Server hardware exception early warning method and apparatus, device and storage medium

Abstract

The present invention relates to the technical field of computers, in particular to a server hardware exception early warning method and apparatus, a device and a storage medium, aiming to improve the accuracy and efficiency of server underlying hardware fault early warning. The method comprises: separately inputting a processing information amount change rate data and end current change rate data of each hardware on a server in a preset time period into a pre-trained fuzzy reasoning model, the fuzzy reasoning model being obtained by training on the basis of a hybrid metaheuristic optimization algorithm (S31); the fuzzy reasoning model obtaining a predicted temperature value of the hardware on the basis of the processing information amount change rate data and the end current change rate data (S32); when the difference between the predicted temperature value and an actual temperature value of the hardware is greater than a preset temperature difference threshold, marking the hardware as abnormal hardware (S33); and generating early warning information corresponding to the abnormal hardware (S34).

Description

A server hardware abnormality warning method, device, equipment and storage medium

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] This application claims priority to the Chinese patent application filed with the China Patent Office on October 31, 2023, with application number 202311433545.7, and entitled “A Server Hardware Abnormality Warning Method, Device, Equipment and Storage Medium”, all contents of which are incorporated by reference into this application. Technical Field

[0003] Some embodiments of the present application relate to the field of computer technology, and in particular, to a server hardware anomaly warning method, apparatus, device, and storage medium. Background Art

[0004] Servers are high-performance computers that provide various services to the outside world. They perform enormous daily computations. If an anomaly in the server's underlying hardware causes an unexpected temporary downtime, it can lead to data loss, application interruptions, and other issues, severely impacting tasks that rely on the server and reducing work efficiency. Therefore, it's necessary to monitor the server's underlying hardware in real time and issue timely warnings when hardware anomalies are detected to ensure normal server operation. In related technologies, using neural networks to provide anomaly warnings for server hardware is a relatively efficient method.

[0005] In related technologies, the limitations of the algorithms used by neural networks affect the accuracy of neural networks in providing abnormal warnings for the underlying hardware of servers. They cannot effectively provide abnormal warnings for devices with low circuit unit repeatability, transistor density, and small scale, and cannot effectively prevent server failures.

[0006] Summary of the Invention

[0007] Some embodiments of the present application provide a server hardware anomaly warning method, apparatus, device, and storage medium, aiming to improve the accuracy and efficiency of server underlying hardware failure warning.

[0008] According to a first aspect of some embodiments of the present application, a server hardware abnormality warning method is provided, comprising:

[0009] The data on the rate of change of information processed by each hardware on the server and the rate of change of terminal current within a preset time period are respectively input into a pre-trained fuzzy inference model, which is trained based on a hybrid meta-heuristic optimization algorithm.

[0010] The fuzzy inference model obtains the predicted temperature value of the hardware based on the data of the rate of change of the processed information volume and the rate of change of the terminal current;

[0011] If the difference between the predicted temperature value and the actual temperature value of the hardware is greater than a preset temperature difference threshold, the hardware is marked as abnormal hardware;

[0012] Generates warning information corresponding to abnormal hardware.

[0013] In some embodiments, the step of training the fuzzy inference model includes:

[0014] Collect training samples to obtain a training sample data set, where the training samples include data on the rate of change of processed information volume, data on the rate of change of terminal current, and measured temperature values ​​of each hardware on the server within a preset time period;

[0015] Input the training sample data set into the fuzzy inference model to be trained;

[0016] Using the fuzzy clustering method, the parameters of the fuzzy inference model to be trained are initialized according to the training samples to obtain the initial fuzzy inference model;

[0017] According to the parameters of the initial fuzzy inference model, the initial parameters of the hybrid metaheuristic optimization algorithm are set;

[0018] The parameters of the initial fuzzy inference model are optimized by a hybrid meta-heuristic optimization algorithm to obtain a trained fuzzy inference model.

[0019] In some embodiments, the method further comprises:

[0020] Set the corresponding temperature difference threshold according to the preset temperature difference threshold setting rule.

[0021] In some embodiments, setting a corresponding temperature difference threshold according to a preset temperature difference threshold setting rule includes:

[0022] Determine the root mean square error between the predicted temperature values ​​obtained by the fuzzy inference model and the corresponding measured temperature values;

[0023] According to the root mean square error, set the corresponding temperature difference threshold.

[0024] In some embodiments, a fuzzy clustering method is used to initialize parameters of a fuzzy inference model to be trained based on training samples to obtain an initial fuzzy inference model, including:

[0025] Randomly selecting a first preset number of initial cluster centers from the training samples;

[0026] According to the preset optimization goal, the initial cluster center is iteratively optimized to obtain the optimized cluster center and the corresponding clustering parameters;

[0027] According to the optimized cluster centers, clustering is performed to generate the corresponding network structure;

[0028] The network parameters of the fuzzy reasoning model are assigned according to the clustering parameters to obtain the initial fuzzy reasoning model.

[0029] In some embodiments, the optimization goal is to minimize the weighted sum of the distance and the degree of membership from a data point in the training sample to its corresponding cluster center.

[0030] In some embodiments, setting initial parameters of the hybrid meta-heuristic optimization algorithm according to the parameters of the initial fuzzy inference model includes:

[0031] According to the number of parameters of the initial fuzzy inference model, the number of particles in the particle swarm of the hybrid metaheuristic optimization algorithm is determined;

[0032] According to the parameter values ​​of the initial fuzzy inference model, the particle parameter value corresponding to each particle in the particle swarm is determined.

[0033] In some embodiments, the method further comprises:

[0034] Determining the numerical range of the particle parameter value of each particle in the particle swarm according to the parameter value of the initial fuzzy inference model;

[0035] According to the numerical range, each particle in the particle swarm is randomly assigned a value.

[0036] In some embodiments, the parameters of the initial fuzzy inference model are optimized by a hybrid meta-heuristic optimization algorithm to obtain a trained fuzzy inference model, including:

[0037] Determining the parameters of each particle in a particle swarm in a hybrid metaheuristic optimization algorithm;

[0038] Through the pre-set fitness function, the parameters of the particle swarm are iteratively updated to obtain the result particles;

[0039] The parameters of the result particles are used as the parameters of the fuzzy inference model to obtain a trained fuzzy inference model.

[0040] In some embodiments, the parameters of the particle swarm are iteratively updated using a preset fitness function to obtain result particles, including:

[0041] Determine the fitness value of each particle in the particle swarm through the preset fitness function;

[0042] When the fitness value does not meet the iteration stopping criterion, the optimized particle swarm algorithm is used to update the particle positions in the particle swarm;

[0043] The particle swarm is updated through the genetic algorithm to obtain the updated particle swarm;

[0044] Determine the fitness value of each particle in the updated particle swarm through the fitness function;

[0045] The above steps are iteratively executed, and when the fitness value meets the preset iteration stopping criterion, the result particle is obtained.

[0046] In some embodiments, updating a particle swarm by a genetic algorithm to obtain an updated particle swarm includes:

[0047] By selecting operators, crossover operators and mutation operators, each particle in the particle swarm is updated to obtain a new particle swarm;

[0048] Determine the fitness value of each particle in the new particle swarm through the fitness function;

[0049] The above steps are iteratively performed, and when the fitness value satisfies the preset stopping rule, an updated particle swarm is obtained.

[0050] In some embodiments, each particle in the particle swarm is updated by selecting an operator, a crossover operator, and a mutation operator to obtain a new particle swarm, including:

[0051] Multiple target particles are obtained by selecting multiple particles in the particle swarm through the selection operator;

[0052] Through the crossover operator, multiple target particles are cross-calculated to obtain the crossover particles;

[0053] The crossed particles replace the original particles and are put into the particle swarm to obtain the initially updated particle swarm;

[0054] Through the mutation operator, the initially updated particle swarm is mutated to obtain a new particle swarm.

[0055] In some embodiments, the fitness function is used to determine a root mean square error corresponding to each particle in the particle swarm.

[0056] In some embodiments, the method further comprises:

[0057] The abnormal information corresponding to the abnormal hardware is stored in the background log.

[0058] In some embodiments, the method further comprises:

[0059] The warning information is sent to the corresponding remote address using the preset warning information sending method.

[0060] In some embodiments, the method further comprises:

[0061] The data on the rate of change of the amount of processed information of each hardware and the data on the rate of change of the terminal current are put into the training set used for training the fuzzy inference model to obtain an updated training set;

[0062] The updated training set is used to update the parameters of the fuzzy inference model to obtain a fuzzy inference model with updated parameters.

[0063] In some embodiments, the method further comprises:

[0064] Determine the root mean square error between the predicted temperature value obtained by the model inference model and the corresponding measured temperature value;

[0065] The preset temperature difference threshold is updated according to the root mean square error.

[0066] According to a second aspect of some embodiments of the present application, a server hardware abnormality warning device is provided, the device comprising:

[0067] A data input module is used to input the data of the rate of change of the amount of processed information and the rate of change of the terminal current of each hardware on the server within a preset time period into a pre-trained fuzzy inference model, which is trained based on a hybrid meta-heuristic optimization algorithm;

[0068] The predicted temperature value acquisition module is used for the fuzzy inference model to obtain the predicted temperature value of the hardware based on the data of the rate of change of the processed information volume and the data of the rate of change of the terminal current;

[0069] An abnormal hardware marking module is used to mark the hardware as abnormal hardware when the difference between the predicted temperature value and the actual temperature value of the hardware is greater than a preset temperature difference threshold;

[0070] The warning information generation module is used to generate warning information corresponding to abnormal hardware.

[0071] In some embodiments, the apparatus further includes a fuzzy inference model training module, the fuzzy inference model training module including:

[0072] The training sample collection submodule is used to collect training samples to obtain a training sample data set. The training samples include the processing information volume change rate data, terminal current change rate data and measured temperature value of each hardware on the server within a preset time period.

[0073] The training sample input submodule is used to input the training sample data set into the fuzzy inference model to be trained;

[0074] The model initialization submodule is used to use the fuzzy clustering method to initialize the parameters of the fuzzy inference model to be trained according to the training samples to obtain the initial fuzzy inference model;

[0075] A parameter setting submodule is used to set initial parameters of the hybrid meta-heuristic optimization algorithm according to the parameters of the initial fuzzy inference model;

[0076] The parameter optimization submodule is used to optimize the parameters of the initial fuzzy inference model through a hybrid meta-heuristic optimization algorithm to obtain a trained fuzzy inference model.

[0077] In some embodiments, the fuzzy inference model training module further includes:

[0078] The temperature difference threshold setting submodule is used to set the corresponding temperature difference threshold according to the preset temperature difference threshold setting rule.

[0079] In some embodiments, the temperature difference threshold setting submodule includes:

[0080] A root mean square error determination submodule is used to determine the root mean square error between the predicted temperature value obtained by the fuzzy inference model and the corresponding measured temperature value;

[0081] The temperature difference threshold determination submodule is used to set the corresponding temperature difference threshold according to the root mean square error.

[0082] In some embodiments, the model initialization submodule includes:

[0083] A cluster center selection submodule, configured to randomly select a first preset number of initial cluster centers from the training samples;

[0084] The iterative optimization submodule is used to iteratively optimize the initial cluster centers according to the preset optimization objectives to obtain the optimized cluster centers and corresponding clustering parameters;

[0085] The network structure generation submodule is used to generate the corresponding network structure based on the optimized cluster centers and cluster divisions;

[0086] The initial fuzzy inference model acquisition submodule is used to assign network parameters of the fuzzy inference model according to clustering parameters to obtain the initial fuzzy inference model.

[0087] In some embodiments, the optimization goal is to minimize the weighted sum of the distance and the degree of membership from a data point in the training sample to its corresponding cluster center.

[0088] In some embodiments, the parameter setting submodule includes:

[0089] The particle number setting submodule is used to determine the number of particles in the particle swarm of the hybrid metaheuristic optimization algorithm according to the number of parameters of the initial fuzzy inference model;

[0090] The particle parameter value determination submodule is used to determine the particle parameter value corresponding to each particle in the particle swarm according to the parameter value of the initial fuzzy inference model.

[0091] In some embodiments, the parameter setting submodule further includes:

[0092] The value range determination submodule is used to determine the numerical range of the particle parameter value of each particle in the particle swarm according to the parameter value of the initial fuzzy inference model;

[0093] The random assignment submodule is used to randomly assign a value to each particle in the particle swarm according to a numerical range.

[0094] In some embodiments, the parameter optimization submodule includes:

[0095] A particle parameter determination submodule is used to determine the parameters of each particle in the particle swarm in the hybrid meta-heuristic optimization algorithm;

[0096] The result particle acquisition submodule is used to iteratively update the parameters of the particle swarm through a preset fitness function to obtain the result particles;

[0097] The model acquisition submodule is used to use the parameters of the result particles as the parameters of the fuzzy inference model to obtain a trained fuzzy inference model.

[0098] In some embodiments, the result particle acquisition submodule includes:

[0099] A first fitness value determination submodule is used to determine the fitness value of each particle in the particle swarm through a preset fitness function;

[0100] The particle position update submodule is used to update the particle positions in the particle swarm using the optimized particle swarm algorithm when the fitness value does not meet the iteration stopping criterion;

[0101] The genetic calculation submodule is used to update the particle swarm through the genetic algorithm to obtain the updated particle swarm;

[0102] A second fitness value determination submodule is used to determine the fitness value of each particle in the updated particle swarm through a fitness function;

[0103] The result particle determination submodule is used to iteratively execute the above steps and obtain the result particle when the fitness value meets the preset iteration stopping criterion.

[0104] In some embodiments, the genetic calculation submodule includes:

[0105] The genetic update submodule is used to update each particle in the particle swarm through the selection operator, crossover operator and mutation operator to obtain a new particle swarm;

[0106] The third fitness value determination submodule is used to determine the fitness value of each particle in the new particle swarm through the fitness function;

[0107] The updated particle swarm acquisition submodule is used to iteratively execute the above steps and obtain the updated particle swarm when the fitness value meets the preset stopping rule.

[0108] In some embodiments, the genetic update submodule includes:

[0109] The target particle acquisition submodule is used to select multiple particles in the particle swarm through the selection operator to obtain multiple target particles;

[0110] The cross particle acquisition submodule is used to perform cross calculations on multiple target particles in pairs through the cross operator to obtain the cross particles;

[0111] The submodule for obtaining the initially updated population is used to replace the original particles with the crossed particles and put them into the particle swarm to obtain the initially updated particle swarm;

[0112] The particle swarm mutation submodule is used to mutate the initially updated particle swarm through the mutation operator to obtain a new particle swarm.

[0113] In some embodiments, the fitness function is used to determine the root mean square error corresponding to each particle in the particle swarm.

[0114] In some embodiments, the apparatus further comprises:

[0115] The abnormal information storage module is used to store the abnormal information corresponding to the abnormal hardware in the background log.

[0116] In some embodiments, the method further comprises:

[0117] The warning information sending module is used to send the warning information to the corresponding remote address in a preset warning information sending method.

[0118] In some embodiments, the apparatus further comprises:

[0119] A training set updating module is used to put the data on the rate of change of the amount of processed information and the rate of change of the terminal current of each hardware into the training set used for training the fuzzy inference model to obtain an updated training set;

[0120] The parameter updating module is used to update the parameters of the fuzzy inference model using the updated training set to obtain the fuzzy inference model with updated parameters.

[0121] In some embodiments, the apparatus further comprises:

[0122] a root mean square error determination module, for determining the root mean square error between the predicted temperature value obtained by the model inference model and the corresponding measured temperature value;

[0123] The temperature difference threshold updating module is used to update the preset temperature difference threshold according to the root mean square error.

[0124] A third aspect of some embodiments of the present application provides a non-volatile readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the steps in the method of the first aspect of the present application are implemented.

[0125] A fourth aspect of some embodiments of the present application provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the method of the first aspect of the present application are implemented.

[0126] Using the hardware anomaly warning method provided by the present application, the information processing rate of change data and the terminal current rate of change data of each hardware on the server within a preset time period are respectively input into a pre-trained fuzzy inference model, and the fuzzy inference model is trained based on a hybrid meta-heuristic optimization algorithm; the fuzzy inference model obtains the predicted temperature value of the hardware based on the information processing rate of change data and the terminal current rate of change data; when the difference between the predicted temperature value and the actual temperature value of the hardware is greater than a preset temperature difference threshold, the hardware is marked as abnormal hardware; and warning information corresponding to the abnormal hardware is generated. In this method, a pre-trained fuzzy inference model is used to obtain the predicted temperature value of the hardware based on the processing information change rate data and the terminal current change rate data of each hardware on the server within a preset time period. The actual temperature value of the current hardware is compared with the predicted temperature value to determine whether the hardware is abnormal. When the hardware is abnormal, corresponding early warning information is generated. The pre-trained fuzzy inference model is trained based on a hybrid meta-heuristic optimization algorithm (IQPSO-GA). The hybrid meta-heuristic optimization algorithm combines IQPSO (optimized particle swarm algorithm) and GA (genetic algorithm). The IQPSO algorithm is optimized on the basis of QPSO (particle swarm algorithm), which is more conducive to global search of data and significantly improves the training effect of the model. The trained fuzzy inference model can better capture the relationship between the processing information volume and the off-current value and the hardware temperature, thereby improving the accuracy and efficiency of hardware abnormality early warning. BRIEF DESCRIPTION OF THE DRAWINGS

[0127] In order to more clearly illustrate the technical solutions of some embodiments of the present application, the following briefly introduces the drawings required for use in the description of some embodiments of the present application. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0128] Figure 1 is a schematic diagram of the fuzzy reasoning model structure;

[0129] FIG2 is a schematic diagram of network result generation proposed in some embodiments of the present application;

[0130] FIG3 is a flow chart of a hybrid metaheuristic optimization algorithm proposed in some embodiments of the present application;

[0131] FIG4 is a flowchart of a hardware anomaly warning method proposed in some embodiments of the present application;

[0132] FIG5 is a schematic diagram of a sample data set proposed in some embodiments of the present application;

[0133] FIG6 is a diagram showing the training and testing results of the fuzzy inference model proposed in some embodiments of the present application;

[0134] FIG7 is a diagram showing the results of model training and testing based on a hybrid meta-heuristic optimization algorithm according to some embodiments of the present application;

[0135] FIG8 is a schematic diagram of a server hardware abnormality simulation warning proposed in some embodiments of the present application;

[0136] FIG9 is a schematic diagram of a hardware anomaly warning device proposed in some embodiments of the present application;

[0137] FIG10 is a schematic diagram of an electronic device according to some embodiments of the present application. DETAILED DESCRIPTION

[0138] The following will be combined with the accompanying drawings of some embodiments of the present application to clearly and completely describe the technical solutions in some embodiments of the present application. Obviously, some of the embodiments described are part of the embodiments of the present application, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on some embodiments of the present application without creative work are within the scope of protection of this application.

[0139] In some embodiments, a fuzzy inference model (ANFIS, Adaptive Network-based Fuzzy Inference System) is first established through training. The training steps of the fuzzy inference model include:

[0140] S11: Collect training samples to obtain a training sample data set, where the training samples include data on a change rate of processed information volume, a change rate of terminal current, and a measured temperature value of each hardware on the server within a preset time period.

[0141] In some embodiments, a training sample data set is a collection of training data used when training a neural network. The training samples include data on the rate of change of information volume processed by each hardware on the server within a preset time period, data on the rate of change of terminal current, and corresponding measured temperature values. The amount of information processed by the hardware is the amount of valid information processed by the hardware per unit time. Different business data processing modes may generate different amounts of information. The rate of change of information volume processed is the rate of change of information volume processed within a period of time, obtained based on the amount of information processed by the hardware within that period of time. The terminal current value refers to the component terminal current of each hardware in the server when processing data, and is correlated with the amount of information. Generally speaking, the more information processed or the more complex the information, the higher the current value. The terminal current change data is the rate of change of the off-state current value within a period of time, obtained based on the amount of information processed by the hardware within that period of time. The temperature value is the heat generated by each hardware in the server during operation. The more electricity consumed, the higher the temperature value.

[0142] In some embodiments, after collecting the processing information volume and terminal current value of each hardware within a preset time period, the collected raw data is input into a data processing unit, and the received data is processed by the data processing unit to obtain the processing information volume change rate and the terminal current value change rate.

[0143] In some embodiments, collecting raw data and generating a training sample data set plays a vital role in the training of the neural network, with the goal that the final training sample data set can enable the neural network to be better trained. The amount of information processed by each hardware in the server, the terminal current and temperature of the components are important parameters related to the hardware. When the various components of the server are working normally, the business data information processed per unit time under different business types is monitored in real time, the values ​​are discretized, and the terminal current and temperature change information of the components within the corresponding time are measured. By analyzing the amount of information processed, the terminal current and the corresponding temperature change information of each hardware, the relationship between the amount of information processed and the terminal current of each hardware and the corresponding temperature can be determined, and the collected data can be standardized, such as normalized, to obtain a training sample data set under each business type.

[0144] For example, the preset time period may be within 1 hour.

[0145] S12: Input the training samples into the fuzzy inference model to be trained.

[0146] In some embodiments, the fuzzy inference model is a model that predicts corresponding output data based on input data based on a fuzzy membership function.

[0147] Refer to Figure 1, which is a schematic diagram of the fuzzy inference model structure. In the figure, circles represent fixed nodes, and squares represent adaptive nodes. The fuzzy inference model consists of five layers: the first layer is the fuzzy layer, the second layer is the rule layer, the third layer is the normalization layer, the fourth layer is the defuzzification layer, and the fifth layer is the output layer. In the fuzzy inference model, the input data is inferred through the membership function to obtain the prediction result. The parameters of the membership function are determined by training the sample data set. The combination or interaction of the membership functions is called a rule. The description of the rules is as follows:

[0148] if x=A1,y=B1,z=C1

[0149] Rule 1: then f1=m1x+p1y+q1z+r1 (1)

[0150] if x=A2,y=B2,z=C2

[0151] Rule 2: then f2=m2x+p2y+q2z+r2(2)

[0152] if x=A3,y=B3,z=C3

[0153] Rule 3: then f3=m3x+p3y+q3z+r3(3)

[0154] Where x, y, and z are the inputs of the node; f is the output; A i 、B i and C i are fuzzy sets related to input x, y, and z, i = 1, 2, 3; m i 、p i ,q i and r i It is the result parameter, usually called the consequent parameter.

[0155] In some embodiments, the temperature change rate data x and the terminal current change rate data y in the training set are used as inputs in the first layer of the model.

[0156] The first fuzzy layer is used to fuzzify the input data, and the function is:

[0157] Where, O 1,i is the output value of the layer; n is the number of input signals; and is the generalized bell-shaped membership function (gbellmf), defined as:

[0158] Where a i 、b i and c i is the antecedent parameter.

[0159] The second rule layer can calculate the incentive strength of each rule, that is, the applicability of the rule to the data. The function is:

[0160] Where w i Represents the incentive intensity of the i-th rule.

[0161] The function of the third normalization layer is:

[0162] Where, is the output value of the third layer, which represents the ratio of the practicality of the i-th rule calculated for the i-th node to the sum of the applicability of all rules.

[0163] The function of the fourth deblurring layer is:

[0164] Where m i 、p i ,q i and r i It is the result parameter, usually called the consequent parameter.

[0165] The function of the fifth output layer is:

[0166] In the formula, the results of each output are summed to obtain the final output result.

[0167] In some embodiments, the final output data O 5，i is the predicted temperature value obtained based on the input data.

[0168] In some embodiments, after obtaining a training sample data set, the obtained training sample data set is input into a fuzzy inference model to be trained to train the fuzzy inference model.

[0169] S13: Using the fuzzy clustering method, according to the training samples, the parameters of the fuzzy inference model to be trained are initialized to obtain an initial fuzzy inference model.

[0170] In some embodiments, the fuzzy inference model processes input training samples using a fuzzy clustering method. This method is an improvement on traditional clustering methods. This method first randomly selects several cluster centers, from which all data points are assigned a certain fuzzy membership to the cluster centers. The cluster centers are then iteratively modified to obtain the optimal cluster centers. Finally, a list of cluster centers and the membership values ​​of each data point to each cluster center are output. This output data is used to establish the fuzzy inference model. The initial fuzzy inference model is a fuzzy inference model whose parameters have been initialized. The initial fuzzy inference model already possesses predictive capabilities, but the accuracy of the predictions cannot be guaranteed, and further parameter adjustment is required.

[0171] In some embodiments, the specific steps of initializing parameters of the fuzzy inference model to be trained based on the training samples to obtain the initial fuzzy inference model include:

[0172] S13-1: Randomly select a first preset number of initial cluster centers from the training samples.

[0173] In some embodiments, the first preset number can be set according to the actual amount of data, and the initial cluster centers are a number of data points randomly specified in the data set of the training sample.

[0174] In some embodiments, a first preset number of cluster centers are randomly selected from the training samples, and the selected cluster centers are used as initial cluster centers.

[0175] S13-2: According to the preset optimization target, the initial cluster centers are iteratively optimized to obtain the optimized cluster centers and corresponding clustering parameters.

[0176] In some embodiments, the preset optimization target is a parameter update target for clustering optimization of the entire network, and the clustering parameters refer to the parameters of the fuzzy inference model when clustering data in the training sample data set.

[0177] In some embodiments, the initial cluster centers are iteratively optimized according to a preset optimization goal. When the cluster centers cannot be optimized further, the iteration is stopped to obtain the optimized cluster centers and corresponding clustering parameters.

[0178] S13-3: Based on the optimized cluster centers, clustering is performed to generate the corresponding network structure.

[0179] In some embodiments, after the optimized cluster centers are determined, corresponding network structures are generated by dividing the plurality of cluster centers as center points.

[0180] S13-4: Assigning network parameters of the fuzzy reasoning model according to the clustering parameters to obtain an initial fuzzy reasoning model.

[0181] In some embodiments, the network parameters of the fuzzy reasoning model are assigned according to the clustering parameters when the cluster center is finally obtained to obtain the initial fuzzy reasoning model.

[0182] Referring to Figure 2, which is a schematic diagram of network result generation according to some embodiments of the present application, a training dataset is input into the fuzzy inference network. The fuzzy inference network then clusters the input samples using a fuzzy clustering method combined with initial clustering parameters. By clustering the input samples, a corresponding network structure is generated. Based on the clustering results, the network parameters are assigned, i.e., the network parameters obtained when the clustering results are obtained are used as the final network parameters of the fuzzy inference network, and finally, the initial fuzzy inference model is output.

[0183] In some embodiments, the optimization objective is to minimize the weighted sum of the distance and the degree of membership from a data point in the training sample to its corresponding cluster center.

[0184] In some embodiments, when the weighted sum of the distance and membership degree between each data point in the training sample data set and its corresponding cluster center reaches a minimum, the iteration ends and the optimization goal is achieved.

[0185] S14: According to the parameters of the initial fuzzy inference model, initial parameters of the hybrid metaheuristic optimization algorithm are set.

[0186] In some embodiments, the hybrid meta-heuristic optimization algorithm (IQPSO-GA) is an algorithm created by combining an optimized particle swarm optimization algorithm (IQPSO) with a genetic algorithm (GA).

[0187] For the optimized particle swarm algorithm, first, the speed and position updates of the ordinary particle swarm optimization (PSO) algorithm example are determined by the following equations:

[0188] Where ω is the inertia weight, which plays a key balancing role in the global search and local search of the algorithm; v i,t represents the velocity vector of the i-th particle in the population at the t-th iteration. i,t represents the position vector of the i-th particle at the t-th iteration; P i,t-1 is the optimal position of the i-th particle; G t-1 is the global optimal position of the entire population; c1 and c2 are acceleration constants; r1 and r2 are random numbers between [0,1] that obey a uniform distribution.

[0189] In some embodiments, a particle is equivalent to a parameter of the fuzzy inference model, and the above formula is used to randomly initialize the number of particles equal to the number of inference model parameters. From the above formula, it can be seen that the fuzzy inference model has a total of 7 unknown parameter values, a i 、b i 、c i 、m i 、p i ,q i 、r i Then set the number of ions to 7, and for each of these 7 particles, use the above formula to initialize the particle parameters.

[0190] OPSO (Quantum Particle Sort Algorithm) is based on the standard PSO algorithm and introduces an optimization evolutionary algorithm based on quantum theory. In QPSO, the position of the particle is determined by the following equation:

[0191] Where, Mbest t represents the average value of the optimal position of all particles at the tth iteration; N is the number of particles in the population; D is the particle dimension; p ij,t P ij,t and G j,t Random positions between ij,t G represents the optimal position of the jth dimension of the i-th particle in the population at the t-th iteration; j,t Represents the j-th dimension position of the global optimal solution of the population; and u ij,t Are all random numbers between [0,1]; x ij,t represents the position of the j-th dimension of the i-th particle at the t-th iteration; β t β is the contraction-expansion coefficient, which is an important parameter used to control the convergence speed of the QPSO algorithm and is usually changed by a linear reduction method. t =0.5(t max -t)+0.5 (15)

[0192] Where t represents the current number of iterations, t max Indicates the maximum number of iterations set.

[0193] QPSO introduces Mbest t and p ij,t The concept of improves inter-particle collaboration and global search capabilities, but it also affects the efficiency of the particle global search. To improve particle search efficiency, we adjust the particle positions from two perspectives: the optimal position and the shrink-expansion coefficient. The weighted average of the optimal positions of all particles at iteration t is set.

[0194] Where, α ij,t For the weight coefficient of each particle, set it as follows:

[0195] Where, Denotes the global optimal solution G t The j-th dimension corresponds to the particle fitness function value, F ij Represents the fitness function value of the jth dimension of the i-th particle.

[0196] In practical applications, this adjustment method of the shrinkage-expansion coefficient cannot be reasonably changed during the execution of the algorithm. Therefore, in some embodiments, an error function (Error_F) is defined. The error function represents the distance relationship between the particle and the current optimal position in the particle swarm. The smaller the error function, the closer the particle is to the current global optimal position, the smaller the search range of the particle becomes, and it is easy to converge prematurely. When the error function is larger, the farther the particle is from the current global optimal position, the larger the particle search range becomes, resulting in a slower convergence speed.

[0197] In some embodiments, the error function is defined as:

[0198] Where, F i represents the fitness function of the i-th particle; F Gbest Denotes the global optimal solution G t The fitness function of the corresponding particle.

[0199] Based on the definition of the error function, the contraction-expansion coefficient β t can be adjusted as follows. κ = log 10 (Error_F) (19)

[0200] Where, β i,t represents the contraction-expansion coefficient of the i-th particle at the t-th iteration; when Error_F is large, β i,t Take a smaller value to speed up the convergence; when Error_F is small, β i,t Take a larger value to expand the search range and avoid falling into the local optimum.

[0201] In some embodiments, the model parameters are adaptively adjusted using an error function, and the shrinkage-expansion coefficient is adjusted according to actual conditions, ultimately obtaining an optimized particle swarm algorithm.

[0202] In some embodiments, in the optimized particle swarm algorithm, the initialization of the particle swarm is random. In the early stages of the algorithm iteration, there are certain blind spots in the population search, which not only reduces the search efficiency of the algorithm but also affects the stability of the algorithm. In addition, as the optimized particle swarm algorithm is continuously iterated and updated, the diversity of the particle swarm will decrease, and it will fall into a local optimum.

[0203] In some embodiments, in order to ensure the diversity of the particle swarm, a genetic algorithm is introduced. The genetic algorithm updates all chromosomes, i.e., particles, in the current population by utilizing its own selection operator, crossover operator, and mutation operator. It is highly efficient in searching for the global optimal solution. When used in combination with the optimized particle swarm algorithm, it effectively compensates for the disadvantage that the optimized particle swarm algorithm is prone to falling into local optimality, thereby obtaining a hybrid metaheuristic optimization algorithm.

[0204] Referring to Figure 3, Figure 3 is a flow chart of a hybrid meta-heuristic optimization algorithm proposed in some embodiments of the present application. As shown in the figure, the data in the data set is used as a particle swarm, the particle swarm is randomly initialized, and the fitness value of each particle is calculated. When the stopping criterion is met, the optimal particle is output and the algorithm ends. When the stopping criterion is not met, the swarm particle position, that is, the particle position in the particle swarm, is updated, and the particle swarm is processed by a GA operation (genetic algorithm) to generate a new population, update the local optimal value and the global optimal value of the population, obtain an updated particle swarm, and then calculate the fitness value of each particle in the updated particle swarm until the stopping criterion is met, and the iterative process ends.

[0205] In some embodiments, the fitness function in the hybrid metaheuristic optimization algorithm is used to obtain the fitness value of each particle, and the root mean square error corresponding to the algorithm particle is used as the current fitness value of the particle. The higher the fitness value, the more accurate the model prediction result, and the closer the model parameters are to the optimal parameters.

[0206] In some embodiments, the fitness function calculation expression of each particle fitness value is:

[0207] Where M represents the total number of data vectors in the training dataset; Represents particle x i,t Predict the prediction result of the mth sample data; Represents the mth sample data in the training data set.

[0208] In some embodiments, by setting the above formula to optimize the particles, the parameters of the fuzzy inference model are optimized. When the particles are optimized to the optimal value, the parameters of the finally optimized particles are used as the parameters of the inference model. The fuzzy inference model has a total of 7 unknown parameter values, a i 、bi 、c i 、m i 、p i ,q i 、r i After the optimization is completed, the optimal parameter values ​​of particles 1 to 7 corresponding to these 7 parameters are obtained, and then the parameters of the fuzzy inference model are determined.

[0209] In some embodiments, the fitness function is to calculate the mean square error between the predicted result and the actual result obtained by the fuzzy inference model under the parameter. The smaller the mean square error, the closer the parameter is to the optimal parameter, that is, the closer the parameter of the particle is to the optimal. Finally, the optimized particle uses the vector value of the particle as the parameter of the inference model, and the prediction result obtained is the best.

[0210] In some embodiments, the specific steps of setting the initial parameters of the hybrid meta-heuristic optimization algorithm according to the parameters of the initial fuzzy inference model include:

[0211] S14-1: Determine the number of particles in the particle swarm of the hybrid metaheuristic optimization algorithm according to the number of parameters of the initial fuzzy inference model.

[0212] In some embodiments, when setting the number of particle swarms in the hybrid meta-heuristic optimization algorithm, the number of parameters of the initial fuzzy inference model is used as the number of particle swarms in the hybrid meta-heuristic optimization algorithm.

[0213] For example, when there are 10 parameters that need to be adjusted in the initial fuzzy inference model, that is, when the number of parameters is 10, the number of particles in the particle swarm is set to 10, that is, there are 10 particles in the particle swarm.

[0214] S14-2: Determine the particle parameter value corresponding to each particle in the particle swarm according to the parameter value of the initial fuzzy inference model.

[0215] In some embodiments, after determining the parameter values ​​of the initial fuzzy inference model, the parameter value of each parameter is assigned to particles in the particle swarm, and the particle parameter value corresponding to each particle in the particle swarm is determined.

[0216] In some other embodiments of the present application, the following steps are also included:

[0217] S14-3 determines the numerical range of the particle parameter value of each particle in the particle swarm according to the parameter value of the initial fuzzy inference model.

[0218] In some embodiments, the value range of the particle parameter value of each particle in the particle swarm is determined based on the parameter value of the initial fuzzy inference model, and the parameter value within a certain range is selected as the particle parameter value with the parameter value of the inference model as the center.

[0219] S14-4: Randomly assign a value to each particle in the particle swarm according to the numerical range.

[0220] In some embodiments, after the value range of the particle parameter value is determined, a random value is assigned to each particle in the particle swarm within the value range.

[0221] In some embodiments, instead of directly determining the parameter values ​​of the particles, the particles are randomly assigned values ​​within a certain range, thereby ensuring the diversity of the particles in the particle swarm and preventing the model from converging prematurely.

[0222] S15: Optimize the parameters of the initial fuzzy reasoning model through a hybrid meta-heuristic optimization algorithm to obtain a trained fuzzy reasoning model.

[0223] In some embodiments, after the initial fuzzy reasoning model is obtained and the initial parameters of the hybrid meta-heuristic optimization algorithm are set, the parameters of the initial fuzzy reasoning model are optimized by the hybrid meta-heuristic optimization algorithm to obtain a trained fuzzy reasoning model.

[0224] In some embodiments, the specific steps of optimizing the parameters of the initial fuzzy inference model using a hybrid meta-heuristic optimization algorithm to obtain a trained fuzzy inference model include:

[0225] S15-1: Determine the parameters of each particle in the particle swarm in the hybrid metaheuristic optimization algorithm.

[0226] In some embodiments, the particle swarm is initialized and the parameters of each particle in the particle swarm are determined, that is, each particle in the particle swarm is assigned a value. When each fuzzy inference model prediction is completed, the parameters of the fuzzy inference model during the current prediction are used as the parameters of the particles in the particle swarm.

[0227] S15-2: The parameters of the particle swarm are iteratively updated through a preset fitness function to obtain result particles.

[0228] In some embodiments, the result particle is the optimal particle output after the hybrid metaheuristic optimization algorithm meets a stopping criterion and stops iterating.

[0229] In some embodiments, a pre-set fitness function is used to determine the root mean square error of each particle for the predicted result, and then the root mean square error between the result predicted by the fuzzy inference model under the current parameters and the actual measured temperature value is determined. Then, a meta-heuristic optimization algorithm is used to iteratively update the parameter value of the particle swarm to obtain the result particle.

[0230] In some embodiments, the specific steps of iteratively updating the parameters of the particle swarm by using a preset fitness function to obtain result particles include:

[0231] S15-2-1: Determine the fitness value of each particle in the particle swarm through a preset fitness function.

[0232] In some embodiments, a pre-set fitness function is used to substitute the vector corresponding to each particle into the fitness function formula, i.e., Equation (21), to obtain the fitness value of each particle in the particle swarm. The fitness value represents the root mean square error between the result predicted based on the parameter and the actual result.

[0233] S15-2-2: When the fitness value does not meet the iteration stopping criterion, the optimized particle swarm algorithm is used to update the particle positions in the particle swarm.

[0234] In some embodiments, the iteration stopping criterion is a rule that needs to be satisfied for the iteration to stop.

[0235] In some embodiments, when the fitness value does not meet the iteration stopping criterion, the optimized particle swarm algorithm is used to update the positions of particles in the particle swarm, that is, to update the parameter values ​​of the fuzzy inference model.

[0236] For example, the iteration stopping criterion is that the fitness value is lower than a preset fitness value threshold, and the fitness value threshold can be set according to the accuracy requirement.

[0237] In some other embodiments of the present application, the number of iterations may be preset. When the number of iterations is reached, the iteration is stopped and the particle obtained last time is output as the optimal particle.

[0238] S15-2-3: Update the particle swarm through a genetic algorithm to obtain an updated particle swarm.

[0239] In some embodiments, the updated particle swarm is a particle swarm obtained by updating the particle swarm through a genetic algorithm.

[0240] In some embodiments, the genetic algorithm is an optimization algorithm that simulates the chromosome mutation process, treating each particle in the particle swarm as a chromosome, performing crossover mutation on the chromosome to generate a new population, and then obtaining an updated particle swarm.

[0241] In some embodiments, the particle swarm is updated by a genetic algorithm to obtain an updated particle swarm, including:

[0242] S15-2-3-1: By selecting operators, crossover operators and mutation operators, each particle in the particle swarm is updated to obtain a new particle swarm.

[0243] In some embodiments, a genetic algorithm is used to update each particle in a particle swarm by using a selection operator, a crossover operator, and a mutation operator to obtain a new particle swarm. The specific steps include:

[0244] Multiple target particles are obtained by selecting multiple particles in the particle swarm through the selection operator;

[0245] Through the crossover operator, multiple target particles are cross-calculated to obtain the crossover particles;

[0246] The crossed particles replace the original particles and are put into the particle swarm to obtain the initially updated particle swarm;

[0247] Through the mutation operator, the initially updated particle swarm is mutated to obtain a new particle swarm.

[0248] S15-2-3-2: Determine the fitness value of each particle in the new particle swarm through the fitness function.

[0249] In some embodiments, after a new particle swarm is obtained, the utility function is used to calculate the fitness value of each particle in the new particle swarm.

[0250] S15-2-3-3: Iterate the above steps to obtain an updated particle swarm when the fitness value meets the preset stopping rule.

[0251] In some embodiments, the crossover, mutation, and selection operations of chromosomes are iteratively performed until a preset stopping rule is satisfied, for example, after a preset number of iterations is reached, an updated particle swarm is obtained.

[0252] S15-2-4: Determine the fitness value of each particle in the updated particle swarm through the fitness function.

[0253] In some embodiments, the updated particle swarm is calculated using a fitness function to obtain a fitness value of each particle in the updated particle swarm.

[0254] S15-2-5: Iterate the above steps and obtain the result particles when the fitness value meets the preset iteration stopping criteria.

[0255] In some embodiments, the population particle position update and population genetic operation are iteratively performed, and when the obtained fitness value meets a preset iteration stopping criterion, the result particle is obtained.

[0256] S15-3: Using the parameters of the result particles as the parameters of the fuzzy inference model to obtain a trained fuzzy inference model.

[0257] In some embodiments, the obtained particle parameter values ​​of the result particles are used as final parameters of the fuzzy inference model to obtain a trained fuzzy inference model.

[0258] In some other embodiments of the present application, the method further includes:

[0259] S21: Setting a corresponding temperature difference threshold according to a preset temperature difference threshold setting rule.

[0260] In some embodiments, the preset temperature difference threshold setting rule is a temperature difference threshold setting rule obtained based on an empirical formula. The preset temperature difference threshold is the maximum value of the root mean square error (RMS) between the predicted temperature and the measured temperature. When the RMS error between the predicted temperature and the measured temperature exceeds the preset temperature difference threshold, it indicates that a hardware abnormality has occurred.

[0261] In some embodiments, according to a preset temperature difference threshold setting rule, the specific steps of setting the corresponding temperature difference threshold include:

[0262] S21-1: Determine the root mean square error between the predicted temperature value obtained by the fuzzy inference model and the corresponding measured temperature value.

[0263] S21-2: Set the corresponding temperature difference threshold according to the root mean square error.

[0264] In some embodiments, the root mean square error between the predicted temperature value obtained by the fuzzy inference model and the corresponding measured temperature value is determined, and then a temperature difference threshold setting rule is obtained based on an empirical formula to set the corresponding temperature difference threshold.

[0265] For example, the temperature difference threshold setting rule is to set the temperature difference threshold to 3 times the root mean square error.

[0266] Referring to FIG4 , FIG4 is a flowchart of a hardware anomaly warning method proposed in some embodiments of the present application. As shown in FIG4 , the method includes the following steps:

[0267] S31: Inputting the information processing rate change data and the terminal current change rate data of each hardware on the server within a preset time period into a pre-trained fuzzy inference model, wherein the fuzzy inference model is trained based on a hybrid meta-heuristic optimization algorithm.

[0268] In some embodiments, after obtaining the trained fuzzy inference model, the processed information volume change rate data and the terminal current change rate data of each hardware on the server within a preset time period are respectively input into the pre-trained fuzzy inference model.

[0269] For example, the hardware in the server includes AI (Artificial Intelligence) acceleration cards, Ethernet cards, RAID (Redundant Array of Independent Disks) cards and other circuit units with low repeatability and small transistor density. Since there are fewer parameters that can be monitored, it is not easy to find suitable early warning patterns. Therefore, it is necessary to use pre-trained fuzzy inference models to issue abnormal early warnings for these hardware devices.

[0270] S32: The fuzzy inference model obtains the predicted temperature value of the hardware based on the data of the rate of change of the processed information volume and the data of the rate of change of the terminal current.

[0271] In some embodiments, after the fuzzy inference model obtains the processing information volume and terminal current value of each hardware, it predicts the predicted temperature value of the hardware based on the obtained data.

[0272] S33: When the difference between the predicted temperature value and the actual temperature value of the hardware is greater than a preset temperature difference threshold, mark the hardware as abnormal hardware.

[0273] In some embodiments, the actual temperature value is the actual temperature of each hardware measured by a sensing device on the server.

[0274] In some embodiments, when the difference between the predicted temperature value and the actual temperature value of the hardware is greater than a preset temperature difference threshold, it indicates that the temperature of the hardware is abnormal, and thus the hardware is marked as abnormal hardware.

[0275] For example, when the root mean square error between the measured temperature and the predicted temperature of an Ethernet card in a server is greater than a preset temperature difference threshold, the Ethernet card is marked as abnormal hardware.

[0276] S34: Generate warning information corresponding to abnormal hardware.

[0277] In some embodiments, when there is a hardware device marked as abnormal hardware, device information of the abnormal device is obtained, and corresponding warning information is generated based on the obtained device information.

[0278] For example, the generated warning information is: "Ethernet card 01 temperature is abnormal."

[0279] In some other embodiments of the present application, the method further includes:

[0280] S41: Storing abnormal information corresponding to the abnormal hardware in the background log.

[0281] In some embodiments, when abnormal hardware is detected, abnormal information corresponding to the abnormal hardware is stored in a background log, and management personnel can view the log at any time to understand the working status of each hardware.

[0282] In some other embodiments of the present application, the method further includes:

[0283] S51: Send the warning information to the corresponding remote address in a preset warning information sending method.

[0284] In some embodiments, after the warning information is generated, the warning information is sent to a corresponding remote address in a preset warning information sending manner.

[0285] In some embodiments, a remote address can be preset. After the warning information is generated, the warning information is sent to the corresponding remote address to ensure that the management personnel receive the warning information in time, repair the server hardware fault, and ensure the normal operation of the server.

[0286] For example, the preset warning information sending method can be text message sending, email sending, etc.

[0287] In some other embodiments of the present application, the method further includes:

[0288] S61: Putting the data of the rate of change of the amount of processed information of each hardware and the prime number of the terminal current rate of change into the training set used for training the fuzzy inference model to obtain an updated training set.

[0289] In some embodiments, the training of the prediction model requires a large amount of sample data, and the trained model should be continuously updated as the data is updated. After obtaining the pre-selected trained fuzzy inference model, when the fuzzy inference model makes predictions based on the data monitored in real time, the received data is put into the training set to obtain an updated training set.

[0290] S62: Using the updated training set, update the parameters of the fuzzy inference model to obtain the fuzzy inference model with updated parameters.

[0291] In some embodiments, after obtaining the updated training set, the model is trained based on the existing fuzzy inference model using the updated training set, and the parameters of the fuzzy inference model are updated to obtain the fuzzy inference model with updated parameters.

[0292] In some embodiments, the update time of the model can be set, for example, when the amount of new data reaches a certain amount, the parameters of the model are updated, or every preset time period, the parameters of the model are updated.

[0293] In some other embodiments of the present application, the method further includes:

[0294] S71: Determine the root mean square error between the predicted temperature value obtained by the model inference model and the corresponding measured temperature value.

[0295] In some embodiments, after the parameters of the fuzzy inference model are updated, the root mean square error between the predicted temperature value obtained by the fuzzy inference model and the corresponding measured temperature value is determined through a fitness function.

[0296] S72: Update the preset temperature difference threshold according to the root mean square error.

[0297] In some embodiments, the preset temperature difference threshold is updated according to the obtained root mean square error.

[0298] In some embodiments, when the parameters of the model change, the root mean square error between the temperature value predicted by the model and the actual temperature value will also change, and the temperature difference threshold will also change accordingly. In this way, the parameters of the model and the temperature difference threshold are continuously adjusted to ensure that the fuzzy inference model can accurately monitor the temperature of the hardware and issue hardware anomaly warnings.

[0299] In some of the above-mentioned embodiments of the present application, a fuzzy inference model is trained through a hybrid meta-heuristic optimization algorithm. This fuzzy inference model is used to provide abnormal warnings for various hardware in the server. When the components are in a healthy state, the mapping relationship between various data can be analyzed based on the data information that can be measured by devices with low circuit unit complexity and small transistor density, thereby monitoring the status of the hardware. In the hybrid meta-heuristic optimization algorithm, in order to improve the efficiency of particle search, based on the particle swarm algorithm, an optimized particle swarm algorithm is designed based on the influence of the contraction-expansion coefficient on the convergence speed and search ability, and the crossover and mutation operations of the genetic algorithm are introduced. Based on this algorithm, the fuzzy inference model is trained to obtain a more accurate fuzzy inference model, which effectively improves the accuracy of the server hardware abnormality warning.

[0300] In other embodiments of the application, the fuzzy reasoning model trained based on the hybrid meta-heuristic optimization algorithm in the present application is simulated and verified.

[0301] Refer to Figure 5, which is a schematic diagram of a sample data set proposed in some embodiments of the present application. As shown in Figure 5, the simulation experiment uses a set of post-processing data sets as sample data sets for simulation experiments. The sample data sets used contain three groups of vectors. The first two groups are the rate of change of the amount of information processed by the server component and the rate of change of the component end current as input data, and the other group is the rate of change of the component temperature as output data. The sample data sets are all mapped to between 0 and 1 through normalization. There are 500 data points in the data set, and each sample data point is a non-periodic sampling result. The data set is divided into two parts. The first part randomly takes 70% of the data points as training data, and the second part takes the remaining 30% of the data as test data.

[0302] The experiment used the fuzzy C-means clustering method. Table 1 lists the optimization algorithm parameter settings, using the number of iterations as the termination criterion. These parameters were chosen based on the experience of our team through trial and error. In Table 1, "selection pressure" represents the ratio of the probability of selecting the best individual in the GA algorithm to the probability of selecting the average individual; "Gamma" is the range of values ​​determined when setting the random array during the crossover operation of the GA algorithm.

[0303] Table 1

[0304] Figure 6 shows the training and testing results of the fuzzy inference model proposed in some embodiments of the present application, and Figure 7 shows the training and testing results of the model based on the hybrid metaheuristic optimization algorithm proposed in some embodiments of the present application. Comparing Figures 6 and 7, it can be clearly seen that the traditional fuzzy inference model has a better ability to fit input-output data than the model trained based on the hybrid metaheuristic optimization algorithm. At the same time, the generalization ability of the models obtained by the two training methods was tested using test data. The results show that the root mean square error (RMSE) of the model trained based on the hybrid metaheuristic optimization algorithm is 0.011. Compared with the various indicators of the traditional fuzzy inference model, the fitting accuracy of the model trained based on the hybrid metaheuristic optimization algorithm is improved by more than 47%.

[0305] To verify the effectiveness of the proposed server anomaly warning method based on the IQPSO-GA optimized ANFIS model, we added several abnormal errors (including Gaussian noise and outlier noise) to the sample data shown in Figure 5. The error values ​​were set as shown in Figure 8 (a), which is a schematic diagram of a server hardware anomaly simulation warning proposed in some embodiments of the application. Outlier mutation noise was added between the 450th and 490th groups of sample data. The predicted values ​​of the ANFIS model optimized by IQPSO-GA were compared with the actual values ​​after adding the error, and the predicted errors are shown in Figure 8 (b) and Figure 8 (c), respectively. It can be seen that the proposed anomaly warning method based on the IQPSO-GA optimized ANFIS model can effectively map the dynamic relationship between the rate of change of the amount of information processed by the component, the rate of change of the terminal current, and the rate of change of the component temperature. In other words, it can observe whether the component temperature change rate is abnormal based on the input information. Once the terminal current or temperature of a component is continuously above the warning threshold, although the component function can still be used, its subsequent operating status needs to be paid attention to and a hardware anomaly warning is issued for the component.

[0306] Based on the same inventive concept, some embodiments of the present application provide a server hardware anomaly warning device. Referring to FIG9 , FIG9 is a schematic diagram of a hardware anomaly warning device 900 proposed in some embodiments of the present application. As shown in FIG9 , the device includes:

[0307] Data input module 901, used to input the data of the rate of change of the processed information volume and the terminal current of each hardware on the server within a preset time period into a pre-trained fuzzy inference model, which is trained based on a hybrid meta-heuristic optimization algorithm;

[0308] The predicted temperature value acquisition module 902 is used for the fuzzy inference model to obtain the predicted temperature value of the hardware according to the data of the rate of change of the processed information volume and the data of the rate of change of the terminal current;

[0309] Abnormal hardware marking module 903, configured to mark the hardware as abnormal hardware when the difference between the predicted temperature value and the actual temperature value of the hardware is greater than a preset temperature difference threshold;

[0310] The warning information generation module 904 is used to generate warning information corresponding to abnormal hardware.

[0311] In some embodiments, the apparatus further includes a fuzzy inference model training module, the fuzzy inference model training module including:

[0312] The training sample collection submodule is used to collect training samples to obtain a training sample data set. The training samples include the processing information volume change rate data, terminal current change rate data and measured temperature values ​​of each hardware on the server within a preset time period.

[0313] The training sample input submodule is used to input the training sample data set into the fuzzy inference model to be trained;

[0314] The model initialization submodule is used to use the fuzzy clustering method to initialize the parameters of the fuzzy inference model to be trained according to the training samples to obtain the initial fuzzy inference model;

[0315] A parameter setting submodule is used to set initial parameters of the hybrid meta-heuristic optimization algorithm according to the parameters of the initial fuzzy inference model;

[0316] The parameter optimization submodule is used to optimize the parameters of the initial fuzzy inference model through a hybrid meta-heuristic optimization algorithm to obtain a trained fuzzy inference model.

[0317] In some embodiments, the fuzzy inference model training module further includes:

[0318] The temperature difference threshold setting submodule is used to set the corresponding temperature difference threshold according to the preset temperature difference threshold setting rule.

[0319] In some embodiments, the temperature difference threshold setting submodule includes:

[0320] A root mean square error determination submodule is used to determine the root mean square error between the predicted temperature value obtained by the fuzzy inference model and the corresponding measured temperature value;

[0321] The temperature difference threshold determination submodule is used to set the corresponding temperature difference threshold according to the root mean square error.

[0322] In some embodiments, the model initialization submodule includes:

[0323] A cluster center selection submodule, configured to randomly select a first preset number of initial cluster centers from the training samples;

[0324] The iterative optimization submodule is used to iteratively optimize the initial cluster centers according to the preset optimization objectives to obtain the optimized cluster centers and corresponding clustering parameters;

[0325] The network structure generation submodule is used to generate the corresponding network structure based on the optimized cluster centers and cluster divisions;

[0326] The initial fuzzy inference model acquisition submodule is used to assign network parameters of the fuzzy inference model according to clustering parameters to obtain the initial fuzzy inference model.

[0327] In some embodiments, the optimization goal is to minimize the weighted sum of the distance and the degree of membership from a data point in the training sample to its corresponding cluster center.

[0328] In some embodiments, the parameter setting submodule includes:

[0329] The particle number setting submodule is used to determine the number of particles in the particle swarm of the hybrid metaheuristic optimization algorithm according to the number of parameters of the initial fuzzy inference model;

[0330] The particle parameter value determination submodule is used to determine the particle parameter value corresponding to each particle in the particle swarm according to the parameter value of the initial fuzzy inference model.

[0331] In some embodiments, the parameter setting submodule further includes:

[0332] The value range determination submodule is used to determine the numerical range of the particle parameter value of each particle in the particle swarm according to the parameter value of the initial fuzzy inference model;

[0333] The random assignment submodule is used to randomly assign a value to each particle in the particle swarm according to a numerical range.

[0334] In some embodiments, the parameter optimization submodule includes:

[0335] A particle parameter determination submodule is used to determine the parameters of each particle in the particle swarm in the hybrid meta-heuristic optimization algorithm;

[0336] The result particle acquisition submodule is used to iteratively update the parameters of the particle swarm through a preset fitness function to obtain the result particles;

[0337] The model acquisition submodule is used to use the parameters of the result particles as the parameters of the fuzzy inference model to obtain a trained fuzzy inference model.

[0338] In some embodiments, the result particle acquisition submodule includes:

[0339] A first fitness value determination submodule is used to determine the fitness value of each particle in the particle swarm through a preset fitness function;

[0340] The particle position update submodule is used to update the particle positions in the particle swarm using the optimized particle swarm algorithm when the fitness value does not meet the iteration stopping criterion;

[0341] The genetic calculation submodule is used to update the particle swarm through the genetic algorithm to obtain the updated particle swarm;

[0342] A second fitness value determination submodule is used to determine the fitness value of each particle in the updated particle swarm through a fitness function;

[0343] The result particle determination submodule is used to iteratively execute the above steps and obtain the result particle when the fitness value meets the preset iteration stopping criterion.

[0344] In some embodiments, the genetic calculation submodule includes:

[0345] The genetic update submodule is used to update each particle in the particle swarm through the selection operator, crossover operator and mutation operator to obtain a new particle swarm;

[0346] The third fitness value determination submodule is used to determine the fitness value of each particle in the new particle swarm through the fitness function;

[0347] The updated particle swarm acquisition submodule is used to iteratively execute the above steps and obtain the updated particle swarm when the fitness value meets the preset stopping rule.

[0348] In some embodiments, the genetic update submodule includes:

[0349] The target particle acquisition submodule is used to select multiple particles in the particle swarm through the selection operator to obtain multiple target particles;

[0350] The cross particle acquisition submodule is used to perform cross calculations on multiple target particles using a cross operator to obtain cross particles.

[0351] The submodule for obtaining the initially updated population is used to replace the original particles with the crossed particles and put them into the particle swarm to obtain the initially updated particle swarm;

[0352] The particle swarm mutation submodule is used to mutate the initially updated particle swarm through the mutation operator to obtain a new particle swarm.

[0353] In some embodiments, the fitness function is used to determine the root mean square error corresponding to each particle in the particle swarm.

[0354] In some embodiments, the apparatus further comprises:

[0355] The abnormal information storage module is used to store the abnormal information corresponding to the abnormal hardware in the background log.

[0356] In some embodiments, the method further comprises:

[0357] The warning information sending module is used to send the warning information to the corresponding remote address in a preset warning information sending method.

[0358] In some embodiments, the apparatus further comprises:

[0359] A training set updating module is used to put the data on the rate of change of the amount of processed information and the rate of change of the terminal current of each hardware into the training set used for training the fuzzy inference model to obtain an updated training set;

[0360] The parameter updating module is used to update the parameters of the fuzzy inference model using the updated training set to obtain the fuzzy inference model with updated parameters.

[0361] In some embodiments, the apparatus further comprises:

[0362] a root mean square error determination module, for determining the root mean square error between the predicted temperature value obtained by the model inference model and the corresponding measured temperature value;

[0363] The temperature difference threshold updating module is used to update the preset temperature difference threshold according to the root mean square error.

[0364] Based on the same inventive concept, some other embodiments of the present application provide a non-volatile readable storage medium on which a computer program is stored. When the program is executed by a processor, the steps in the hardware anomaly warning method of some of the above embodiments of the present application are implemented.

[0365] Based on the same inventive concept, some other embodiments of the present application provide an electronic device 1000, as shown in Figure 10. Figure 10 is a schematic diagram of an electronic device shown in some embodiments of the present application, including a memory 1002, a processor 1001, and a computer program stored in the memory and executable on the processor. When executed by the processor, the steps of the network service providing method in some of the above embodiments of the present application are implemented.

[0366] As for some embodiments of the device, since they are basically similar to some embodiments of the method, the description is relatively simple, and the relevant parts can be referred to the partial description of some embodiments of the method.

[0367] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0368] Those skilled in the art will appreciate that some embodiments of the present application may be provided as methods, devices, or computer program products. Therefore, some embodiments of the present application may take the form of some embodiments entirely in hardware, some embodiments entirely in software, or some embodiments combining software and hardware. Furthermore, some embodiments of the present application may take the form of a computer program product implemented on one or more non-volatile readable storage media (including but not limited to magnetic disk storage, CD-ROM (Compact Disc Read-Only Memory), optical storage, etc.) containing computer-usable program code.

[0369] Some embodiments of the present application are described with reference to the flowcharts and / or block diagrams of the methods, terminal devices (systems), and computer program products according to some embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing terminal device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing terminal device generate a device for implementing the functions specified in one or more processes in the flowchart and / or one or more boxes in the block diagram.

[0370] These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing terminal device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce a manufactured product including an instruction device that implements the functions specified in one or more processes in the flowchart and / or one or more boxes in the block diagram.

[0371] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal device so that a series of operating steps are executed on the computer or other programmable terminal device to produce computer-implemented processing, so that the instructions executed on the computer or other programmable terminal device provide steps for implementing the functions specified in one or more processes in the flowchart and / or one or more boxes in the block diagram.

[0372] Although preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they become aware of the basic inventive concepts. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0373] Finally, it should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal device that includes a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or terminal device. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of additional identical elements in the process, method, article, or terminal device that includes the element.

[0374] The above is a detailed introduction to the hardware anomaly warning method, device, equipment and storage medium provided by this application. Specific examples are used in this article to illustrate the principles and implementation methods of this application. The descriptions of some of the above embodiments are only used to help understand the method of this application and its core ideas. At the same time, for general technical personnel in this field, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as a limitation on this application.

Claims

1. A server hardware abnormality early warning method, characterized in that: The method comprises: Inputting the processing information volume change rate data and the terminal current change rate data of each hardware on the server within a preset time period into a pre-trained fuzzy inference model, wherein the fuzzy inference model is trained based on a hybrid meta-heuristic optimization algorithm; The fuzzy inference model obtains the predicted temperature value of the hardware according to the processed information amount change rate data and the terminal current change rate data; When the difference between the predicted temperature value and the actual temperature value of the hardware is greater than a preset temperature difference threshold, marking the hardware as abnormal hardware; Generate warning information corresponding to the abnormal hardware.

2. The method according to claim 1, characterized in that The training steps of the fuzzy inference model include: Collect training samples to obtain a training sample data set, wherein the training samples include data on the rate of change of the amount of processed information, terminal current value data, and measured temperature value of each hardware on the server within a preset time period; Inputting the training sample data set into the fuzzy inference model to be trained; Using a fuzzy clustering method, initializing parameters of the fuzzy inference model to be trained according to the training samples to obtain an initial fuzzy inference model; According to the parameters of the initial fuzzy inference model, setting the initial parameters of the hybrid meta-heuristic optimization algorithm; The parameters of the initial fuzzy reasoning model are optimized by the hybrid meta-heuristic optimization algorithm to obtain the trained fuzzy reasoning model.

3. The method according to claim 2, characterized in that The method further comprises: According to the preset temperature difference threshold setting rule, set the corresponding temperature difference threshold.

4. The method according to claim 3, characterized in that The step of setting the corresponding temperature difference threshold according to a preset temperature difference threshold setting rule includes: Determining a root mean square error between a predicted temperature value obtained by the fuzzy inference model and a corresponding measured temperature value; According to the root mean square error, the corresponding temperature difference threshold is set.

5. The method according to claim 2, characterized in that: The fuzzy clustering method is used to initialize the parameters of the fuzzy inference model to be trained according to the training samples to obtain an initial fuzzy inference model, including: Randomly selecting a first preset number of initial cluster centers from the training samples; According to a preset optimization target, the initial cluster center is iteratively optimized to obtain an optimized cluster center and corresponding cluster parameters; According to the optimized cluster centers, clustering is performed to generate a corresponding network structure; The network parameters of the fuzzy reasoning model are assigned values ​​according to the clustering parameters to obtain the initial fuzzy reasoning model.

6. The method according to claim 5, characterized in that The optimization goal is to minimize the weighted sum of the distance from the data point in the training sample to the cluster center corresponding to the data point in the training sample and the degree of membership.

7. The method according to claim 2, characterized in that The step of setting the initial parameters of the hybrid meta-heuristic optimization algorithm according to the parameters of the initial fuzzy inference model comprises: Determining the number of particles in the particle swarm in the hybrid meta-heuristic optimization algorithm according to the number of parameters of the initial fuzzy inference model; According to the parameter value of the initial fuzzy inference model, a particle parameter value corresponding to each particle in the particle swarm is determined.

8. The method according to claim 7, characterized in that The method further comprises: Determining a numerical range of a particle parameter value of each particle in the particle swarm according to a parameter value of the initial fuzzy inference model; According to the numerical range, each particle in the particle group is randomly assigned a value.

9. The method according to claim 2, characterized in that: The step of optimizing the parameters of the initial fuzzy reasoning model by the hybrid meta-heuristic optimization algorithm to obtain the trained fuzzy reasoning model includes: Determining parameters of each particle in the particle swarm in the hybrid meta-heuristic optimization algorithm; Iteratively updating the parameters of the particle swarm through a preset fitness function to obtain result particles; The parameters of the result particles are used as the parameters of the fuzzy inference model to obtain the trained fuzzy inference model.

10. The method according to claim 9, characterized in that The method of iteratively updating the parameters of the particle swarm through a preset fitness function to obtain result particles includes: Determining the fitness value of each particle in the particle swarm by using a preset fitness function; When the fitness value does not meet the iteration stopping criterion, using the optimized particle swarm algorithm to update the particle positions in the particle swarm; Updating the particle swarm through a genetic algorithm to obtain an updated particle swarm; Determining the fitness value of each particle in the updated particle swarm through the fitness function; The above steps are iteratively performed, and when the fitness value meets a preset iteration stop criterion, the result particle is obtained.

11. The method according to claim 10, characterized in that The updating of the particle swarm by the genetic algorithm to obtain an updated particle swarm includes: By selecting an operator, a crossover operator and a mutation operator, each particle in the particle swarm is updated to obtain a new particle swarm; Determining the fitness value of each particle in the new particle swarm by using the fitness function; The above steps are iteratively performed to obtain the updated particle swarm when the fitness value satisfies the preset stopping rule.

12. The method according to claim 11, characterized in that The updating of each particle in the particle swarm by selecting an operator, a crossover operator and a mutation operator to obtain a new particle swarm comprises: Selecting a plurality of particles in the particle group by the selection operator to obtain a plurality of target particles; By using the crossover operator, performing crossover calculations on the plurality of target particles in pairs to obtain crossover particles; The crossover particles are placed in the particle group to replace the original particles, so as to obtain a preliminarily updated particle group; The initially updated particle swarm is mutated by the mutation operator to obtain the new particle swarm.

13. A method according to any one of claims 9 to 12, characterized in that: The fitness function is used to determine the root mean square error corresponding to each particle in the particle swarm.

14. The method according to claim 1, characterized in that The method further comprises: The abnormal information corresponding to the abnormal hardware is stored in the background log.

15. The method according to claim 1, characterized in that The method further comprises: The warning information is sent to the corresponding remote address in a preset warning information sending method.

16. The method according to claim 1, characterized in that The method further comprises: Putting the processed information amount change rate data and the terminal current change rate data of each hardware into a training set used for training the fuzzy inference model to obtain an updated training set; The updated training set is used to update the parameters of the fuzzy inference model to obtain a fuzzy inference model with updated parameters.

17. The method according to claim 16, characterized in that The method further comprises: Determine the root mean square error between the predicted temperature value obtained by the model inference model and the corresponding measured temperature value; The preset temperature difference threshold is updated according to the root mean square error.

18. A server hardware abnormality warning device, characterized in that: The device comprises: A data input module, used to input the processing information volume change rate data and the terminal current change rate data of each hardware on the server within a preset time period into a pre-trained fuzzy inference model, wherein the fuzzy inference model is trained based on a hybrid meta-heuristic optimization algorithm; A predicted temperature value acquisition module, used for the fuzzy inference model to obtain the predicted temperature value of the hardware according to the processed information amount change rate data and the terminal current change rate data; An abnormal hardware marking module, used for marking the hardware as abnormal hardware when the difference between the predicted temperature value and the actual temperature value of the hardware is greater than a preset temperature difference threshold; The warning information generation module is used to generate warning information corresponding to the abnormal hardware.

19. A non-volatile readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps in the method according to any one of claims 1 to 17 are implemented.

20. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 17 are implemented.

Images