter described herein will now be discussed with reference to example implementations. It should be understood that the discussion of these implementations is only to enable those skilled in the art to better understand and implement the subject matter described herein, and the functions and arrangements of the elements discussed may be changed without departing from the scope of protection of the present specification. Various examples may omit, replace, or add various processes or components as needed. In addition, the features described in some examples may also be combined in other examples.

[0023] At least one embodiment of the present invention discloses an intelligent control method for a stamping robot, such as Figures 1 to 6 As shown, the following steps are included: Step 1: construct a knowledge graph containing the multidimensional relationship of material, process, equipment and quality, identify the causal relationship between parameters and form a process knowledge base; The specific steps include: Step 1.1, data collection and preprocessing; Multi-source heterogeneous data are collected from stamping process databases, equipment operation logs and expert experience documents, and natural language processing technology is used to extract process parameters, operating rules and experience knowledge from text data. Through data cleaning and standardization, a structured data set is formed.

[0024] Step 1.2, entity and relationship extraction; The named entity recognition algorithm is used to identify key entities (such as material type, process parameters, equipment status, quality indicators, etc.) from the preprocessed data, and the relationship extraction algorithm is applied to detect the dependencies between entities, and the identified entities and relationships are mapped to the predefined ontology model.

[0025] Step 1.3, knowledge graph construction; Based on the extracted entities and relationships, a knowledge graph containing multidimensional relationships of material, process, equipment and quality is constructed ,in, represents the knowledge graph, Represents a collection of entities, Represents a set of relationships. Each node in the graph Represents a specific process entity (such as steel plate hardness, stamping speed, etc.). Each edge Represents the relationship between entities (such as influence, decision, etc.).

[0026] Step 1.4, causal relationship mining; Apply causal reasoning algorithms to analyze entity association patterns in the knowledge graph and calculate conditional independence test statistics: ; in Represents the conditional independence test statistic, which is used to measure the Under the condition of and independence, Represents a variable , , Sum all the values ​​of , Indicates the calculation of the log-likelihood ratio of conditional independence. and In a given When the condition is independent, the value is 0; otherwise, the larger the value, the and The stronger the correlation, the better. The causal relationship between parameters is identified through algorithms and added to the knowledge graph to form a stamping process knowledge base rich in causal information.

[0027] Step 2: Convert process knowledge into formal logic rules and integrate them with deep neural networks to build a neural symbolic reasoning model; The specific steps include: Step 2.1, logical rule encoding; Extract process knowledge from the knowledge graph and encode it into a set of formal logic rules: ; in, , , Respectively represent , , Logical rules, Represents the total number of logical rules, each of which is represented by first-order predicate logic.

[0028] Step 2.2, neural network model training; Training a deep neural network based on historical stamping data , which is used to capture implicit patterns in the data. The network input is the stamping condition feature vector (including material properties, stamping parameters, etc.), the output is the predicted quality index or process parameters The network structure adopts an architecture that combines a multi-layer perceptron (MLP) with an attention mechanism to enhance the perception of key features.

[0029] Step 2.3, neural symbolic fusion model construction; Design joint optimization function: ; represents the total loss function of the neural symbolic reasoning model, Represents the neural network loss The weight coefficient of Indicates logical consistency loss The weight coefficient of is the neural network loss function, It is the loss of logical consistency; ; ; in, is the sample size, For neural networks, For the input samples, For the The true output of samples is is a set of logical rules, Representation Rules In neural networks Satisfaction under output.

[0030] Step 2.4, interpretability enhancement; During inference, for each prediction, the importance score of the input features is calculated using a backward attribution technique: ; in, Represents input features The importance score of Represents the neural network output About input features The partial derivative of Indicates that the input sample Calculate partial derivatives at ; Combined with the activated logic rules, an explainable report is generated, allowing operators to understand the basis and process of system decisions.

[0031] Step 3, model the stamping process as a discrete event system and use linear temporal logic specifications to verify the correctness and consistency of the process control rules; The specific steps include: Step 3.1, discrete event system modeling; Model the stamping process as a discrete event system: ; in, is a discrete event system, is a combination of process parameters and equipment status. is the initial state, To control the command set, is the state transfer function; Step 3.2, time automaton construction; Create a timed automaton based on a discrete event model: ; in, represents a timed automaton, is the location set, is the initial position, is a set of clock variables, is a set of transition edges with clock constraints, is a position-invariant function; Step 3.3, linear temporal logic specification definition; The safety, activity and constraints of the stamping process are converted into linear temporal logic (LTL) formulas: ; in, Represents the entire LTL specification formula, It is a global temporal operator, requiring the condition to hold in all states; It is a final tense operator, requiring the condition to be true in a certain future state; , , is an atomic proposition, representing a Boolean property of the system state. For example, It means that when the pressure is greater than 200, good quality will be obtained in the end.

[0032] Step 3.4, model verification execution; Verifying Timed Automata Using Model Checking Algorithms Whether it meets the LTL regulations The algorithm traverses the system state space to check whether there is an execution path that violates the specification. If a counterexample is found, a detailed counterexample path is generated to point out the problem in the rule; if the verification passes, it proves that the process rule remains consistent and correct in all reachable states.

[0033] Step 4, using a satisfiability modulo theory solver to detect and resolve parameter constraint conflicts in process control rules; The specific steps include: Step 4.1, constraint formula conversion; The logical rules in step 2 are combined The constraint formula set converted to SMT problem is: ; in, Represents a set of rules Transformed into a set of constraint formulas for the SMT problem, , , Respectively represent , , SMT constraint formula, Represents the total number of SMT constraint formulas, each constraint formula Corresponding to a process rule , describing the constraint relationship between parameters. During conversion, the variable types in the rules are mapped to data types supported by SMT theory, such as real numbers, integers, Boolean values, etc., and the functions, predicates, and quantifiers in the rules are retained.

[0034] Step 4.2, conflict detection; Apply SMT solver to check constraint formula set The solver uses the DPLL (T) algorithm combined with the theoretical solver to find a solution that satisfies all constraints, or proves that the constraint set is unsatisfiable. For unsatisfiable constraint sets, the solver generates an unsatisfiable core (UNSATCore) ,in is the smallest unsatisfiable subset, indicating the specific constraints that conflict.

[0035] Step 4.3, conflict analysis; For each conflict detected, the relationship between the conflicting constraints is analyzed and a conflict graph is constructed: ; in, Represents the conflict graph, node Indicates the constraints involved in the conflict, Indicates the dependency or conflict relationship between constraints. Through graph analysis, identify the core cause of the conflict and determine whether the conflict is caused by overlapping parameter value ranges, logical contradictions, or incomplete conditions.

[0036] Step 4.4, conflict resolution strategy generation; Based on the conflict analysis results, a set of conflict resolution strategies is automatically generated: ; in represents a set of conflict resolution strategies, , , Respectively represent , , conflict resolution strategies, Indicates the total number of conflict resolution strategies. Different resolution strategies are applied to different types of conflicts: For parameter range conflicts, the parameter domain is re-divided through mathematical programming methods to find a feasible solution that satisfies all constraints: ; in, Indicates conflict resolution strategies, Represents solving the optimization problem The solution, As parameters, is the original value, is the weight, is the distance function.

[0037] For logical conflicts, a priority-based rule coordination mechanism is applied to assign priorities to conflicting rules: ; in, Represents a set of rule priorities, , , Respectively represent , , The priority of a rule, Indicates the total number of rule priorities. Rules are applied in order of priority at execution time.

[0038] Automatically generate supplementary rules for incomplete conditions Fill in the logic gaps to ensure that the control rules have defined behavior for all possible operating conditions.

[0039] Step 4.5, rule base update; Apply the solution to the original rule set to generate an optimized rule base: ; in, represents the optimized rule base, Represents a set of conflict resolution strategies Applied to the original rule set Function on ; And through formal verification, it is confirmed that the updated rule base no longer contains conflicts. The traceability of the rules is retained during the update process, and the source, modification history and application conditions of each rule are recorded to facilitate subsequent maintenance and improvement.

[0040] Step 5: Generate and execute adaptive stamping control strategy based on the verified rule base and production process feedback, and realize continuous optimization of the control system through closed-loop feedback mechanism; The specific steps include: Step 5.1, real-time data acquisition and preprocessing; Real-time collection of process parameter data through the sensor network on the stamping equipment , including material properties, equipment status, environmental conditions and quality indicators, among which, Represents a set of process parameter data collected in real time, , , Respectively represent , , The collected process parameter data points.

[0041] Data segmentation using sliding window technology ,in, Indicates A subset of data within a sliding window, is the window length, and statistical methods are applied to perform anomaly detection and data smoothing to ensure the quality of input data.

[0042] Step 5.2, state estimation and prediction; Based on the currently collected data, a Kalman filter is used to estimate the current state of the system : ; in, express The estimated value of the system state at time is the state transfer matrix, is the control input matrix, is the control input, is the observed value, is the observation matrix, is the Kalman gain. At the same time, the time series prediction model is used to predict the future The system state at each time step ,in , , Respectively represent , , system status, Represents the total number of system states and provides forward-looking information for control decisions.

[0043] Step 5.3, control decision generation; Call the neural symbolic reasoning model and combine it with the current state estimation and future state prediction, inference to generate optimal control decisions ,in, represents the control strategy function, represents the optimal control decision, Indicates the current state. The decision-making process combines two modes: Rule-based decision making: From a validated rule base Select the subset of rules that apply to the current state , through forward reasoning, we get the control decision candidate set ,in, Indicates that it applies to the current estimated state A subset of rules, Represents a validated rule base, represents the control decision candidate set obtained based on rule reasoning, , , Respectively represent , , A candidate for control decision.

[0044] Neural network-based decision making: Input the current state vector into a deep neural network to predict the optimal control parameters ,in, represents the optimal control parameters predicted by the neural network, represents the decision function based on neural network, is the network parameter.

[0045] The final control decision is generated by weighted combination of the two decision results: ,in, express The optimal control decision at the moment, is the rule reasoning result, It is an adaptive weight coefficient, which is dynamically adjusted according to the accuracy of historical decisions.

[0046] Step 5.4, execution monitoring and feedback; The control decision generated Converted into specific control instructions for stamping robots ,in, According to the optimal control decision ∗Generated specific control instructions, Represents a function that converts abstract control decisions into specific control instructions, which are sent to the execution unit through the control interface.

[0047] At the same time, a closed-loop feedback mechanism is established to monitor the execution of control instructions and the changing trend of process parameters, and calculate the control effect evaluation index. ,in, express The control effect evaluation index at each moment, It represents a function that calculates the control effect based on the current state, control decision and the state at the next moment, which is used for optimizing and adjusting the subsequent control strategy.

[0048] Step 5.5, online learning and optimization; Based on the execution feedback data, the online learning and optimization of the control system are realized. Specifically, it includes: Model parameter update: using the newly acquired state-action-reward triplet Update neural network model parameters , using stochastic gradient descent to minimize the prediction error: ,in, , , , Respectively The state, action, reward and next state at the moment. represents the parameter vector of the neural network model, is the learning rate, is the actual observed value, is the loss function, Representation parameters The gradient operator.

[0049] Rule optimization: Based on execution data analysis, regularly evaluate the effectiveness of each rule in the rule base, and measure the importance of rules (Contribution of rules to decision making) Identify inefficient rules, where Representation Rules The importance measure, represents the total number of historical decisions considered when evaluating the importance of the rule, Representation Rules right The contribution of the optimal decision at any time and the rules are simplified or enhanced.

[0050] Control parameter adaptation: Adaptively adjust key parameters in the control algorithm, such as neural symbolic fusion weights, based on dynamic changes in the process and quality feedback , predict the time domain length and control response sensitivity, etc., to optimize control performance.

[0051] The intelligent control method for a stamping robot proposed in this embodiment has the following technical advantages and effects: Improve system safety and reliability: Through the formal verification mechanism, the correctness and consistency of the process control rules can be strictly verified to ensure that the control system will not produce operations that violate safety constraints in any state. According to experimental tests, this method can reduce the safety accident rate in stamping production by more than 95%, which is far better than the traditional control method based on empirical rules.

[0052] Enhanced explainability of decision-making process: The neural symbolic fusion model combines the predictive power of deep learning with the explainability of symbolic logic to make the decision-making process of the control system transparent. The system can generate detailed decision explanation reports, including the key factors affecting the decision and their weights, so that operators can understand and trust the system decision, and the acceptance rate is increased by 87%.

[0053] Improve parameter optimization efficiency: The parameter constraint conflict detection and resolution mechanism based on the SMT solver can quickly identify and resolve conflicting constraints between process parameters, greatly reducing parameter tuning time. Compared with the traditional trial and error method, the parameter optimization efficiency has been improved by about 75%, and the new product development cycle has been shortened by more than 40%.

[0054] Enhanced knowledge accumulation and inheritance capabilities: Through knowledge graph construction and causal relationship mining, the system can transform expert experience and historical production data into structured knowledge, and express and apply it through formal rules. This method transforms the inheritance of expert knowledge from the traditional master-apprentice model to a systematic and automated knowledge acquisition and application process, and improves knowledge acquisition efficiency by about 3 times.

[0055] Achieve precise control and adaptive optimization: The adaptive control strategy combines rule reasoning and deep learning to dynamically adjust control parameters according to changes in working conditions. In actual production tests, the stamping robot control system using this method has improved product quality consistency by 28%, reduced scrap rate by 32%, and increased production efficiency by 18% compared to traditional PID control.

[0056] Reduce system development and maintenance costs: Through automated knowledge acquisition, rule verification and conflict resolution mechanisms, the labor costs of system development and debugging are greatly reduced. At the same time, the system's online learning and adaptive optimization capabilities enable it to be continuously improved, reducing the workload of later maintenance and reducing the total cost of ownership (TCO) by about 25%.

[0057] Strong adaptability and easy expansion: This method adopts modular design, and the interfaces between functional modules are clear, which is convenient for customization and expansion for different stamping process scenarios. The system can adapt to the stamping process requirements of different materials, different equipment and different products without changing the core architecture, and the adaptability of application scenarios is expanded to 23 times that of traditional methods.

[0058] Real-world application examples of this implementation: This implementation method has been put into practical application on a high-strength steel plate stamping production line of a certain auto parts manufacturing company. The production line mainly produces auto body structural parts, including A-pillar reinforcements, B-pillars, crossbeams and other safety-critical parts, which have extremely high requirements for stamping accuracy and quality consistency. The production line is equipped with 6 315-ton servo presses and 4 robots, with an annual production capacity of approximately 1.5 million pieces.

[0059] Implementation process example: Examples of knowledge acquisition and graph construction: In the application of this automotive parts company, the following multi-source heterogeneous data were first collected: Historical production data: The production records of the past three years were exported from the MES system, including about 280,000 stamping process data, recording the process parameters and quality inspection results under different product, material and equipment combinations.

[0060] Equipment operation log: About 15GB of operation log data was exported from the control systems of 6 presses and 4 robots, including equipment status, operating parameters, and alarm information.

[0061] Expert knowledge documents: About 120 process guidance documents, standard operating procedures and experience summary reports from factory technical experts were compiled.

[0062] Quality Management System Data: Quality inspection data of approximately 45,000 products were collected, including key dimensions, surface quality and mechanical properties test results.

[0063] Through data processing and knowledge extraction, a stamping process knowledge graph containing 5 categories of entities and 12 relationships was constructed. The core content statistics of the knowledge graph are shown in Table 1: Table 1: Statistics of stamping process knowledge graph;

[0064] In the causal relationship mining stage, the PC algorithm and NOTEARS algorithm were used to perform causal analysis on the association relationships in the graph, and 3,245 relationships with clear causality were identified from 15,634 relationships.

[0065] Example of building a neural symbolic reasoning model: During the model building phase, 642 logical rules were extracted from the knowledge graph, covering aspects such as material parameter selection, quality control, and fault handling. Some typical rule examples are shown in Table 2: Table 2: Examples of typical logic rules;

[0066] Next, a neural network model combining a multi-layer perceptron and an attention mechanism was trained based on historical production data. The network structure includes: Input layer: 68 feature nodes, representing material properties, equipment status, and process parameters 3 hidden layers: containing 256, 128, and 64 nodes respectively, using the ReLU activation function Attention layer: 8-head self-attention mechanism to capture the relationship between features Output layer: 12 nodes, representing key quality characteristic predictions and process parameter recommendations.

[0067] Finally, by designing a joint optimization function, the neural network and the logic rules are integrated for training, where the logic consistency loss weight It is initially set to 0.3 and gradually increases to 0.7 as training progresses to ensure that the model can both learn implicit patterns in the data and follow expert experience rules.

[0068] Formal verification mechanism implementation example: In this application, the stamping process is first modeled as a discrete event system with 7 main states, including: standby state, material loading, preheating stage, stamping preparation, stamping execution, product unloading and quality inspection. For each state, the corresponding parameter constraints and transition conditions are defined.

[0069] Based on this discrete event model, a timed automaton was constructed to describe the timing behavior of the stamping process. The timed automaton contains 12 position nodes, 17 transition edges, and 4 clock variables, and describes in detail the timing constraints of the process flow, such as material preheating time, stamping execution time, and equipment reset time.

[0070] The key process safety and activity requirements are converted into linear temporal logic (LTL) formulas. Some key LTL specification examples are shown in Table 3: Table 3: Examples of key LTL specifications;

[0071] The constructed timed automaton was verified using the UPPAAL model checking tool to check whether it meets all defined LTL specifications. In the initial verification phase, 23 violations of specifications were detected, mainly focusing on safety protection and timing constraints under extreme conditions. Some verification results and corresponding corrective measures are shown in Table 4: Table 4: Some verification results and corrective measures;

[0072] After three rounds of rule revision and verification, all protocols finally passed formal verification, ensuring the safety and consistency of the control system in all reachable states.

[0073] Example of parameter constraint conflict detection and resolution: In this application example, 642 process rules were converted into SMT constraint formulas and satisfiability was checked using the Z3 solver. The initial check found 87 potential conflicts, which are classified by conflict type as shown in Table 5: Table 5: Classification statistics of parameter constraint conflicts;

[0074] The system automatically generates a solution strategy for the detected conflict. The following is a specific example of resolving a parameter range conflict: Conflict rule 1 (R127): When the tensile strength of the steel plate is >980MPa and the thickness is >1.6mm, the punching speed should be <180mm / s; Conflict rule 2 (R135): When the steel plate thickness is >2.0mm, the punching speed should be >160mm / s to ensure the forming quality; These two rules have a conflicting parameter range when the tensile strength of the steel plate is >980MPa and the thickness is >2.0mm. The stamping speed must be both <180mm / s and >160mm / s. The parameter space is too narrow to adapt to fluctuations in the production process.

[0075] The system applies optimization algorithms, re-divides the parameter domain, and gives the solution: Rule R127 is modified to: When the tensile strength of the steel plate is greater than 980 MPa and the thickness is within the range of [1.6 mm, 2.0 mm], the punching speed should be less than 180 mm / s; New rule R127b: When the tensile strength of the steel plate is >980MPa and the thickness is >2.0mm, the punching speed should be controlled within the range of [165mm / s, 175mm / s]; Keep the original conditions of rule R135 but lower its application priority; Similarly, corresponding resolution strategies are generated for all detected conflicts, and these strategies are applied to the original rule base to generate an optimized rule base. The optimized rule base has passed formal verification to confirm that there are no conflicts.

[0076] Adaptive control strategy generation and execution example: In this application example, the sensor network of the stamping production line was first upgraded, and the following sensor equipment was added: 12 sets of high-precision pressure sensors for real-time monitoring of the pressure distribution of each station 8 sets of temperature sensors for monitoring mold and material temperature 6 sets of acceleration sensors for monitoring equipment vibration 4 sets of image recognition systems for online detection of product surface quality; Through these sensor networks, the system can collect process parameter data at a frequency of 100Hz, processing about 3,600 data points per second. A 30-second sliding window is used for data segmentation with an overlap rate of 50%, ensuring good capture of both short-term fluctuations and long-term trends.

[0077] Based on the real-time collected data, the system uses a 5-state Kalman filter to estimate the current production state. The state vector contains: material properties, pressure distribution, speed curve, temperature distribution and equipment status. At the same time, the prediction model estimates the state evolution of the next 30 sampling points based on the current state, providing forward-looking information for control decisions, as shown in Table 6: Table 6: Adaptive control decision example (B-pillar reinforcement production process);

[0078] Table 6 shows examples of adaptive control decisions generated by the system for different working conditions during the production of B-pillar reinforcements. It can be seen that the system can automatically adjust the fusion weight according to different working conditions and find the best balance between empirical rules and data-driven decisions. When faced with obvious data features such as hardness fluctuations of new materials, the neural network decision weight is higher; when encountering defect processing that requires specific domain knowledge, the rule reasoning decision weight is automatically increased.

[0079] In actual applications, the control system always maintains online learning and optimization capabilities. After every 500 products are produced, the system will analyze the production data of the batch and update the neural network model parameters and rule importance scores. Through long-term operation, the system has gradually accumulated the best control strategies under different working conditions and formed an adaptive knowledge base.

[0080] Technical effect verification: After six months of actual use in an automotive parts manufacturing company, the intelligent control system was fully evaluated, mainly from two key technical effects: enhanced decision interpretability and improved product quality consistency.

[0081] Verification of decision interpretability effect: An important technical effect of this system is that it enhances the interpretability of control decisions, enabling operators to understand and trust system decisions. Before its application, traditional neural network-based control systems could only give parameter adjustment suggestions without explaining the reasons for the adjustments, resulting in low operator acceptance and an actual adoption rate of only 43%.

[0082] After applying this system, each control decision will generate a detailed explanation report, including decision basis, activation rules and feature importance analysis. The evaluation results of decision interpretability are shown in Table 7: Table 7: Comparison of decision explainability effects;

[0083] The system generates differentiated explanatory reports for different types of operators, ranging from detailed technical explanations at the engineer level to concise instructions at the operator level.

[0084] Through questionnaire surveys and on-site observations, it was found that the operators' understanding of system decisions increased from 4.2 points (out of 10 points) before application to 8.7 points, an increase of 107.1%; the decision adoption rate increased from 43.2% to 86.5%, an increase of 100.2%; the average decision time of operators decreased from 127 seconds to 42 seconds, a decrease of 66.9%. These results show that the neural symbolic fusion method of this system enhances the interpretability of the decision-making process, transforms the intelligent control system from a black box to a transparent box, and greatly improves the operators' acceptance and trust in the system.

[0085] Product quality consistency effect verification: Another key technical effect of the system is improved product quality consistency. This effect was verified by comparing the product dimensional accuracy and surface quality before and after the system was applied.

[0086] In actual production, three products with different process difficulties (A-pillar reinforcement, B-pillar and front longitudinal beam) were randomly selected, and 1,000 pieces were sampled before and after the system was applied for precision measurement, and the deviation distribution of key dimensions was statistically analyzed. The comparison results of dimensional accuracy are shown in Table 8: Table 8: Comparison of product dimensional accuracy (unit: mm);

[0087] As can be seen from Table 8, after applying this system, the average value of product dimensional deviation decreased from 0.81mm to 0.37mm, a decrease of 54.3%; the standard deviation of dimensional deviation decreased from 0.56mm to 0.21mm, a decrease of 62.5%; the deviation rate decreased from 8.64% to 1.08%, a decrease of 87.5%. This shows that this system has greatly improved the dimensional accuracy and consistency of stamping products through the adaptive control strategy based on neural symbolic reasoning and formal verification.

[0088] At the same time, the surface defects of the products were counted through the surface quality inspection system. It was found that after the application of the system, the surface defect rate of the products decreased from 5.8% to 1.9%, a decrease of 67.2%; in particular, defects related to parameter control (such as wrinkles, cracks, etc.) decreased by 76.4%, which further verified the effectiveness of this system in improving product quality consistency.

[0089] The above describes an embodiment of the present invention, but this embodiment is not limited to the above-mentioned specific implementation mode. The above-mentioned specific implementation mode is merely illustrative and not restrictive. Under the guidance of this embodiment, ordinary technicians in this field can also make more forms of equivalent embodiments, all of which are within the protection of this embodiment.

Claims

1. An intelligent control method for a stamping robot, characterized in that: The following steps are involved: Construct a knowledge graph containing the multi-dimensional relationship between material, process, equipment and quality, identify the causal relationship between parameters and form a process knowledge base; Convert process knowledge into formal logic rules and integrate them with deep neural networks to build a neural symbolic reasoning model; The stamping process is modeled as a discrete event system, and the correctness and consistency of the process control rules are verified using linear temporal logic specifications; Detect and resolve parameter constraint conflicts in process control rules using satisfiability modulo theory solvers; Generate and execute adaptive stamping control strategies based on a proven rule base and production process feedback, and continuously optimize the control system through a closed-loop feedback mechanism.

2. The intelligent control method for a stamping robot according to claim 1, characterized in that: The step of constructing a knowledge graph containing multidimensional relationships of material, process, equipment and quality comprises: Collect multi-source heterogeneous data from stamping process database, equipment operation log and expert experience documents; Use named entity recognition and relation extraction algorithms to identify key entities and dependency relationships from preprocessed data; Building a knowledge graph: ； in, represents the knowledge graph, Represents a collection of entities, Represents a set of relations; Causal inference algorithms are applied to identify causal relationships between parameters, and the strength and direction of causal relationships are determined through conditional independence test statistics: ； in Represents the conditional independence test statistic, which is used to measure the Under the condition of and independence, Represents a variable , , Sum all the values ​​of , Indicates the calculation of the log-likelihood ratio of conditional independence. and In a given When the condition is independent, the value is 0; otherwise, the larger the value, the and The stronger the correlation, Indicator , , The joint probability distribution of and Respectively indicate that in the given Under the conditions and The conditional probability distribution of .

3. The intelligent control method for a stamping robot according to claim 1, characterized in that: The joint optimization function of the neural symbolic reasoning model is: ； in, represents the total loss function of the neural symbolic reasoning model, Represents the neural network loss The weight coefficient of Indicates logical consistency loss The weight coefficient of is the neural network loss function, It is the loss of logical consistency; ； ； in, is the sample size, For neural networks, For the input samples, For the The true output of samples is is a set of logical rules, Representation Rules In neural networks Satisfaction under output.

4. The intelligent control method for a stamping robot according to claim 1, characterized in that: The step of modeling the stamping process as a discrete event system comprises: Define a discrete event system: ； in, is a discrete event system, is a combination of process parameters and equipment status. is the initial state, To control the command set, is the state transfer function; Create a time automaton: ； in, represents a timed automaton, is the location set, is the initial position, is a set of clock variables, is a set of transition edges with clock constraints, is a position-invariant function; The process safety and activity conditions are transformed into linear temporal logic formulas for specification definition.

5. The intelligent control method for a stamping robot according to claim 1, characterized in that: The steps of using the satisfiability modulo theory solver to detect and resolve parameter constraint conflicts include: Convert logical rules into constraint formulas for SMT problems ; Apply SMT solver to check the satisfiability of constraint formula set; Constructing a conflict graph for analysis of detected conflicts; Generate resolution strategies for different types of conflicts , including parameter range redivision, rule priority assignment, and supplementary rule generation; Update the rule base and confirm conflict elimination through formal verification.

6. The intelligent control method for a stamping robot according to claim 1, characterized in that: The steps of generating and executing the adaptive stamping control strategy include: Real-time collection of process parameter data through sensor networks ; Apply Kalman filter to estimate the current state of the system and predict the future state; Calling the neural symbolic reasoning model to generate control decisions, where the control decisions are obtained by weighted combination of rule reasoning results and neural network prediction results, and the weighting coefficients are dynamically adjusted according to the accuracy of historical decisions; Convert control decisions into robot control instructions and establish a closed-loop feedback mechanism; Neural network parameter update, rule optimization and control parameter adaptive adjustment are achieved based on execution feedback data.

7. The intelligent control method for a stamping robot according to claim 6, characterized in that: The formula for applying the Kalman filter to estimate the current state of the system is: ； in, express The estimated value of the system state at time express The estimated value of the system state at time is the state transfer matrix, is the control input matrix, is the control input, is the observed value, is the observation matrix, is the Kalman gain; The formula for the weighted combination rule reasoning result and the neural network prediction result is: ； in, express The optimal control decision at the moment, is the rule reasoning result, For the neural network prediction results, is the adaptive weight coefficient.

8. The intelligent control method for a stamping robot according to claim 6, characterized in that: The rule optimization is performed based on the rule importance metric: ； in, Representation Rules The importance measure, represents the number of time steps in a control cycle, Representation Rules Decision-making The contribution of Indicates Logical rules, Indicates The optimal control decision for each time step; The neural network parameters are updated using stochastic gradient descent: ； in, represents the parameter vector of the neural network, is the learning rate, is the actual observed value, is the loss function, Represents the loss function About parameters The gradient of Indicates in status Below is the predicted value output by the neural network.

9. The intelligent control method for a stamping robot according to claim 1, characterized in that: During the reasoning process of the neural symbolic reasoning model, the importance scores of the input features are calculated through the back-attribution technique: ； in, Represents input features The importance score of Represents the neural network output About input features The partial derivative of Indicates that the input sample Calculate partial derivatives at ; Combined with the activated logic rules, a decision explainability report is generated, allowing operators to understand the basis and process of system decisions.

10. An intelligent control system for a stamping robot, used to execute the steps of the intelligent control method for a stamping robot as claimed in any one of claims 1 to 9, characterized in that: include: The knowledge graph construction module is used to construct a knowledge graph containing the multi-dimensional relationship between materials, processes, equipment and quality, and to identify the causal relationship between parameters to form a process knowledge base; The neural symbolic reasoning module is used to convert process knowledge into formal logic rules and integrate it with the deep neural network to build a neural symbolic reasoning model; Formal verification module, which is used to model the stamping process as a discrete event system and verify the correctness and consistency of the process control rules using linear temporal logic specifications; A conflict detection and resolution module is used to detect and resolve parameter constraint conflicts in process control rules using a satisfiability modulus theory solver; The adaptive control module is used to generate and execute adaptive stamping control strategies based on a proven rule base and production process feedback, and to achieve continuous optimization of the control system through a closed-loop feedback mechanism.

Images