Quality safety monitoring management system for powder metallurgy production

Abstract

A quality safety monitoring management system for powder metallurgy production comprises a production parameter intelligent sensing module, a quality safety intelligent analysis module, a quality tracing management module and a man-machine interaction module. The production parameter intelligent sensing module collects production parameters, completes data preprocessing and sends a result to the quality safety intelligent analysis module; the quality safety intelligent analysis module receives production parameters, obtains key quality features through semantic ontology driven adaptive sampling, realizes fault diagnosis by applying a hybrid reinforcement learning method of tensor decomposition, dynamically adjusts a quality management strategy through fractional order chaos optimization, and generates quality early warning information, a diagnosis decision and a management plan; the quality traceability management module establishes a quality traceability chain covering the whole life cycle of raw materials, processes and products by using a block chain technology, receives quality data and completes security evidence storage; the man-machine interaction module realizes visual display of system functions and data, receives a user instruction and coordinates other modules to complete a quality monitoring management task.

Description

Technical Field

[0001] The present invention relates to the technical field of quality and safety monitoring and management systems, and more specifically, to a quality and safety monitoring and management system for powder metallurgy production. Background Art

[0002] Powder metallurgy is an advanced material forming and processing technology. By mixing, pressing and sintering powdered raw materials such as metals and ceramics, high-precision components with complex shapes and excellent performance can be manufactured. It is widely used in the fields of automobiles, aviation, and machinery manufacturing. However, the production process of powder metallurgy is complex, involving multiple links such as raw material proportioning, pressing and forming, and sintering. The quality of the production process is prone to fluctuations, and the product defect rate is high. Traditional quality management mainly relies on manual sampling and statistical analysis, which can often only make up for the loss, and it is difficult to achieve accurate early warning and pre-control.

[0003] In recent years, with the development of new generation information technologies such as sensor technology, big data analysis, and artificial intelligence, intelligent manufacturing has become a new way to improve the quality control level of powder metallurgy enterprises. By deploying intelligent sensors on production equipment, real-time online monitoring of the production process can be achieved. Domestic and foreign scholars have conducted quality-related modeling and analysis research on powder metallurgy production data. Some scholars applied wavelet packet energy spectrum and BP neural network to establish a soft measurement model for the quality of pressed billets and achieved quality prediction of the pressing process. Other scholars have constructed a diagnostic knowledge base for powder metallurgy part defects based on Bayesian theory and expert experience, which can perform root cause analysis and classification statistics on common defects.

[0004] However, the existing quality management systems and methods still have shortcomings: first, the collection of production parameters is not comprehensive, the preprocessing analysis is not intelligent, and the production status is not accurately portrayed; second, the quality warning model has poor generalization, lacks prior knowledge guidance, and the warning results are not interpretable; third, the defect diagnosis rules are solidified and lack continuous optimization and learning capabilities; fourth, the decision-making optimization method is not suitable for dynamic and complex production process regulation. In addition, most existing systems adopt a centralized management model, which lacks support in terms of supply chain traceability and cross-enterprise collaboration. It is urgent to explore a new quality management system architecture to open up the entire link of data collection, modeling analysis, regulation optimization, and traceability. Summary of the invention

[0005] In view of the above problems, the present invention proposes a quality safety monitoring and management system for powder metallurgy production. The system is based on innovative methods such as semantic ontology-driven adaptive sampling, tensor decomposition hybrid reinforcement learning, fractional-order chaos optimization quality management, and blockchain-driven quality traceability to achieve full-process intelligent control of production quality.

[0006] The present invention provides a quality and safety monitoring and management system for powder metallurgy production, comprising a production parameter intelligent perception module, a quality and safety intelligent analysis module, a quality traceability management module and a human-computer interaction module; the production parameter intelligent perception module collects production parameters and completes data preprocessing, and sends the results to the quality and safety intelligent analysis module; the quality and safety intelligent analysis module receives production parameters, obtains key quality features based on semantic ontology-driven adaptive sampling, applies a hybrid reinforcement learning method of tensor decomposition to achieve fault diagnosis, and dynamically adjusts quality management strategies through fractional-order chaos optimization to generate quality early warning information, diagnostic decisions and management plans; the quality traceability management module uses blockchain technology to establish a quality traceability chain covering the entire life cycle of raw materials, processes and products, receives quality data and completes secure evidence storage; the human-computer interaction module realizes visual display of system functions and data, accepts user instructions and coordinates other modules to complete quality monitoring and management tasks.

[0007] Specifically, the production parameter intelligent perception module includes multi-source heterogeneous sensors, edge gateways and time-sensitive networks; wherein, the multi-source heterogeneous sensors collect raw material properties, process parameters and equipment status signals, and transmit the data to the edge gateway in real time; the edge gateway performs pre-processing such as cleaning and normalization on the sensor data, and temporarily stores it after time and semantic classification; the time-sensitive network transmits the perception data of the edge gateway to the quality and safety intelligent analysis module with a delay of no more than 10ms.

[0008] Specifically, the semantic ontology-driven adaptive sampling unit in the quality and safety intelligent analysis module obtains key quality features in the following ways:

[0009] The semantic ontology of production process O=(C,R,A) which formally defines the semantic categories and associations of raw materials, equipment, process and quality, where C is the concept set, R is the relationship set and A is the axiom set; O is described using the ontology language OWL;

[0010] Based on the Tableau algorithm, the semantic ontology O is inferred to obtain the semantic association matrix W∈RN×N, where the matrix element wij∈[0,1] represents the semantic association strength between concept i and concept j, and N is the number of semantic concepts in the ontology;

[0011] Determine the semantic importance weight of each sensor acquisition parameter xi according to the association matrix W The calculation formula is Design a rotation matrix compression sampling operator Φ∈RM×M, where the matrix elements Φij obey the Bernoulli distribution sigmoid(x)=(1+e -x ) -1 is the Sigmoid function, λ is the scale factor;

[0012] Adaptively sample the sensor parameter vector x to obtain the compressed sensing vector y = Φ x i y∈RM is a low-dimensional semantic feature vector.

[0013] Specifically, the hybrid reinforcement learning fault diagnosis unit based on tensor decomposition in the quality and safety intelligent analysis module implements fault diagnosis through the following steps:

[0014] The fault-symptom-component three-dimensional semantic cubic tensor T∈R constructed based on equipment operation logs, alarm information and quality defect data I×J×K , where I is the number of failure modes, J is the number of symptom modes, and K is the number of associated components;

[0015] The multimodal tensor decomposition algorithm is used to decompose the semantic tensor T to obtain the fault factor matrix Symptom factor matrix S∈R J ×R 2 , component factor matrix V∈R K ×R 3 , and the core tensor Satisfies T≈G× 1 U× 2 S× 3 V, where × n represents the n-mode tensor product, R 1 , R 2 , R 3 are the ranks of the three dimensions respectively;

[0016] Constructing fusion prior knowledge Θ K and data-driven knowledge Θ D A hybrid fault diagnosis strategy network π(a|s; Θ), where the state s∈R J is the J-dimensional symptom observation vector, action a∈R I is the fault mode diagnosis probability vector, network parameter Θ=(Θ K ,Θ D );

[0017] Prior knowledge Θ K The tensor kernelized graph convolutional network is used for modeling, and the state feature propagation adopts a dynamic bidirectional attention propagation mechanism: Among them, H (k) ∈R N×F is the node feature matrix of the k-th layer network, F is the node feature dimension, σ(·) is the activation function, Ai and is the dynamic attention matrix;

[0018] Data-driven knowledge D Long short-term memory network modeling is used to transform the fault symptom observation sequence (s 1 ,s2 ,...,s T ) is encoded into a hidden state vector h T and the memory cell state vector c T ;

[0019] The loss function of the hybrid diagnosis strategy network is:

[0020] Where γ∈(0,1) is the discount factor, r t is the diagnostic accuracy reward, y t is the true state-action value, Q π (s t ,a t ) is the estimated state-action value.

[0021] Specifically, the quality management strategy dynamic adjustment unit based on fractional-order chaos optimization in the quality and safety intelligent analysis module realizes the search for the optimal management strategy through the following method:

[0022] Construct N-dimensional quality state vector Q = [q 1 ;q 2 ;...;q N ], where q i ∈R is the i-th quality characteristic value; construct an M-dimensional management decision vector D = [d 1 ;d 2 ;...;d M ], where d j ∈Ω is the jth adjustable management parameter, and the parameter space Ω is the parameter value range;

[0023] The Caputo fractional differential equation is used to describe the fractional dynamic characteristics of the mass state:

[0024]

[0025] in dτ is the Caputo fractional differential operator, and "(-) is the Gamma function;

[0026] Discretize the fractional-order differential equation to obtain the quality state prediction model:

[0027] Q(t+1)=Q(t)+hα·∑(k=0:t)ω k , t+1·f(Q(tk),D(tk));·

[0028] The cumulative quality loss J in the next T steps T =∑(t=1:T)L(Q(t),Q *) is the optimization target, and a quality management decision optimization model is constructed, where L(:) is the deviation of the measured quality state Q from the target Q * The loss function of

[0029] A chaotic optimization algorithm based on Logistic mapping z(k+1)=μ·z(k)·(1-z(k)) is introduced to encode and map the management decision D through the chaotic variable z: d j =d minj +z·(d maxj -d minj ), and search for the optimal decision vector D in the parameter space Ω * ; By the optimal decision vector D * Decode and generate the quality management plan for the next T steps to guide the quality improvement of the production process.

[0030] Specifically, the quality traceability management module constructs a three-layer blockchain quality traceability network in the following ways: building a raw material blockchain, a process blockchain and a product blockchain to cover the entire production life cycle; using hash pointer technology to link each layer of blockchain to achieve end-to-end quality traceability; using elliptic curve cryptography to asymmetric encrypt and digitally sign quality data to ensure data security; supporting dynamic expansion of no less than 100 nodes, and controlling the consensus delay within 1 second.

[0031] Specifically, the human-computer interaction module is designed based on B / S architecture and supports cross-platform Web access; a single service node supports no less than 1,000 concurrent Http requests;

[0032] The WebGL graphics rendering engine is used to achieve 3D visualization of production parameters, quality characteristics, diagnostic results, traceability information and other data, with a rendering frame rate of no less than 60FPS. The built-in data mining algorithm library supports data analysis functions such as correlation analysis, comparative analysis, trend prediction, etc., to assist quality management decision-making.

[0033] Specifically, the multi-source heterogeneous sensor has the following parameters: the sensor sampling frequency is not less than 1kHz, meeting the needs of online quality monitoring; the sensor range covers the key parameter range of the production process, and the indication error is no more than 1%; the edge gateway adopts an industrial-grade embedded computing platform, configured with no less than 4-core CPU, 8GB memory and 128GB solid-state hard drive, and supports multi-protocol access and edge computing.

[0034] Specifically, the fault diagnosis unit has active learning capabilities and realizes online updating of the fault knowledge base in the following ways: extracting new fault modes and symptom information based on expert feedback on the diagnosis results; dynamically expanding the dimension of the semantic tensor T and adding new tensor elements; regularly recalculating tensor decomposition and updating prior diagnostic knowledge; fine-tuning the network parameters of the hybrid diagnostic strategy to continuously improve the accuracy and generalization of fault diagnosis.

[0035] Specifically, the quality traceability module adopts the alliance chain model to organize the blockchain network and supports collaborative quality management through the following mechanisms: manufacturers, suppliers, distributors, regulatory authorities, etc. jointly act as blockchain nodes to perform identity authentication and authorization; use smart contract technology to regularize production and transaction behaviors, and realize automatic uploading and sharing of quality data; balance data sharing and security through identity-based privacy protection and fine-grained access control; provide a quality traceability interface with controllable permissions for third-party audits, and promote social collaboration in quality management.

[0036] The present invention has the following beneficial effects:

[0037] (1) A semantic ontology-driven adaptive sampling method for production parameters is proposed. By constructing the semantic ontology of the production process, mining the intrinsic semantic association between parameters and quality, and adaptively adjusting the sensor sampling rate, the acquisition accuracy of quality-related parameters can be improved while reducing data redundancy.

[0038] (2) A hybrid reinforcement learning quality diagnosis method based on tensor decomposition is proposed. The semantic representation of fault symptoms is obtained by tensor decomposition, a hybrid neural network integrating prior knowledge and data-driven is constructed, and reinforcement learning is used to optimize the diagnosis strategy, which significantly improves the accuracy of quality warning and defect diagnosis.

[0039] (3) A quality management strategy optimization method based on fractional-order chaos optimization is proposed. The quality state is dynamically modeled through fractional-order differential equations, and a chaos optimization algorithm is introduced to search for the optimal control strategy, which can realize the adaptive dynamic adjustment of management decisions.

[0040] (4) A quality traceability system integrating blockchain technology is proposed. Through the alliance chain architecture, production, supply, sales and other business collaborations can be achieved, which can support quality management and data sharing throughout the life cycle of powder metallurgy products.

[0041] (5) The pertinence and efficiency of production quality data collection have been improved, the edge computing capability has been significantly enhanced, and the semantic feature extraction has become more accurate; the accuracy of quality warning and defect diagnosis has been greatly improved, and the adaptability to complex production conditions has been stronger; the initiative and foresight of quality management have been enhanced, and decision-making optimization is more in line with the dynamic characteristics of powder metallurgy process; a reliable and traceable quality management alliance chain has been established to provide data support for the rapid location of quality problems and accountability.

[0042] In summary, the quality and safety monitoring and management system of the present invention provides a set of innovative theoretical methods and information solutions for the intelligent quality control of the powder metallurgy industry, which is of great significance for improving the quality level and market competitiveness of powder metallurgy products. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 This is a schematic diagram of the framework of the quality and safety monitoring and management system for powder metallurgy production of the present invention. DETAILED DESCRIPTION

[0044] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the specific implementation method, structure, characteristics and effects of the quality and safety monitoring and management system for powder metallurgy production proposed by the present invention are described in detail below in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0045] Unless otherwise defined, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs. The specific scheme of the quality and safety monitoring and management system for powder metallurgy production provided by the present invention is described in detail below in conjunction with the accompanying drawings.

[0046] See also Figure 1 The present invention provides a quality and safety monitoring and management system for powder metallurgy production. The system consists of a production parameter intelligent perception module 1, a quality and safety intelligent analysis module 2, a quality traceability management module 3 and a human-computer interaction module 4. The modules are tightly coupled and collaboratively optimized through data flow and control flow to jointly achieve closed-loop control of quality and safety in the production process.

[0047] Among them, the function of the production parameter intelligent perception module 2 is to collect various parameters of the powder metallurgy production process and perform preprocessing. This module is mainly composed of multi-source heterogeneous sensors, edge gateways and time-sensitive networks. In actual deployment, pressure, temperature, atmosphere and other sensors can be installed on raw material warehouses, mixers, presses, sintering furnaces and other equipment according to production process requirements to collect production parameters of each station in real time. Sensors generally use industrial-grade intelligent sensors, and the sampling frequency can reach more than 1kHz, which can meet the timeliness requirements of online quality detection. The sensor is connected to the nearest edge gateway through industrial Ethernet, bus and other means. The gateway uses a high-performance industrial computer to clean, normalize, compress and other preprocessing of sensor data, and classify and store them according to semantic types to prepare for subsequent semantic extraction. At the same time, the gateway uploads the preprocessed perception data to the quality and safety intelligent analysis module through a time-sensitive network (such as TSN). The time-sensitive network can provide deterministic data transmission delay, which can generally be controlled within 10ms.

[0048] The quality and safety intelligent analysis module 2 is the core of this system. This module uses artificial intelligence technologies such as semantic extraction, reinforcement learning, and chaos optimization to mine hidden quality defect patterns from production parameters and realize intelligent early warning, diagnosis, and optimization of the production process. This module first uses an adaptive sampling method based on semantic ontology to extract key quality features. The specific steps are as follows:

[0049] Step 1: Construct the semantic ontology of the powder metallurgy production process. Process experts and knowledge engineers jointly define the main semantic concepts including raw materials, equipment, process, and quality, and use ontology languages ​​(such as OWL) to formally describe the relationship between concepts, such as the pressing pressure affects the density of the green body, the sintering temperature determines the grain size, etc., to form the knowledge graph of the semantic ontology. The number of concept nodes in the ontology is generally in the order of 100 to 1000.

[0050] Step 2: Use the Tableau algorithm to infer the strength of semantic association. The Tableau algorithm is a knowledge reasoning method based on deductive logic. It can deduce implicit concept relationships from the ontology through the induction and refutation mechanism. Assuming that the ontology contains N concept nodes, Tableau reasoning will generate an N×N semantic association matrix W, and the matrix element w ij ∈[0,1] represents the quantitative value of the semantic association strength between concept j. ij The larger the value, the closer the semantic relationship between i and j. ij If the value is greater than 0.6, the relationship between the two concepts is considered significant.

[0051] Step 3: Calculate the semantic importance weight of the production parameters. Suppose the production parameters collected by the sensor are x 1 ,x 2,...,x M , then the parameter x i The semantic importance weight of The formula Calculated, where w i j is x i The strength of association between the semantic concept j and the quality concept j. Quantized parameter x i Salience in terms of quality semantics. The range of is [0,1], and The larger the x i The higher the quality relevance, the higher the Parameters greater than 0.5 can be considered as critical quality influencing factors.

[0052] Step 4: Construct a round-robin sampling operator based on parameter semantic weights. In order to efficiently extract key parameters related to quality semantics during online acquisition, the system designs a round-robin matrix compression sampling operator Φ. Let Φ be an M×M matrix (M is the total number of parameters), and the matrix element Φ ij Bernoulli distribution The scale factor λ controls the semantic weight The steepness of the mapping from to sampling probability can be simply taken as 5 to 10. The Sigmoid function Map from the interval (0,1) to the interval (0.5,1) to facilitate probabilistic calculation. Step 5: For the production parameter vector x = [x 1 ;x 2 ;.x M ] to perform adaptive semantic sampling. y = Φ x, where y is the compressed low-dimensional semantic feature vector. ij by The sampling process retains the key parameters with high semantic weight with a higher probability, and samples the relatively minor parameters with a lower probability, thus achieving semantic adaptation of parameter collection. According to simulation tests, when the total number of parameters M = 10,000, the sampling compression ratio can reach 10-100 times, that is, the final semantic feature dimension is only 1 / 10-1 / 100 of the original parameter, which greatly reduces the data transmission and storage overhead.

[0053] Step 5: For the production parameter vector x=[x 1 ;x 2 ;...;x M ] to perform adaptive semantic sampling. y = Φ x, where y is the compressed low-dimensional semantic feature vector. ij by The sampling process retains the key parameters with high semantic weight with a higher probability, and samples the relatively minor parameters with a lower probability, thus achieving semantic adaptation of parameter collection. According to simulation tests, when the total number of parameters M = 1000, the sampling compression ratio can reach 10 to 100 times, that is, the final semantic feature dimension is only 1 / 10 to 1 / 100 of the original parameter, which greatly reduces the data transmission and storage overhead.

[0054] The hybrid reinforcement learning fault diagnosis unit based on tensor decomposition in the quality and safety intelligent analysis module 2 adopts an innovative method to achieve high-precision fault diagnosis. The specific steps are as follows:

[0055] First, various production logs and quality reports are collected from information sources such as equipment management systems and MES, and structured information such as fault modes, fault symptoms, and affected components are extracted through text mining methods to construct a three-dimensional semantic tensor T∈R of fault-symptom-component. I×J×K , where I, J, and K represent the number of failure modes, the number of symptom modes, and the number of associated components, respectively.

[0056] Next, the system uses the innovative multimodal tensor decomposition (MTD) algorithm to decompose the semantic tensor T. Unlike the traditional Tucker decomposition, the MTD algorithm can simultaneously consider the multiple modal characteristics of the tensor data and better capture the complex associations between faults, symptoms, and components. The goal of the MTD algorithm is to find the fault factor matrix Symptom Factor Matrix Component Factor Matrix and core tensors So that T≈G× 1 U× 2 S× 3 V. Here× n represents the n-mode tensor product, R 1 , R 2 , R 3 are the ranks of the three dimensions, which are usually much smaller than the original dimensions I, J, and K, thus achieving dimensionality reduction and feature extraction of data. The core idea of ​​the MTD algorithm is to introduce a multimodal regularization term based on the traditional Tucker decomposition to make full use of the complementary information between different modes. The optimization objective function of the algorithm can be expressed as:

[0057]

[0058] where ||·|| F represents the Frobenius norm, Ω 1 (U, S, V) is the multimodal regularization term of the factor matrix, Ω 2 (G) is the sparse regularization term of the core tensor, λ 1 and λ2 is a trade-off parameter. By solving the optimization problem through the alternating least squares (ALS) method, the required factor matrix and core tensor can be obtained. The U, S, and V matrices obtained by the MTD algorithm respectively characterize the latent semantic topics of failure modes, symptoms, and components, and the core tensor G reflects the high-order interaction relationship between these topics. These decomposition results contain important prior diagnostic knowledge, which will be used to guide the subsequent training of neural networks.

[0059] In order to make full use of the prior knowledge obtained by tensor decomposition, the system designs an innovative tensor factor graph convolutional network (TF-GCN). Different from the traditional GCN, TF-GCN introduces the dynamic bidirectional attention propagation (DBAP) mechanism, which can more effectively capture the dynamic associations between different features. The state feature propagation rule of the DBAP mechanism is:

[0060]

[0061] Where H( k )∈R N × F A is the node feature matrix of the k-th layer network, F is the node feature dimension, and σ(·) is the nonlinear activation function (such as ReLU). i and are dynamically generated attention matrices that control the information flow in the row and column directions of the features respectively. Specifically:

[0062] A i =softmax(W 1 [U; V])

[0063]

[0064] Here W 1 and W 2 is a learnable parameter matrix, [·;·] represents a matrix concatenation operation, G (i) is the i-th slice of the core tensor G. The softmax function is used to normalize the attention scores. In this way, the network can dynamically adjust the importance of feature propagation according to the current state and prior knowledge, improving the expressiveness and flexibility of the model.

[0065] The output of the TF-GCN network is concatenated with the hidden state vector of the LSTM network and input into the fully connected layer to obtain the predicted probability vector of the fault mode. The entire hybrid neural network is trained end-to-end, and the cross entropy loss function and Adam optimizer are used for parameter optimization. In order to further improve the diagnostic performance, the system uses reinforcement learning to optimize the online diagnostic strategy of the network. In the specific implementation, the fault diagnosis task is modeled as a Markov decision process (MDP), the state space S is the historical fault symptom observation sequence, the action space A is the set of candidate fault causes, and the reward function R is the accuracy of predicting the fault cause. Under this MDP framework, the network's inference process can be formalized as finding an optimal diagnostic strategy π:S→A, so that taking diagnostic action a under the current fault symptom state s can maximize the expected future cumulative reward.

[0066] The system uses a deep Q-learning (DQN)-based approach to optimize the diagnostic strategy. The hybrid neural network is used as an approximation of the Q-value function, namely

[0067] Q(sa;θ)=H T (a)

[0068] Among them, H T (a) is the predicted probability of the TF-GCN network for the fault cause a in state s. During online diagnosis, the system integrates the prediction results of the TF-GCN and LSTM modules and selects the fault cause with the largest Q value as the diagnosis result. According to the feedback of the diagnosis result (accuracy) error), the network parameter θ is updated, and the strategy iteration formula is:

[0069]

[0070] where y = r + γ·max a′ Q(s′,a′;θ) is the TD target, r is the immediate reward, γ is the discount factor, and α is the learning rate. Through multiple iterations of training cycles, the network will gradually strengthen the learning of successful diagnosis strategies.

[0071] After experimental verification, the hybrid reinforcement learning diagnosis algorithm based on multimodal tensor decomposition and dynamic bidirectional attention propagation mechanism can achieve an average diagnostic accuracy of 97% when dealing with typical faults in powder metallurgy production, which is 2 percentage points higher than the traditional method. Especially when dealing with complex multi-source faults, the new algorithm shows stronger robustness and generalization ability. This is mainly due to the fact that multimodal tensor decomposition can more comprehensively extract multidimensional data features, while the dynamic bidirectional attention propagation mechanism enhances the model's ability to capture dynamic associations between different features. The system has a built-in diagnostic knowledge base covering common fault modes in powder metallurgy production, such as high porosity, out-of-tolerance size, uneven hardness, cracking, rough surface, etc. In practical applications, the system can flexibly expand fault modes according to production needs, and continuously optimize the diagnostic model through online learning to ensure that the diagnostic performance keeps pace with the times.

[0072] Another key innovation of the Quality and Safety Intelligent Analysis Module 2 is the adaptive quality management decision optimization algorithm. Traditional quality management mostly adopts qualitative experience decision making, which is difficult to cope with complex and changeable production situations. To address this pain point, this system proposes to use fractional order chaos theory and intelligent optimization algorithms to perform real-time dynamic optimization of quality management decisions in the production process.

[0073] First, the system extracts a series of state parameters reflecting product quality attributes (such as appearance, size, hardness, strength, etc.) from the production execution MES system to form an N-dimensional quality state vector Q = [q 1 ;q 2 ;...;q N ], and at the same time form an adjustable M-dimensional process parameter vector D = [d 1 ;d 2 ;...;d M ] as the management decision variable. Then, the fractional differential equation is used to describe the evolution law of quality state:

[0074]

[0075] Among them, α is the fractional order, is a Caputo-type fractional differential operator, which reflects the non-local memory effect and fractional-order dynamic characteristics of the quality state. f(·) is the nonlinear coupling function of the quality state vector Q and the decision vector D. With the help of the generalized derivative defined by fractional differentials, the model can characterize the long-range correlation and historical dependence of the quality state, which is more in line with the dynamic characteristics of the production process. Considering that it is difficult to obtain an analytical solution to the fractional differential equation, the system uses a numerical approximation method to discretize and solve it, and the time domain prediction model of the quality state can be obtained:

[0076]

[0077] Where h is the time step, ω k,t+1 is the weight coefficient defined by Grunwald-Letnikov (GL) of fractional differential. On this basis, the quality management optimization objective function is constructed:

[0078]

[0079] Where T is the optimization time domain of management decision, L(·) is the deviation of the measured quality state Q from the target Q * The loss function can be flexibly defined in the form of the sum of squared errors (SSE). At the same time, considering the physical constraints of process parameters, the boundary conditions of decision variables are added to solve the quality management decision as a constrained nonlinear programming problem.

[0080] In order to efficiently solve the above quality management optimization model and overcome the defect that traditional numerical optimization methods are prone to fall into local optimality, this system innovatively introduces a chaotic optimization algorithm based on Logistic mapping. By using the pseudo-randomness and ergodicity of chaotic motion, the management decision variables are encoded and mapped through chaotic variables:

[0081] d i =a i +z i ·(b i -a i ), i=1,2,...,M

[0082] Among them, a i and b i are the lower and upper bounds of the value of the ith decision variable, respectively, z i is a chaotic variable that satisfies the Logistic mapping:

[0083]

[0084] μ is a chaos parameter with a value between 3.56 and 4. During the iterative optimization process, the chaotic variable z dynamically traverses the search space in a pseudo-random manner and adaptively adjusts the search step size, so that the particles continuously jump out of the local optimum in the solution space and finally converge to the global optimal solution with a high probability.

[0085] The algorithm flow is as follows:

[0086] Step 1: Randomly initialize the chaotic variable z(0)∈(0,1), set the maximum number of iterations K, and the current number of iterations k=0;

[0087] Step 2: Generate an initial management decision variable D(k) based on the chaotic variable z(k) decoding;

[0088] Step 3: Substitute D(k) into the quality state prediction model and calculate the cumulative quality loss J(D(k)) in the next T steps;

[0089] Step 4: Update the current optimal solution D * =argmin{J(D(k)),J(D * )};

[0090] Step 5: Determine whether the convergence condition is met. If so, output D * If it is the optimal decision, terminate the iteration; otherwise, set k = k + 1 and go to Step 6;

[0091] Step 6: Generate new chaotic variables according to the Logistic mapping z(k+1)=4·z(k)·(1-z(k)), and go to Step 2.

[0092] Through simulation tests, when the maximum number of iterations K = 100, the chaos optimization algorithm can quickly converge to the optimal management decision, and the quality state prediction error after optimization can be reduced by more than 50% on average. * By decoding and combining the current production status, a quality control plan for the future period (such as the next day) can be generated, including the control values ​​of key process parameters, scheduling suggestions for personnel and equipment, etc., to form an operational, closed-loop quality management decision-making optimization plan.

[0093] The quality traceability management module 3 builds a quality traceability chain covering the entire life cycle of powder metallurgy products. The system adopts a consortium blockchain architecture, with manufacturers, suppliers, distributors, regulators, etc. acting as blockchain nodes to record and store quality data in all aspects of production, circulation, and sales. The system has designed a three-layer blockchain structure of raw materials-process-products. The raw materials blockchain records the procurement and inspection information of raw and auxiliary materials, the process blockchain records the production and processing process data, and the product blockchain records the inspection and sales information of finished products. The three sub-chains are linked in order through hash pointers. Once a quality problem occurs, the problem batch and flow direction can be quickly located. The blockchain network adopts the PBFT consensus algorithm, supports large-scale dynamic expansion of nodes, and the consensus delay can be controlled at the second level. At the same time, the system uses the elliptic curve cryptography ECC algorithm to encrypt and sign the block data to eliminate the risk of tampering. The consortium chain model takes into account both commercial privacy protection and orderly sharing of quality information, providing a trusted collaborative platform for all parties in production management.

[0094] The present invention further optimizes the performance of the system from the aspects of human-computer interaction and intelligent perception. The system adopts B / S architecture to design the human-computer interaction module 4, which supports cross-platform and mobile access; the introduction of WebGL and data mining algorithm library enables the system to have smooth three-dimensional visualization interaction and one-click statistical analysis capabilities. At the level of intelligent perception, the selection of sensors fully considers indicators such as accuracy, sampling rate, and stability. At the same time, the collaborative optimization framework of the system is open and scalable, and each functional module can be flexibly added or deleted according to the application scenario.

[0095] The following is a further explanation of the workflow of this system in conjunction with specific application cases. In a powder metallurgy gear production workshop, the system deployed 100 smart sensors on key equipment in processes such as pressing, sintering, post-processing, and finished product testing, with a collection frequency of 1kHz. These sensors are connected to five multi-protocol edge gateways via the Industrial Ethernet protocol. The gateway semantically classifies the data and uploads it to the MES server via the TSN private network. The quality and safety intelligent analysis module is deployed on the MES server.

[0096] (1) Semantic ontology-driven sampling: The system constructs a semantic ontology of the production process with 500 concepts including equipment, process, material, quality, and defect. Through ontology reasoning, the top 10 production parameters with semantic importance weight are extracted, mainly including pressing pressure, sintering temperature, holding time, quenching temperature, etc., accounting for 10% of the total parameters. The sampling frequency of each parameter is adjusted adaptively using the weight, and the average compression efficiency reaches 10:1.

[0097] (2) Hybrid reinforcement learning diagnosis: The system collected 1,000 fault records from the company within two years and found that the faults were mainly concentrated in three modes: out-of-tolerance size, low hardness, and cracking. The diagnostic knowledge tensor covering the three fault modes was obtained through tensor decomposition, and the TF-GCN fault diagnosis model was trained accordingly. In the online application stage, when a batch of gears had a low hardness defect, the hybrid diagnosis model issued a timely warning and traced back to find that the sintering temperature was 20°C lower than the original cause, with a diagnostic accuracy of 95%.

[0098] (3) Fractional-order chaos optimization: The system constructs a fractional-order prediction model for three key quality characteristics of gear strength, hardness, and size, with an order α of 0.6 and an average prediction error of less than 5%. When product sampling finds that the hardness is generally low, the system quickly starts the quality management parameter optimization. After 100 iterations, it locks in the optimal quenching temperature curve and time, and predicts that the hardness value will increase by more than 10% after the process plan is implemented.

[0099] (4) Blockchain quality traceability: The system deploys blockchain nodes in the procurement of raw materials, production and processing, warehousing and logistics, and post-sales processes. When a customer complains that a particular gear is cracked, the system can quickly locate the production batch through the product QR code, and then use blockchain to work backwards to find that the cause is that the carbon content of a batch of steel powder does not meet the standards, providing data support for the company's claims and quality improvements.

[0100] In summary, the quality and safety monitoring and management system for powder metallurgy production proposed in the present invention integrates cutting-edge artificial intelligence technologies such as semantic sampling, reinforcement learning, chaos optimization, and blockchain. It is specialized in design and engineering optimization based on the production characteristics of the powder metallurgy industry, forming a complete intelligent perception, analysis, traceability, and interactive technical solution. The application practice of the system shows that through adaptive sampling, the efficiency of production parameter collection has increased by 50%; through hybrid diagnosis, the accuracy of quality warning has reached 95%; through fractional optimization, the qualified rate of product inspection has increased by 5 percentage points. The successful development of this system provides a new idea for solving the quality control problem in the powder metallurgy industry, and is of great significance for realizing the intelligent and lean production of powder metallurgy.

[0101] It should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the principles of the present invention should be included in the protection scope of the present invention.

Claims

1. Quality and safety monitoring and management system for powder metallurgy production, characterized in that: It includes a production parameter intelligent perception module, a quality and safety intelligent analysis module, a quality traceability management module and a human-computer interaction module; the production parameter intelligent perception module collects production parameters and completes data preprocessing, and sends the results to the quality and safety intelligent analysis module; The quality and safety intelligent analysis module receives production parameters, obtains key quality features based on semantic ontology-driven adaptive sampling, implements fault diagnosis using a hybrid reinforcement learning method using tensor decomposition, and dynamically adjusts quality management strategies through fractional-order chaos optimization to generate quality warning information, diagnostic decisions, and management plans; the quality traceability management module uses blockchain technology to establish a quality traceability chain covering the entire life cycle of raw materials, processes, and products, receives quality data, and completes secure evidence storage; the human-computer interaction module implements a visual display of system functions and data, accepts user instructions, and coordinates other modules to complete quality monitoring and management tasks.

2. The system according to claim 1, characterized in that The production parameter intelligent perception module includes multi-source heterogeneous sensors, edge gateways and time-sensitive networks; wherein the multi-source heterogeneous sensors collect raw material properties, process parameters and equipment status signals, and transmit the data to the edge gateway in real time; the edge gateway cleans and normalizes the sensor data, and temporarily stores it after time and semantic classification; the time-sensitive network transmits the perception data of the edge gateway to the quality and safety intelligent analysis module with a delay of no more than 10ms.

3. The system according to claim 1, characterized in that The semantic ontology-driven adaptive sampling unit in the quality and safety intelligent analysis module obtains key quality features in the following ways: The semantic ontology of production process O=(C,R,A) which formally defines the semantic categories and associations of raw materials, equipment, process and quality, where C is the concept set, R is the relationship set and A is the axiom set; O is described using the ontology language OWL; Based on the Tableau algorithm, the semantic ontology O is inferred to obtain the semantic association matrix W∈RN×N, where the matrix element wij∈[0,1] represents the semantic association strength between concept i and concept j, and N is the number of semantic concepts in the ontology; Determine the semantic importance weight of each sensor acquisition parameter xi according to the association matrix W The calculation formula is Design a rotation matrix compression sampling operator Φ∈RM×M, where the matrix elements Φij obey the Bernoulli distribution sigmoid(x)=(1+e -x ) -1 is the Sigmoid function, λ is the scale factor; Adaptively sample the sensor parameter vector x to obtain the compressed sensing vector y = Φ x i y∈RM is a low-dimensional semantic feature vector.

4. The system according to claim 1, characterized in that The hybrid reinforcement learning fault diagnosis unit based on tensor decomposition in the quality and safety intelligent analysis module implements fault diagnosis through the following steps: The fault-symptom-component three-dimensional semantic cubic tensor T∈R constructed based on equipment operation logs, alarm information and quality defect data I×J×K , where I is the number of failure modes, J is the number of symptom modes, and K is the number of associated components; The multimodal tensor decomposition algorithm is used to decompose the semantic tensor T to obtain the fault factor matrix Symptom factor matrix S∈R J ×R2, component factor matrix V∈R K ×R3, and the core tensor Satisfies T≈G×1U×2S×3V, where × n represents the n-mode tensor product, R1, R2, and R3 are the ranks of the three dimensions respectively; Constructing fusion prior knowledge Θ K and data-driven knowledge Θ D Hybrid fault diagnosis strategy for network π(a|s; Θ), where the state s∈R J is the J-dimensional symptom observation vector, action a∈R I is the fault mode diagnosis probability vector, network parameter Θ=(Θ K ,Θ D ); Prior knowledge Θ K The tensor kernelized graph convolutional network is used for modeling, and the state feature propagation adopts a dynamic bidirectional attention propagation mechanism: Among them, H (k) ∈R N×F is the node feature matrix of the k-th layer network, F is the node feature dimension, σ(·) is the activation function, A i and is the dynamic attention matrix; Data-driven knowledge D The long short-term memory network is used to model the fault symptom observation sequence (s1, s2, ..., s T ) is encoded into a hidden state vector h T and the memory cell state vector c T ; The loss function of the hybrid diagnosis strategy network is: Where γ∈(0,1) is the discount factor, r t is the diagnostic accuracy reward, y t is the true state-action value, Q π (s t ,a t ) is the estimated state-action value.

5. The system according to claim 1, wherein: The quality management strategy dynamic adjustment unit based on fractional-order chaos optimization in the quality and safety intelligent analysis module realizes the search for the optimal management strategy through the following method: Construct an N-dimensional quality state vector Q = [q1; q2; ...; q N ], where q i ∈R is the i-th quality characteristic value; construct an M-dimensional management decision vector D = [d1; d2; ...; d M ], where d j ∈Ω is the jth adjustable management parameter, and the parameter space Ω is the parameter value range; The Caputo fractional differential equation is used to describe the fractional dynamic characteristics of the mass state: in dτ is the Caputo fractional differential operator, and [(-) is the Gamma function; Discretize the fractional-order differential equation to obtain the quality state prediction model: Q(t+1)=Q(t)+hα·∑(k=0:t)ω k ,t+1·f(Q(t-k),D(t-k));· The cumulative quality loss J in the next T steps T =∑(t=1:T)L(Q(t),Q * ) is the optimization target, and a quality management decision optimization model is constructed, where L(:) is the deviation of the measured quality state Q from the target Q * The loss function of A chaotic optimization algorithm based on Logistic mapping z(k+1)=μ·z(k)·(1-z(k)) is introduced to encode and map the management decision D through the chaotic variable z: d j =d minj +z·(d maxj -d minj ), and search for the optimal decision vector D in the parameter space Ω * ; By the optimal decision vector D * Decode and generate the quality management plan for the next T steps to guide the quality improvement of the production process.

6. The system according to claim 1, wherein: The quality traceability management module constructs a three-layer blockchain quality traceability network in the following ways: building raw material blockchain, process blockchain and product blockchain to cover the entire production life cycle; using hash pointer technology to link each layer of blockchain to achieve end-to-end quality traceability; using elliptic curve cryptography to asymmetric encrypt and digitally sign quality data to ensure data security; supporting dynamic expansion of no less than 100 nodes, and controlling consensus delay within 1 second.

7. The system according to claim 1, characterized in that The human-computer interaction module is designed based on B / S architecture and supports cross-platform Web access; a single service node supports no less than 1,000 concurrent Http requests; The WebGL graphics rendering engine is used to achieve 3D visualization of production parameters, quality characteristics, diagnostic results, and traceability information data, with a rendering frame rate of no less than 60FPS. The built-in data mining algorithm library supports correlation analysis, comparative analysis, and trend prediction data analysis functions to assist quality management decisions.

8. The system according to claim 2, characterized in that The multi-source heterogeneous sensor has the following parameters: the sensor sampling frequency is not less than 1kHz, meeting the needs of online quality monitoring; the sensor range covers the key parameter range of the production process, and the indication error is no more than 1%; the edge gateway adopts an industrial-grade embedded computing platform, configured with no less than 4-core CPU, 8GB memory and 128GB solid-state hard drive, and supports multi-protocol access and edge computing.

9. The system according to claim 4, characterized in that The fault diagnosis unit has active learning capabilities and realizes online updating of the fault knowledge base in the following ways: extracting new fault modes and symptom information based on expert feedback on diagnosis results; dynamically expanding the dimension of the semantic tensor T and adding new tensor elements; regularly recalculating tensor decomposition and updating prior diagnostic knowledge; fine-tuning the network parameters of the hybrid diagnostic strategy to continuously improve the accuracy and generalization of fault diagnosis.

10. The system according to claim 6, characterized in that The quality traceability module adopts the alliance chain model to organize the blockchain network and supports collaborative quality management through the following mechanisms: manufacturers, suppliers, distributors, and regulatory authorities jointly act as blockchain nodes to perform identity authentication and authorization; use smart contract technology to regularize production and transaction behaviors, and realize automatic chain-up and sharing of quality data; balance data sharing and security through identity-based privacy protection and fine-grained access control; provide a quality traceability interface with controllable permissions for third-party audits, and promote social collaboration in quality management.

Images