Method and system for evaluating reliability of communication network of industrial equipment

Abstract

The invention discloses a method and a system for evaluating the reliability of an industrial equipment communication network. The method comprises the following steps: S1, collecting multi-source time sequence data of the industrial equipment communication network; s2, preprocessing the multi-source time sequence data; s3, training by using an extreme learning machine to generate a network state evaluation model, and predicting a future network state; s4, optimizing hyper-parameters of the network state evaluation model by using a cat swarm optimization algorithm; s5, predicting the real-time data by the network state evaluation model to generate a reliability evaluation result; s6, dynamically adjusting communication network parameters of the industrial equipment according to an evaluation result; s7, continuously monitoring a network state, updating model prediction in real time and feeding back latest data; and S8, triggering an early warning mechanism and implementing an optimization and adjustment scheme in case of a fault or a performance bottleneck. According to the invention, through fusion of deep learning and the improved cat swarm optimization algorithm, real-time accurate monitoring and adaptive regulation and control of the industrial equipment communication network are realized, so that the network stability and the operation efficiency are greatly improved.

Description

Technical Field

[0001] The present invention relates to the field of communication technologies, and particularly to a method and system for evaluating the reliability of an industrial equipment communication network. Background Art

[0002] With the continuous development of industrial automation and intelligent manufacturing, the industrial equipment communication network has become a key infrastructure for supporting real-time data transmission, remote monitoring, and automated control of production lines. However, the communication network between devices in the industrial environment often faces a harsh working environment, complex network topologies, and variable network load conditions, resulting in increasingly prominent network reliability problems. Existing industrial equipment communication networks mainly rely on traditional network monitoring technologies and static evaluation methods, which evaluate the network status by collecting basic indicators such as network traffic, latency, and packet loss rate. However, these methods have many deficiencies. First, since traditional evaluation methods mainly use fixed thresholds and simple statistical analysis, they cannot comprehensively reflect the dynamic characteristics of industrial equipment communication networks in complex environments, making it difficult to detect potential abnormal conditions in the network in a timely manner. Second, most existing methods are evaluated based on manually set parameters and lack intelligence and adaptability. As a result, in the case of a sharp increase in network load, equipment failure, or environmental interference, the evaluation results often have a lag and cannot achieve real-time monitoring and early warning.

[0003] In addition, when traditional network evaluation technologies face multi-source time-series data, they often rely on simple statistical features such as means and variances, and it is difficult to capture the complex non-linear relationships and time-dependent changes hidden in industrial equipment communication networks. Especially in a high-speed dynamic industrial production environment, the prediction error is large, which affects the effect of network optimization and control. When traditional methods are used for fault diagnosis and prediction, they mostly adopt rule-based algorithms or simple regression models. These methods show low robustness and accuracy when dealing with sudden faults and complex abnormal conditions in industrial equipment communication networks. Since industrial equipment communication networks vary greatly in different production scenarios, the network status evaluation method should have high adaptability, while traditional methods often cannot meet this requirement.

[0004] In recent years, with the development of machine learning and swarm intelligence algorithms, more and more research has begun to focus on applying deep learning, extreme learning machine and other technologies to the field of industrial network status prediction, and combining swarm intelligence optimization algorithms to adjust the parameters of the model, thereby improving the accuracy of prediction and the adaptive ability of the system. However, although the existing network status assessment model based on extreme learning machine has the advantages of fast training speed and high computational efficiency, it relies on manual adjustment or simple random search methods in parameter selection, and it is difficult to achieve adaptive adjustment of the global optimal parameters. On the other hand, although traditional swarm intelligence optimization algorithms such as particle swarm optimization and genetic algorithms can optimize model parameters to a certain extent, their search process is prone to fall into local optimality, and when faced with high-dimensional complex parameter space, the convergence speed is slow and the stability is insufficient. Therefore, how to effectively integrate advanced deep learning technology and swarm intelligence algorithms to achieve accurate evaluation of the communication network status of industrial equipment, and realize real-time optimization and control in dynamic environments such as network load fluctuations and equipment failures, has become a technical problem that needs to be solved urgently.

[0005] In the prior art, some studies have attempted to combine extreme learning machines with traditional cat swarm optimization algorithms to achieve hyperparameter optimization of network state assessment models, but this method has two major defects. First, the traditional cat swarm optimization algorithm only simulates the exploration and development behavior of cats when looking for food, and lacks the simulation of the alert behavior of cat groups under environmental changes. As a result, when the network state fluctuates violently or the model converges lagging, the optimization process is prone to fall into local optimality and cannot give full play to the advantages of global search. Secondly, the traditional objective function design is relatively simple, only considering the prediction error but failing to take into account the parameter diversity and model adaptability at the same time, which makes it difficult to obtain the optimal parameter configuration in complex industrial environments, thus affecting the accuracy and stability of network reliability assessment.

[0006] Therefore, how to provide a method and system for evaluating the reliability of industrial equipment communication networks is an urgent problem that those skilled in the art need to solve. Summary of the invention

[0007] One purpose of the present invention is to propose an evaluation method and system for the reliability of industrial equipment communication networks. The present invention makes full use of extreme learning machines and improved cat swarm optimization algorithms, generates network status evaluation models by preprocessing and training multi-source time series data, and predicts the operating status of industrial equipment communication networks in combination with real-time data, and then dynamically adjusts network key parameters based on prediction errors, and describes in detail the entire process of realizing intelligent monitoring, prediction and adaptive optimization of industrial equipment communication networks. The present invention not only realizes efficient and accurate prediction of network status, but also has global jump search capabilities, can effectively avoid local optimal solutions, and enable the network to maintain high stability and robustness in complex industrial environments, thereby significantly improving the scientificity and reliability of network resource configuration.

[0008] An evaluation method for the reliability of an industrial equipment communication network according to an embodiment of the present invention includes the following steps:

[0009] S1. Collect multi-source time-series data of the industrial equipment communication network;

[0010] S2. Preprocess the multi-source time-series data and standardize the multi-source time-series data;

[0011] S3. Use an extreme learning machine to train the preprocessed multi-source time-series data to generate a network state evaluation model and predict the future state of the industrial equipment communication network;

[0012] S4. Use the cat swarm optimization algorithm to optimize the hyperparameters of the network state evaluation model, including the number of hidden layer nodes, activation function, and learning rate;

[0013] S5. Based on the optimized network state evaluation model, predict the real-time collected industrial equipment communication network data to generate a reliability evaluation result of the industrial equipment communication network;

[0014] S6. According to the reliability evaluation result, combined with the cat swarm optimization algorithm, dynamically adjust the parameters of the industrial equipment communication network;

[0015] S7. Continuously monitor the state of the industrial equipment communication network, feedback the latest data, use the optimized network state evaluation model for real-time update and prediction, and adjust the industrial equipment communication network configuration through the cat swarm optimization algorithm;

[0016] S8. When predicting industrial equipment communication network failures or performance bottlenecks, trigger corresponding optimization and adjustment schemes by setting a warning mechanism.

[0017] Optionally, the multi-source time-series data specifically includes network delay, packet loss rate, throughput, and data transmission path, which are used to train the extreme learning machine and generate a network state evaluation model.

[0018] Optionally, the parameters of the industrial equipment communication network specifically include bandwidth allocation, routing selection, and load balancing strategies, which are used to achieve dynamic management and optimization of industrial equipment communication network resources.

[0019] Optionally, S3 specifically includes:

[0020] S31. Input the preprocessed multi-source time-series data into the extreme learning machine, and define the input data matrix X = [x1, x2,..., x n T , where x i is the preprocessed multi-source time-series data at each moment t i ;​

[0021] S32. Initialize the number of hidden layer nodes and the learning rate of the extreme learning machine, and randomly initialize the weight matrix W from the input layer to the hidden layer and the hidden layer bias vector b;

[0022] S33. Use the activation function combination to perform a non-linear mapping on the input multi-source time series data to generate the hidden layer output matrix H:

[0023] H = [f1(WX + b)|f2(WX + b)] + λ·[f3(WX + b)] 2 ;

[0024] where f1(), f2() and f3() are activation functions, and λ is a regulation coefficient for regulating the weight of the quadratic activation term;

[0025] S34. Use the least squares optimization based on adaptive weighted regularization to optimize the difference between the hidden layer output matrix H and the target output matrix Y, and calculate the output layer weight β:

[0026] β = (H T H + α·I + γ·diag(H T H)) -1 H T Y;

[0027] where α is the regularization coefficient, I is the identity matrix, γ is the weighted regularization parameter, and diag(H T H) is the diagonal matrix of the hidden layer output matrix H T H;

[0028] S35. Generate a network state evaluation model through the trained extreme learning machine. The network state evaluation model can evaluate the current and future states of the industrial equipment communication network based on the input preprocessed multi-source time series data, and predict the performance indicators of the industrial equipment communication network;

[0029] S36. Use the network state evaluation model to predict the newly collected and preprocessed time series data in real time to generate the future state prediction value Y of the industrial equipment communication network pred .

[0030] Optionally, the specific steps of S4 include:

[0031] S41. Initialize the population size of the cat swarm optimization algorithm to N, and set the position of each cat as the hyperparameter vector P i = [p i1 , p i2 ,..., p im , where p ij represents the j-th hyperparameter of the i-th cat, including the number of hidden layer nodes, activation function, and learning rate;

[0032] S42. Define the objective function f(P i ), which is used to evaluate the performance of the given hyperparameter vector P i :

[0033]

[0034] where Y pred is the predicted value of the future state of the industrial equipment communication network, Y is the actual target output, n is the number of data points, m is the number of hyperparameters, p ij is the j-th hyperparameter of the i-th cat, p ij-1 is the value of the (j - 1)-th hyperparameter of the i-th cat, and δ is the weight of the adjustment term;

[0035] S43. Simulate that individual cats in the cat swarm search according to the evaluation value of the objective function f(P i ), and perform global and local searches through position update and speed update rules:

[0036]

[0037] where is the updated position of the cat, G best is the position of the global optimal solution, P best is the position of the optimal solution of the individual cat, β1 and β2 are acceleration constants, α1 is the jump factor, and P random is a randomly generated position;

[0038] S44. Update the hyperparameters through the updated positions of the cat swarm, substitute the hyperparameter values corresponding to the new positions into the network state evaluation model, calculate the new network prediction result Y curr , and update the evaluation function value f(P i ):

[0039] Y curr = Hβ + ∈·(|Y curr - Y pred |)·sign(Y curr - Y pred );

[0040] where ∈ is the adjustment parameter, sign() is the sign function, H is the hidden layer output matrix, and β is the output layer weight;

[0041] S45. According to the evaluation function value f(P i ), determine whether the cat swarm has converged. If it has converged, output the current hyperparameter combination as the hyperparameters of the final network state evaluation model;

[0042] S46. Introduce an alert behavior. When the fitness change of the cat swarm is small, or no better solution is found for several consecutive generations, the cat swarm automatically triggers the alert behavior, increases the proportion of global exploration, reduces local optimization search, and promotes jumping to the areas in the solution space that have not been fully explored for optimization;

[0043] S47. Through the alert behavior, increase the proportion of the global search of the cat swarm, and adaptively adjust to control the balance between global exploration and local optimization, dynamically adjusting the proportion of exploration behavior and exploitation behavior;

[0044] S48. After several iterations, after the cat swarm optimization algorithm completes local optimization and global search, the optimal hyperparameter combination P best is obtained, and the network state evaluation model is updated to complete the optimization of the network state evaluation model;

[0045] S49. Perform real-time performance evaluation of the industrial equipment communication network with the optimized network state evaluation model, and dynamically adjust the key parameters of the industrial equipment communication network.

[0046] Optionally, the S5 specifically includes:

[0047] S51. Obtain real-time data from the industrial equipment communication network and define it as X new ;

[0048] S52. Input the real-time data X new into the optimized network state evaluation model, process the real-time data, and generate a hidden layer output;

[0049] S53. Use the optimized output layer weights to combine the hidden layer output with the weights to obtain the real-time prediction value of the industrial equipment communication network state The real-time prediction value represents the current performance evaluation of the industrial equipment communication network;

[0050] S54. Based on the real-time prediction value, extract the reliability evaluation index of the industrial equipment communication network and evaluate the operation status of the industrial equipment communication network;

[0051] S55. Evaluate the prediction error of the industrial equipment communication network according to the deviation between the real-time prediction result and the actual monitoring value, and generate the reliability evaluation result of the industrial equipment communication network to identify potential problem areas;

[0052] S56. According to the reliability evaluation result of the industrial equipment communication network, continuously monitor the industrial equipment communication network, trigger the optimization strategy, and adjust the key parameters of bandwidth allocation, routing selection, and load balancing in combination with the actual performance of the industrial equipment communication network.

[0053] Optionally, the reliability evaluation indicators of the industrial equipment communication network specifically include network stability, throughput, and latency, which are used to reflect the performance level and operation reliability of the industrial equipment communication network under different operating conditions.

[0054] Optionally, S6 specifically includes:

[0055] S61. Obtain the reliability evaluation result of the industrial equipment communication network and determine the reliability error value of the current industrial equipment communication network;

[0056] S62. Compare the reliability error with the preset threshold θ. When the reliability error exceeds the threshold, trigger the industrial equipment communication network parameter optimization process;

[0057] S63. Construct the parameter vector Q of the industrial equipment communication network to be optimized;

[0058] S64. Initialize the cat swarm optimization algorithm, set the position of each cat as the network parameter vector Q i , and set the cat swarm size, maximum number of iterations, and behavior adjustment factor;

[0059] S65. For each cat, evaluate the fitness of the parameter vector according to the objective function G(Q i ). The objective function simultaneously considers the prediction error of the industrial equipment communication network and the diversity penalty term of the parameter configuration:

[0060]

[0061] Among them, represents the predicted value of the kth data point under the parameter combination Q i , Y (k) is the true value of the kth data point, n is the number of data points, q ij is the value of the ith cat on the jth parameter, is the average value of the jth parameter in the current population, μ is a small constant to avoid division by zero, and λ o is a hyperparameter for adjusting the weight of the diversity penalty term;

[0062] S66. Use the cat swarm optimization algorithm to optimize the parameter vector Q. By introducing an adaptive behavior weight coefficient and a Gaussian perturbation jump mechanism, adjust the position of each cat in each iteration:

[0063]

[0064] Among them, Q i is the parameter vector of the current ith cat, P best is the optimal parameter vector of this cat in history, and G bestis the global optimal parameter vector in the entire cat population, r1 and r2 are independently generated random numbers, ω is the exploration and exploitation balance coefficient, η is the jump intensity coefficient, and ε is a random perturbation term with a mean of 0. is the position vector of the i-th cat after update;

[0065] S67. In each iteration, update the individual fitness according to the objective function G(Q i ) and output the global optimal parameter combination Q after reaching the maximum number of iterations or the convergence condition. opt ;

[0066] S68. Apply the optimal parameter vector Q opt to the industrial equipment communication network, configure and update the bandwidth allocation, routing selection, and load balancing strategies, and continuously monitor the state of the adjusted industrial equipment communication network, and feedback new data for optimization iteration.

[0067] Optionally, the parameter vector of the industrial equipment communication network to be optimized specifically includes the bandwidth allocation ratio, routing priority weight, and load balancing strategy factor, which are used to dynamically regulate network resource allocation and transmission paths.

[0068] An evaluation system for the reliability of an industrial equipment communication network according to an embodiment of the present invention includes the following modules:

[0069] A data acquisition module for real-time acquisition of data of the industrial equipment communication network;

[0070] A data preprocessing module for denoising, missing value filling, and standardization processing of the acquired data;

[0071] A network state evaluation model module for constructing and training a network state evaluation model using extreme learning machine to predict the network state;

[0072] A cat swarm optimization module for dynamically optimizing the hyperparameters of the network state evaluation model and the key network parameters;

[0073] A reliability evaluation module for evaluating the prediction error and overall reliability of the industrial equipment communication network according to the prediction results;

[0074] A parameter optimization control module for real-time adjusting the bandwidth allocation, routing priority, and load balancing strategy according to the evaluation results;

[0075] A monitoring and warning module for continuously monitoring the network state and triggering a warning mechanism when an anomaly is detected.

[0076] The beneficial effects of the present invention are:

[0077] By combining the extreme learning machine with the improved cat swarm optimization algorithm, the present invention realizes the efficient and accurate evaluation and dynamic optimization of the state of the industrial equipment communication network. The present invention adopts advanced algorithm technologies in all links of data acquisition, preprocessing, model training, real-time prediction, and parameter dynamic adjustment, enabling the network state evaluation model to fully capture the non-linear characteristics in multi-dimensional time series data such as network load, latency, and packet loss rate in the industrial environment. By introducing an adaptive weight and a Gaussian perturbation jump mechanism, it ensures that the optimization process has stronger global search ability under complex working conditions and avoids falling into local optima. Thus, the present invention can output the network reliability evaluation result in real time, accurately reflect the actual operation state of the network, and timely adjust the bandwidth allocation, routing selection, and load balancing strategies according to the prediction error, thereby realizing the intelligent adaptive control of the industrial equipment communication network. This system not only greatly improves the accuracy and real-time performance of network state prediction, but also shows extremely high robustness and stability in industrial field applications, effectively reducing the network failure risk, improving the overall operation efficiency of the system, and fully meeting the requirements for network reliability and dynamic response ability in high-demand industrial production environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0078] The accompanying drawings are used to provide a further understanding of the present invention, and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention, and do not constitute a limitation to the present invention. In the drawings:

[0079] Figure 1 is a flowchart of an evaluation method for the reliability of an industrial equipment communication network proposed by the present invention;

[0080] Figure 2 is a schematic structural diagram of an evaluation system for the reliability of an industrial equipment communication network proposed by the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0081] Now, the present invention will be further described in detail with reference to the accompanying drawings. These drawings are all simplified schematic diagrams, only showing the basic structure of the present invention in a schematic manner, so they only show the components related to the present invention.

[0082] Refer to Figure 1 , an evaluation method for the reliability of an industrial equipment communication network, includes the following steps:

[0083] S1. Collect multi-source time series data of the industrial equipment communication network;

[0084] S2. Preprocess the multi-source time series data, and perform standardization processing on the multi-source time series data;

[0085] S3. Use the extreme learning machine to train the preprocessed multi-source time series data to generate a network state evaluation model, and predict the future state of the industrial equipment communication network;

[0086] S4. Optimize the hyperparameters of the network state evaluation model using the cat swarm optimization algorithm, including the number of hidden layer nodes, activation function, and learning rate;

[0087] S5. Predict the industrial equipment communication network data collected in real time based on the optimized network state evaluation model, and generate a reliability evaluation result of the industrial equipment communication network;

[0088] S6. According to the reliability evaluation result, combine with the cat swarm optimization algorithm to dynamically adjust the parameters of the industrial equipment communication network;

[0089] S7. Continuously monitor the state of the industrial equipment communication network, feedback the latest data, use the optimized network state evaluation model for real-time update and prediction, and adjust the configuration of the industrial equipment communication network through the cat swarm optimization algorithm;

[0090] S8. When predicting industrial equipment communication network failures or performance bottlenecks, trigger corresponding optimization and adjustment schemes by setting an early warning mechanism.

[0091] For the industrial equipment communication network reliability evaluation method and system of the present invention, by using an extreme learning machine to train the network state evaluation model and combining with an improved cat swarm optimization algorithm to dynamically optimize the model hyperparameters (including the number of hidden layer nodes, activation function, learning rate) and network key parameters, real-time acquisition, preprocessing, prediction, and feedback closed-loop control of multi-source time series data are achieved. This system can accurately predict the future state of the industrial equipment communication network, and timely adjust bandwidth allocation, routing selection, and load balancing strategies according to real-time monitoring results, effectively reducing network latency and packet loss rate, and improving data throughput. Through the global jump search and adaptive parameter adjustment mechanism, the system has strong robustness and anti-interference ability in complex industrial environments, significantly shortening the fault response time, reducing the equipment downtime rate, ensuring the stability and efficiency of network operation, and thus providing reliable technical support for industrial automation and intelligent manufacturing.

[0092] In this embodiment, the multi-source time series data specifically includes network latency, packet loss rate, throughput, and data transmission path, which are used to train the extreme learning machine and generate a network state evaluation model.

[0093] In this embodiment, the parameters of the industrial equipment communication network specifically include bandwidth allocation, routing selection, and load balancing strategies, which are used to realize the dynamic management and optimization of industrial equipment communication network resources.

[0094] In this embodiment, the specific content of S3 includes:

[0095] S31. Input the preprocessed multi-source time-series data into the extreme learning machine, and define the input data matrix X = [x1, x2,..., x n T , where x i is the preprocessed multi-source time-series data at each moment t i .

[0096] S32. Initialize the number of hidden layer nodes and the learning rate of the extreme learning machine, and randomly initialize the weight matrix W from the input layer to the hidden layer and the hidden layer bias vector b;

[0097] S33. Use the activation function combination to perform a non-linear mapping on the input multi-source time-series data to generate the hidden layer output matrix H:

[0098] H = [f1(WX + b)|f2(WX + b)] + λ·[f3(WX + b)] 2 ;

[0099] where f1(), f2() and f3() are activation functions, and λ is a regulation coefficient for regulating the weight of the quadratic activation term;

[0100] S34. Use the least squares method optimization based on adaptive weighted regularization to optimize the difference between the hidden layer output matrix H and the target output matrix Y, and calculate the output layer weight β:

[0101] β = (H T H + α·I + γ·diag(H T H)) -1 H T Y;

[0102] where α is the regularization coefficient, I is the identity matrix, γ is the weighted regularization parameter, and diag(H T H) is the diagonal matrix of the hidden layer output matrix H T H;

[0103] S35. Generate a network state evaluation model through the trained extreme learning machine. The network state evaluation model can evaluate the current and future states of the industrial equipment communication network based on the input preprocessed multi-source time-series data, and predict the performance indicators of the industrial equipment communication network;

[0104] S36. Use the network state evaluation model to predict the newly collected and preprocessed time-series data in real time, and generate the future state prediction value Y pred of the industrial equipment communication network.

[0105] ​The present invention trains the preprocessed multi-source time-series data by using an extreme learning machine to generate a network state evaluation model, and optimizes the calculation of the output layer weights by using the least squares method based on adaptive weighted regularization, thereby realizing the efficient prediction of the state of the industrial equipment communication network. When the system is initialized, the number of hidden layer nodes and the learning rate are optimized, and at the same time, the input data is non-linearly mapped by using the activation function combination, ensuring the adaptability of the model to the changing data in the complex network environment. This method can accurately evaluate the current state and future trend of the network, thereby providing a reliable basis for real-time monitoring and fault warning. Compared with the traditional evaluation method, the present invention greatly improves the prediction accuracy and response speed, effectively reduces the network delay and data packet loss rate, and can dynamically predict the network performance indicators, providing accurate support for network scheduling and resource optimization, and significantly improving the operation efficiency and security of the industrial equipment communication network.

[0106] In this embodiment, step S4 specifically includes:

[0107] S41. Initialize the population size of the cat swarm optimization algorithm to N, and set the position of each cat as the hyperparameter vector P i =[p i1 ,p i2 ,...,p im , where p ij represents the j-th hyperparameter of the i-th cat, including the number of hidden layer nodes, the activation function, and the learning rate;

[0108] S42. Define the objective function f(P i ) for evaluating the performance of the given hyperparameter vector P i :

[0109]

[0110] where Y pred is the predicted value of the future state of the industrial equipment communication network, Y is the actual target output, n is the number of data points, m is the number of hyperparameters, p ij is the j-th hyperparameter of the i-th cat, p ij-1 is the value of the (j - 1)-th hyperparameter of the i-th cat, and δ is the weight of the adjustment term;

[0111] S43. Simulate that the individual cats in the cat swarm search according to the evaluation value of the objective function f(P i ), and perform global and local searches through the position update and speed update rules:

[0112]

[0113] where is the updated position of the cat, G best is the position of the global optimal solution, Pbest is the optimal solution position of an individual cat, β1 and β2 are acceleration constants, α1 is a jump factor, and P random is a randomly generated position;

[0114] S44. Update the hyperparameters based on the updated cat population positions, substitute the hyperparameter values corresponding to the new positions into the network state evaluation model, and calculate the new network prediction result Y curr , and update the evaluation function value f(P i ):

[0115] Y curr = Hβ + ∈·|Y curr - Y pred |)·sign(Y curr - Y pred );

[0116] where ∈ is a tuning parameter, sign() is the sign function, H is the hidden layer output matrix, and β is the output layer weight;

[0117] S45. According to the evaluation function value f(P i ), determine whether the cat population has converged. If it has converged, output the current hyperparameter combination as the final hyperparameters of the network state evaluation model;

[0118] S46. Introduce an alert behavior. When the fitness change of the cat population is small, or no better solution has been found for several consecutive generations, the cat population automatically triggers the alert behavior, increases the proportion of global exploration, reduces local optimization search, and promotes jumping to the areas in the solution space that have not been fully explored for optimization;

[0119] S47. Through the alert behavior, increase the proportion of global search of the cat population, and dynamically adjust the balance between global exploration and local optimization by adaptive adjustment, and dynamically adjust the proportion of exploration behavior and exploitation behavior;

[0120] S48. After several iterations, the cat swarm optimization algorithm obtains the optimal hyperparameter combination P best after completing local optimization and global search, updates the network state evaluation model, and completes the optimization of the network state evaluation model;

[0121] S49. Evaluate the real-time performance of the industrial equipment communication network using the optimized network state evaluation model, and dynamically adjust the key parameters of the industrial equipment communication network.

[0122] By introducing the vigilant behavior in the improved cat swarm optimization algorithm, the present invention effectively overcomes the problems of traditional optimization methods being prone to falling into local optima and having low search efficiency in complex industrial environments. In the present invention, when the cat swarm is searching for the hyperparameters (including the number of hidden layer nodes, activation function, learning rate) of the network state evaluation model, if it is found that the fitness change is extremely slow or there is no obvious improvement for several consecutive generations, the system automatically triggers the vigilant behavior, actively increases the proportion of global search and introduces random perturbations, enabling the algorithm to jump from the current local optimal solution to the region in the solution space that has not been fully explored, thereby achieving a more comprehensive parameter search. This mechanism not only improves the global search ability of the model, ensures that the hyperparameters of the network state evaluation model reach the optimal configuration, but also further enhances the accuracy and robustness of the model's prediction of the future state of the industrial equipment communication network. By adjusting the key network parameters in real time, the system can respond in a timely manner to network load fluctuations and abnormal situations, effectively reduce latency and packet loss rate, improve data throughput, provide reliable technical support for the stable operation of the industrial field communication network, and is significantly superior to the existing technology.

[0123] In this embodiment, step S5 specifically includes:

[0124] S51. Obtain real-time data from the industrial equipment communication network and define it as X new ;

[0125] S52. Input the real-time data X new into the optimized network state evaluation model, process the real time, and generate the hidden layer output;

[0126] S53. Combine the hidden layer output with the weights using the optimized output layer weights to obtain the real-time prediction value of the industrial equipment communication network state The real-time prediction value represents the current performance evaluation of the industrial equipment communication network;

[0127] S54. Based on the real-time prediction value, extract the reliability evaluation index of the industrial equipment communication network and evaluate the operation status of the industrial equipment communication network;

[0128] S55. Evaluate the prediction error of the industrial equipment communication network according to the deviation between the real-time prediction result and the actual monitoring value, and generate the reliability evaluation result of the industrial equipment communication network to identify potential problem areas;

[0129] S56. According to the reliability evaluation result of the industrial equipment communication network, continuously monitor the industrial equipment communication network, trigger the optimization strategy, and adjust the key parameters of bandwidth allocation, routing selection, and load balancing in combination with the actual performance of the industrial equipment communication network.

[0130] By using the optimized network state evaluation model to process the data in the industrial equipment communication network in real time, the present invention can quickly obtain an accurate prediction of the current network state, and extract the network reliability evaluation index from the real-time prediction value, so as to effectively monitor the network operation status. During the system operation, by evaluating the deviation between the real-time prediction result and the actual monitoring value, the abnormal and potential problem areas in the network can be identified in time, providing a scientific basis for network maintenance and fault warning. Based on this evaluation result, the system can automatically trigger the optimization strategy and dynamically adjust the key parameters such as bandwidth allocation, routing selection and load balancing, so that the network can adapt to sudden load changes and abnormal interferences, ensuring the stable and efficient operation of the communication network. This method not only greatly improves the real-time monitoring accuracy and response speed of the industrial equipment communication network, but also realizes the adaptive optimization of the network state through closed-loop feedback, effectively reducing the network delay and data packet loss risk, and further ensuring the continuity and security of industrial automation production. Through actual application tests, the system performs excellently in continuous monitoring and dynamic regulation, with significantly reduced prediction errors, greatly improving the overall operation efficiency and reliability of the network.

[0131] In this embodiment, the reliability evaluation indexes of the industrial equipment communication network specifically include network stability, throughput and delay, which are used to reflect the performance level and operation reliability of the industrial equipment communication network under different operation states.

[0132] In this embodiment, S6 specifically includes:

[0133] S61. Obtain the reliability evaluation result of the industrial equipment communication network, and determine the reliability error value of the current industrial equipment communication network;

[0134] S62. Compare the reliability error with the preset threshold θ. When the reliability error exceeds the threshold, trigger the optimization process of the industrial equipment communication network parameters;

[0135] S63. Construct the parameter vector Q of the industrial equipment communication network to be optimized;

[0136] S64. Initialize the cat swarm optimization algorithm, set the position of each cat as the network parameter vector Q i , and set the cat swarm size, the maximum number of iterations and the behavior adjustment factor;

[0137] S65. For each cat, evaluate the fitness of the parameter vector according to the objective function G(Q i ), and the objective function takes into account both the prediction error of the industrial equipment communication network and the diversity penalty term of the parameter configuration:

[0138]

[0139] Wherein, Denote the predicted value of the k-th data point under the parameter combination Q i as Y, (k) where the true value of the k-th data point is Y, n is the number of data points, and q ij is the value of the i-th cat on the j-th parameter, is the average value of the j-th parameter in the current population, μ is a small constant to avoid division by zero, and λ o is a hyperparameter for adjusting the weight of the diversity penalty term;

[0140] S66. Optimize the parameter vector Q using the cat swarm optimization algorithm. By introducing an adaptive behavior weight coefficient and a Gaussian perturbation jump mechanism, adjust the position of each cat in each iteration:

[0141]

[0142] where Q i is the parameter vector of the current i-th cat, P best is the optimal parameter vector of this cat in history, G best is the global optimal parameter vector in the entire cat swarm, r1 and r2 are independently generated random numbers, ω is the exploration and exploitation balance coefficient, η is the jump intensity coefficient, and ε is a random perturbation term with a mean of 0, is the updated position vector of the i-th cat;

[0143] S67. In each iteration, update the individual fitness according to the objective function G(Q i ), and after reaching the maximum number of iterations or the convergence condition, output the global optimal parameter combination Q opt ;

[0144] S68. Apply the optimal parameter vector Q opt to the industrial equipment communication network, configure and update the bandwidth allocation, routing selection, and load balancing strategies, and monitor the state of the adjusted industrial equipment communication network in real time, and feedback new data for optimization iteration.

[0145] The present invention realizes real-time monitoring and closed-loop feedback regulation of network reliability errors by introducing an improved cat swarm optimization algorithm to dynamically optimize the communication network parameters of industrial equipment. The system first obtains various operation data of the industrial equipment communication network through a real-time acquisition module, and completes denoising, complementing, and standardizing processing in the data preprocessing module. Then, the network state evaluation model processes the preprocessed data to generate real-time prediction values, thereby evaluating the network reliability error. Based on this, when the reliability error exceeds the preset threshold, the system automatically triggers the parameter optimization process, constructs the network parameter vector to be optimized, and performs global search and local adjustment on key parameters such as bandwidth allocation, routing selection, and load balancing strategies. The improved cat swarm optimization algorithm introduces an adaptive behavior weight coefficient and a Gaussian perturbation jump mechanism in each iteration, effectively avoiding the problem of traditional optimization methods falling into local optimal solutions, so that the parameter search is more comprehensive and the update is more accurate. After several iterations, the system outputs the global optimal parameter combination and applies it to the actual network configuration to realize the automatic regulation of the industrial equipment communication network. Thus, the present invention not only significantly reduces network latency and data packet loss rate, improves throughput, but also can quickly respond to network anomalies, shorten the fault handling time, ensure the stability and efficiency of network operation, and provide reliable communication guarantee and safe operation support for the industrial production site.

[0146] In this embodiment, the parameter vector of the industrial equipment communication network to be optimized specifically includes the bandwidth allocation ratio, the routing priority weight, and the load balancing strategy factor, which are used to dynamically regulate the network resource allocation and transmission path.

[0147] Reference Figure 2 , an evaluation system for the reliability of an industrial equipment communication network, includes the following modules:

[0148] A data acquisition module for real-time acquisition of data of the industrial equipment communication network;

[0149] A data preprocessing module for denoising, missing value filling, and standardizing the acquired data;

[0150] A network state evaluation model module for constructing and training a network state evaluation model using extreme learning machine to predict the network state;

[0151] A cat swarm optimization module for dynamically optimizing the hyperparameters of the network state evaluation model and the key network parameters;

[0152] A reliability evaluation module for evaluating the prediction error and overall reliability of the industrial equipment communication network according to the prediction results;

[0153] A parameter optimization control module for real-time adjusting the bandwidth allocation, routing priority, and load balancing strategy according to the evaluation results;

[0154] The monitoring and early warning module is used to continuously monitor the network status and trigger the early warning mechanism when anomalies are detected.

[0155] Example 1:

[0156] To verify the feasibility of the present invention in implementation, the present invention is applied to a large-scale iron and steel production plant. The industrial equipment communication network in this plant area undertakes tasks such as real-time data transmission, equipment monitoring, and remote control. Due to the large scale of the plant area, the variety of equipment, and the complex operating environment, traditional network monitoring systems often have inaccurate evaluations and slow responses when dealing with problems such as network latency, data packet loss, and bandwidth bottlenecks, resulting in an increased risk of equipment failures and production interruptions. In view of the above problems, the present invention proposes an evaluation method and system for the reliability of an industrial equipment communication network. This system makes full use of the extreme learning machine and the improved cat swarm optimization algorithm, and through preprocessing, model training, real-time prediction, and dynamic parameter optimization of multi-source time-series data, realizes the accurate evaluation and adaptive regulation of the network status.

[0157] In practical applications, distributed sensors and network monitoring devices are installed on each production line in the plant area. These devices can collect key operation data such as network latency, packet loss rate, data throughput, and connection status in real time and upload the data to the central monitoring platform. After the data preprocessing module performs denoising, missing value filling, and standardization processing on the collected data, it is sent to the network status evaluation model module. This module uses the extreme learning machine to quickly train the preprocessed data to construct a network status evaluation model, and then adaptively optimizes the model hyperparameters through the improved cat swarm optimization algorithm to further improve the accuracy and robustness of model prediction. After the model is optimized, the system automatically updates the real-time prediction data every 5 minutes and outputs the reliability evaluation result of the industrial equipment communication network.

[0158] During an actual operation in the plant area, through continuous monitoring and automatic optimization of the system, the system can timely identify the situation where the network latency and packet loss rate increase abnormally. Between 2 am and 10 am on a certain day, the traditional network monitoring system showed that the network latency in this area could reach up to 120 milliseconds at most, and the packet loss rate was as high as over 7%, resulting in frequent interruptions in production line data transmission. However, through the rapid prediction and feedback of real-time data, the system of the present invention controls the prediction error within 5%, and under the drive of the monitoring module, automatically adjusts the bandwidth allocation, routing priority, and load balancing strategy, so that the network latency in this area is stabilized below 60 milliseconds, the packet loss rate is reduced to about 2%, and the overall data throughput is increased by about 20%. Therefore, the operation and maintenance personnel can quickly take measures after receiving the early warning to repair the faulty nodes, avoiding long-term downtime and production losses.

[0159] Table 1 Network Status Evaluation Data Table

[0160]

[0161] According to the data in Table 1, it can be clearly seen that the system of the present invention is superior to the traditional system in multiple key performance indicators. Especially in terms of communication latency, packet loss rate, and network throughput, it shows obvious advantages. In the data at 2:00 am on November 15, 2024, the average communication latency of the traditional system was 110 milliseconds, while the system of the present invention controlled the latency within 65 milliseconds, a reduction of 45 milliseconds, and the latency improvement rate was approximately 41%. At the same time point, the packet loss rate also significantly decreased from 6.5% to 1.8%, indicating that when the network is unstable, the system can quickly identify problems and adjust network parameters through an optimization mechanism, thus effectively reducing data loss during transmission. In terms of network throughput, the traditional system was 150 Mbps, while the present system reached 180 Mbps, an increase of 20%, fully demonstrating its optimization ability in bandwidth resource allocation and transmission efficiency.

[0162] The data between 4:00 am and 10:00 am also shows the continuous and stable advantages of the system. Especially in the 4:00 period, the latency of the traditional system was as high as 120 milliseconds, approaching the network bottleneck, while the present system controlled it within 60 milliseconds, and the packet loss rate also decreased from 7.0% to 2.0%. The network performance after system optimization is more stable. The data at 6:00, 8:00, and 10:00 further verify the robustness of the system of the present invention under continuous operation conditions. As the network traffic gradually increases, the latency and packet loss rate of the traditional system begin to fluctuate, while the present system maintains a relatively stable performance output.

[0163] Generally speaking, within the entire test time period, the communication latency of the system of the present invention was reduced by more than 40% on average, the packet loss rate was reduced by approximately 70%, and the throughput was increased between 15% and 20%. This fully shows that the network state evaluation and regulation mechanism proposed by the present invention, which combines extreme learning machine and improved cat swarm optimization algorithm, can effectively improve the overall performance of industrial equipment communication networks and enhance their adaptability to network fluctuations in complex industrial environments. The system can not only predict potential network performance decline trends in advance but also take timely parameter optimization measures to deal with them, truly achieving the closed-loop control goal of "prediction + feedback + self-regulation". These empirical data also confirm the feasibility and promotion value of the present invention in industrial actual scenarios.

[0164] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention, according to the technical solution and inventive concept of the present invention, makes equivalent substitutions or changes, and all should be covered within the protection scope of the present invention.

Claims

1. A method and system for evaluating the reliability of an industrial equipment communication network, characterized in that: The steps include: S1, collect multi-source time series data of industrial equipment communication network; S2, preprocessing the multi-source time series data, and standardizing the multi-source time series data; S3, using extreme learning machine to train the preprocessed multi-source time series data to generate a network status assessment model and predict the future status of the industrial equipment communication network; S4, using a cat swarm optimization algorithm to optimize the hyperparameters of the network state assessment model, including the number of hidden layer nodes, activation function, and learning rate; S5. Based on the optimized network status assessment model, predict the real-time collected industrial equipment communication network data to generate a reliability assessment result of the industrial equipment communication network; S6. According to the reliability evaluation results, combined with the cat swarm optimization algorithm, dynamically adjust the parameters of the industrial equipment communication network; S7, continuously monitor the status of the industrial equipment communication network, feedback the latest data, use the optimized network status assessment model for real-time updates and predictions, and adjust the industrial equipment communication network configuration through the cat swarm optimization algorithm; S8. When a communication network failure or performance bottleneck of industrial equipment is predicted, an early warning mechanism is set up to trigger corresponding optimization and adjustment plans.

2. The method for evaluating the reliability of an industrial equipment communication network according to claim 1, characterized in that: The multi-source time series data specifically includes network delay, packet loss rate, throughput, and data transmission path, which are used to train the extreme learning machine and generate a network status evaluation model.

3. The method for evaluating the reliability of an industrial equipment communication network according to claim 1, characterized in that: The parameters of the industrial equipment communication network specifically include bandwidth allocation, routing selection and load balancing strategy, which are used to achieve dynamic management and optimization of industrial equipment communication network resources.

4. The method for evaluating the reliability of an industrial equipment communication network according to claim 1, characterized in that: The S3 specifically includes: S31, input the preprocessed multi-source time series data into the extreme learning machine, and define the input data matrix X = [x1, x2, ..., x n ] T , where x i For each moment t i Preprocessed multi-source time series data; S32, initializing the number of hidden layer nodes and the learning rate of the extreme learning machine, and randomly initializing the weight matrix W from the input layer to the hidden layer and the hidden layer bias vector b; S33, using the combination of activation functions, nonlinearly map the input multi-source time series data to generate the hidden layer output matrix H: H=[f1(WX+b)|f2(WX+b)]+λ·[f3(WX+b)] 2 ; Among them, f1(), f2() and f3() are activation functions, and λ is the adjustment coefficient for adjusting the weight of the secondary activation term; S34. Utilize the least squares optimization based on adaptive weighted regularization to optimize the difference between the hidden layer output matrix H and the target output matrix Y, and calculate the output layer weight β: β=(H T H+α·I+γ·diag(H T H)) -1 H T Y; Among them, α is the regularization coefficient, I is the unit matrix, γ is the weighted regularization parameter, diag(H T H) is the hidden layer output matrix H T The diagonal matrix of H; S35. Generate a network status evaluation model through the trained extreme learning machine, where the network status evaluation model can evaluate the current and future status of the industrial equipment communication network based on the input preprocessed multi-source time series data, and predict the performance indicators of the industrial equipment communication network; S36. Use the network status assessment model to predict the new real-time collected pre-processed time series data to generate the future state prediction value Y of the industrial equipment communication network pred .

5. The method for evaluating the reliability of an industrial equipment communication network according to claim 1, characterized in that: The S4 specifically includes: S41. Initialize the population size of the cat swarm optimization algorithm to N, and set the position of each cat to the hyperparameter vector P. i =[p i1 ,p i2 ,...,p im ], where p ij represents the jth hyperparameter of the i-th cat, including the number of hidden layer nodes, activation function, and learning rate; S42, define the objective function f(P i ), used to evaluate a given hyperparameter vector P i Performance: Among them, Y pred is the predicted value of the future state of the industrial equipment communication network, Y is the actual target output, n is the number of data points, m is the number of hyperparameters, and p ij is the jth hyperparameter of the i-th cat, p ij-1 is the j-1th hyperparameter value of the i-th cat, and δ is the weight of the adjustment term; S43, simulate the individual cats in the cat group according to the objective function f(P i ) to search, and perform global and local searches through position update and velocity update rules: in, is the updated position of the cat, G best is the global optimal solution position, P best is the optimal solution position of the individual cat, β1 and β2 are acceleration constants, α1 is the jump factor, P random A randomly generated position; S44, update the hyperparameters through the updated cat group positions, and substitute the hyperparameter values ​​corresponding to the new positions into the network state evaluation model to calculate the new network prediction result Y curr , and update the evaluation function value f(P i ): AND curr =Hβ+∈·(|Y curr -AND pred |)·sign(Y curr -AND pred ); Among them, ∈ is the adjustment parameter, sign() is the sign function, H is the hidden layer output matrix, and β is the output layer weight; S45, according to the evaluation function value f(P i ), determine whether the cat group has converged, and if so, output the current hyperparameter combination as the final network state evaluation model hyperparameter; S46, introduce alert behavior. When the fitness of the cat group changes little, or no better solution is found for several consecutive generations, the cat group automatically triggers alert behavior, increases the proportion of global exploration, reduces local optimization search, and promotes jumping to the area in the solution space that has not been fully explored for optimization; S47, through alert behavior, increase the proportion of global search of the cat group, and through adaptive adjustment to control the balance between global exploration and local optimization, dynamically adjust the proportion of exploration behavior and exploitation behavior; S48. After several iterations, the cat swarm optimization algorithm obtains the optimal hyperparameter combination P after completing local optimization and global search. best , update the network status assessment model and complete the optimization of the network status assessment model; S49. Use the optimized network status evaluation model to perform real-time performance evaluation of the industrial equipment communication network, and dynamically adjust key parameters of the industrial equipment communication network.

6. The method for evaluating the reliability of an industrial equipment communication network according to claim 1, characterized in that: The S5 specifically includes: S51, obtain real-time data from industrial equipment communication network, defined as X new ; S52, real-time data X new Input into the optimized network status assessment model, process the real-time, and generate hidden layer output; S53, using the optimized output layer weights, combine the hidden layer output with the weights to obtain the real-time prediction value of the industrial equipment communication network status The real-time prediction value represents the current performance assessment of the industrial equipment communication network; S54, extracting reliability evaluation indicators of the industrial equipment communication network based on the real-time prediction value, and evaluating the operation status of the industrial equipment communication network; S55, evaluating the prediction error of the industrial equipment communication network according to the deviation between the real-time prediction result and the actual monitoring value, and generating a reliability evaluation result of the industrial equipment communication network to identify potential problem areas; S56. Based on the reliability assessment results of the industrial equipment communication network, the industrial equipment communication network is continuously monitored, the optimization strategy is triggered, and the key parameters of bandwidth allocation, routing selection and load balancing are adjusted based on the actual performance of the industrial equipment communication network.

7. The method for evaluating the reliability of an industrial equipment communication network according to claim 7, characterized in that: The reliability evaluation indicators of the industrial equipment communication network specifically include network stability, throughput and latency, which are used to reflect the performance level and operation reliability of the industrial equipment communication network under different operating conditions.

8. The method for evaluating the reliability of an industrial equipment communication network according to claim 1, characterized in that: The S6 specifically includes: S61, obtaining a reliability evaluation result of the industrial equipment communication network, and determining a reliability error value of the current industrial equipment communication network; S62, comparing the reliability error with a preset threshold value θ, and when the reliability error exceeds the threshold value, triggering an industrial equipment communication network parameter optimization process; S63, constructing a parameter vector Q of the industrial equipment communication network to be optimized; S64, initial cat group optimization algorithm, set the position of each cat as the network parameter vector Q i , and set the cat group size, maximum number of iterations and behavior adjustment factors; S65. For each cat, according to the objective function G(Q i ) performs fitness evaluation on the parameter vector, and the objective function considers the prediction error of the industrial equipment communication network and the diversity penalty term of the parameter configuration at the same time: in, Indicates that in the parameter combination Q i The predicted value of the kth data point, Y (k) is the true value of the kth data point, n is the number of data points, q ij is the value of the i-th cat at the j-th parameter, is the average value of the jth parameter in the current population, μ is a small constant to avoid division by zero, and λ o A hyperparameter for adjusting the weight of the diversity penalty term; S66. Optimize the parameter vector Q using the cat swarm optimization algorithm. By introducing the adaptive behavior weight coefficient and Gaussian perturbation jump mechanism, the position of each cat is adjusted in each iteration: Among them, Q i is the parameter vector of the current i-th cat, P best is the optimal parameter vector of the cat in history, G best is the global optimal parameter vector in the entire cat population, r1 and r2 are independently generated random numbers, ω is the exploration and development balance coefficient, η is the jump intensity coefficient, ε is a random perturbation term with a mean of 0, Q i new is the updated position vector of the i-th cat; S67. In each round of iteration, according to the objective function G(Q i ) updates individual fitness and outputs the global optimal parameter combination Q after reaching the maximum number of iterations or convergence conditions opt ; S68, the optimal parameter vector Q opt Applied to industrial equipment communication networks, it updates the configuration of bandwidth allocation, routing selection, and load balancing strategies, monitors the adjusted industrial equipment communication network status in real time, and feeds back new data for optimization and iteration.

9. The method for evaluating the reliability of an industrial equipment communication network according to claim 8, characterized in that: The parameter vector of the industrial equipment communication network to be optimized specifically includes a bandwidth allocation ratio, a routing priority weight, and a load balancing strategy factor, which are used to dynamically regulate network resource allocation and transmission paths.

10. An industrial equipment communication network reliability evaluation system, an industrial equipment communication network reliability evaluation method according to any one of claims 1 to 9, characterized in that: Includes the following modules: Data acquisition module, used to collect data from industrial equipment communication networks in real time; Data preprocessing module, used to remove noise, fill missing values ​​and standardize the collected data; The network status assessment model module is used to build and train the network status assessment model using the extreme learning machine to predict the network status; Cat swarm optimization module, used to dynamically optimize the hyperparameters of the network status assessment model and key network parameters; A reliability evaluation module, used to evaluate the prediction error and overall reliability of the industrial equipment communication network based on the prediction results; Parameter optimization control module, used to adjust bandwidth allocation, routing priority and load balancing strategy in real time according to the evaluation results; The monitoring and early warning module is used to continuously monitor the network status and trigger the early warning mechanism when an anomaly is detected.

Images