Artificial intelligence-based operating system failure automatic repair method

Abstract

This invention discloses an artificial intelligence-based automatic operating system fault repair method, belonging to the field of automatic fault repair technology. It utilizes a pre-trained fault feature extraction model to perform feature analysis on operating system operation data, generating fault feature vectors. These vectors are input into a fault classifier, which uses a multi-classification algorithm to identify the current fault type. Based on the fault type, primary fault cause tracing is performed to obtain fault occurrence factors. Secondary fault cause tracing is then performed on these factors to obtain fault impact factors. Based on the fault occurrence factors, the fault location is determined, and corresponding repair strategies are matched from a knowledge base. Historical system operation data with correlation is obtained based on the fault impact factors. After performing repair operations, updated real-time system operation data is acquired to calculate system optimization coefficients. A positive trend of the repair strategy is determined based on a preset optimization threshold and updated in the knowledge base, achieving rapid and efficient automatic repair.

Description

Technical Field

[0001] This invention relates to the field of automatic fault repair technology, and in particular to an automatic fault repair method for operating systems based on artificial intelligence. Background Technology

[0002] As enterprises continue to improve their level of informatization, IT infrastructure has become a core component of enterprise operations. As a core tool to ensure the stable operation of enterprise IT infrastructure, IT operation and maintenance system can monitor the usage status of hardware resources such as servers, network devices, and operating systems (such as CPU, memory, and disk) in real time and trigger alarm mechanisms through preset thresholds. Traditional IT operations and maintenance methods typically involve building staff notifying the operations and maintenance department, who then need to arrive on-site to determine the type of IT equipment failure before taking targeted measures. However, due to the complex network environment and the distribution and diversity of equipment, maintenance personnel often struggle to respond quickly when dealing with IT equipment in large office buildings. This makes it difficult for operations and maintenance personnel to locate faulty IT equipment and to address nearby equipment that is experiencing simultaneous failures, thus prolonging the fault repair time, affecting the stability of operations and maintenance, and increasing labor costs. To address the aforementioned technical deficiencies, a solution is proposed. Summary of the Invention

[0003] The purpose of this invention is to: locate faults based on fault occurrence factors and match corresponding repair strategies from a knowledge base; obtain relevant historical system operation data based on fault impact factors; obtain updated real-time system operation data after performing repair operations to calculate system optimization coefficients; and determine the positive trend of the repair strategy based on a preset optimization threshold and update it to the knowledge base, so as to achieve fast and efficient automatic repair.

[0004] To achieve the above objectives, the present invention adopts the following technical solution: an artificial intelligence-based automatic operating system fault repair method, comprising the following steps: Step 1: Acquire the operating system running data of the IT operation and maintenance system in real time through the multi-source data acquisition module. The operating system running data includes log files, performance indicators, network traffic and configuration parameters. Use a pre-trained fault feature extraction model to perform feature analysis on the operating system running data to generate fault feature vectors. The fault feature extraction model is built based on a deep neural network and trained using historical fault samples. Step 2: Input the fault feature vector into the fault classifier, identify the current fault type through a multi-classification algorithm. The fault classifier is trained using an ensemble learning algorithm. Based on the fault type, the fault cause is traced at the first level to obtain the fault occurrence factor, and the fault occurrence factor is traced at the second level to obtain the fault impact factor. Step 3: Locate the fault based on the fault occurrence factor and match the corresponding repair strategy from the knowledge base. The knowledge base contains the mapping relationship between fault types and repair scripts and execution constraints. The effectiveness of the repair strategy is verified in a sandbox environment through an automated execution engine. When the verification is successful, the repair operation in the repair strategy is executed in the production environment. Step 4: Obtain relevant historical system operation data based on fault impact factors, and obtain updated real-time system operation data after performing repair operations. Compare the historical system operation data with the real-time system operation data to calculate the system optimization coefficient. Determine the positive trend of the repair strategy based on the preset optimization threshold and update it to the knowledge base.

[0005] Furthermore, the specific process for generating fault feature vectors is as follows: S101. Obtain historical datasets under standard operating conditions and historical datasets under various fault conditions. Use an autoencoder structure to reduce the dimensionality of the historical datasets, extract key features, and integrate them into training samples. Divide the generated training samples into training, validation, and test sets in a 7:2:1 ratio. Construct a fault feature extraction model based on a deep neural network. S102. Download the weight file and load it onto the corresponding network to initialize the migration network parameters; S103. Modify the last fully connected layer of the network, keep the input unchanged, set the output to the fault feature vector, initialize the weights of the last layer, use the gradient descent algorithm for learning, and use fixed step size decay to optimize the training parameters. Retrain the entire network to obtain the fault feature extraction model. S104. During the training process, small batches of samples are randomly and non-repeatedly drawn from the training set for training. One training cycle is completed after all training samples are drawn. The training is completed after a certain number of cycles, and the fault feature extraction model is obtained. S105. Obtain real-time operating system data of the IT operations and maintenance system, and input it into the fault feature extraction model to obtain the fault feature vector.

[0006] Furthermore, the fault classifier adopts a hybrid model of gradient boosting decision tree and long short-term memory network. The gradient boosting decision tree is used to process structured feature data, and the long short-term memory network is used to process temporal feature data. The output results of the gradient boosting decision tree and the long short-term memory network are integrated by a weighted fusion module to obtain the final current fault type.

[0007] Furthermore, the specific process for obtaining the failure occurrence factor is as follows: S201. Obtain the fault feature vector, parse the fault location based on the fault feature vector, and obtain the current fault type corresponding to the fault feature vector; S202. Construct a fault propagation tree based on the qualitative reasoning results of the causal graph. Take the fault location as the root node and trace the possible impact path of the fault occurrence upwards. Mark the fault probability nodes on the possible impact path and calculate the fault factor probability on each fault probability node. S203. Use a Bayesian network to quantify the probability dependencies between fault probability nodes to obtain the fault factor probability corresponding to the optimized fault probability node. According to the preset fault probability judgment threshold, mark the fault probability nodes that exceed the fault probability judgment threshold as fault occurrence factors. After judging each fault probability node one by one, integrate and output the fault occurrence factor.

[0008] Furthermore, the specific process for obtaining the fault impact factor is as follows: S301. Obtain the fault occurrence factors to determine the time, environmental conditions and fault manifestations of the fault, and define the fault scope according to the current fault type to clarify the system components, functional modules and business processes affected by the fault. S302. Set up a fault simulation environment based on the time of the fault occurrence, environmental conditions, and fault manifestations. At the same time, construct a causal chain from the cause of the fault to the fault phenomenon based on the system components, functional modules, and business processes affected by the fault, and clarify the logical relationship between each fault occurrence factor. S303. Set multiple sets of fault occurrence conditions in the fault simulation environment, obtain simulation operation data, extract simulation feature vectors through the fault feature extraction model, calculate the similarity between the simulation feature vectors and the fault feature vectors, and when the similarity is greater than or equal to the similarity threshold, the current fault occurrence condition is the fault influence factor.

[0009] Furthermore, the sandbox environment verification includes the following steps: S401. Build a virtualized test environment isolated from the production environment; S402. Simulate fault scenarios and execute repair scripts in a test environment; S403. Evaluate the repair effect through a preset set of verification indicators, which includes service availability, performance recovery rate and resource utilization. When all data in the verification indicator set are greater than the preset standard indicator set, the repair strategy is verified.

[0010] Furthermore, the specific process for calculating the system optimization coefficients and determining the positive trend of the repair strategy is as follows: S501. Obtain relevant historical system operation data based on fault impact factors. The historical system operation data includes log files, monitoring data, and user feedback. Select relevant historical items for fault impact factors and mark them as data to be verified, Gi. S502. After performing the repair operation, obtain the updated real-time system operation data, and select the verifiable data Hi of the same type as the data to be verified in the real-time system operation data. S503. Calculate the system optimization coefficients according to the following formula. , where i = 1, 2, 3, ..., n, n is the number of data items to be verified and verifiable data, and the system optimization coefficient is used to evaluate the degree of optimization of the operating system after performing the repair operation; S504. Obtain the preset optimization threshold. If the system optimization coefficient is greater than or equal to the optimization threshold, the repair strategy is a positive trend. If the system optimization coefficient is less than the optimization threshold, the repair strategy will be a negative trend.

[0011] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are: This AI-based automatic operating system fault repair method utilizes a pre-trained fault feature extraction model to perform feature analysis on operating system operation data, generating fault feature vectors. These vectors are then input into a fault classifier, which uses a multi-classification algorithm to identify the current fault type. Based on the fault type, the method performs primary source tracing to obtain fault occurrence factors, followed by secondary source tracing to obtain fault impact factors. The method then locates the fault based on these factors and matches corresponding repair strategies from a knowledge base. Furthermore, it retrieves relevant historical system operation data based on the fault impact factors and, after performing repair operations, obtains updated real-time system operation data to calculate system optimization coefficients. Finally, it determines the positive trend of the repair strategy based on a preset optimization threshold and updates the knowledge base accordingly, achieving rapid and efficient automatic repair. Attached Figure Description

[0012] Figure 1 A schematic diagram of the overall structure of the present invention is shown. Detailed Implementation

[0013] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0014] Example: like Figure 1 As shown, the artificial intelligence-based automatic operating system fault repair method includes the following steps: Step 1: Acquire the operating system running data of the IT operation and maintenance system in real time through the multi-source data acquisition module. The operating system running data includes log files, performance indicators, network traffic and configuration parameters. Use a pre-trained fault feature extraction model to perform feature analysis on the operating system running data to generate fault feature vectors. The fault feature extraction model is built based on a deep neural network and trained using historical fault samples. The specific process for generating fault feature vectors is as follows: S101. Obtain historical datasets under standard operating conditions and historical datasets under various fault conditions. Use an autoencoder structure to reduce the dimensionality of the historical datasets, extract key features, and integrate them into training samples. Divide the generated training samples into training, validation, and test sets in a 7:2:1 ratio. Construct a fault feature extraction model based on a deep neural network. S102. Download the weight file and load it onto the corresponding network to initialize the migration network parameters; S103. Modify the last fully connected layer of the network, keep the input unchanged, set the output to the fault feature vector, initialize the weights of the last layer, use the gradient descent algorithm for learning, and use fixed step size decay to optimize the training parameters. Retrain the entire network to obtain the fault feature extraction model. S104. During the training process, small batches of samples are randomly and non-repeatedly drawn from the training set for training. One training cycle is completed after all training samples are drawn. The training is completed after a certain number of cycles, and the fault feature extraction model is obtained. S105. Obtain real-time operating system data of the IT operations and maintenance system, and input it into the fault feature extraction model to obtain the fault feature vector.

[0015] Step 2: Input the fault feature vector into the fault classifier, identify the current fault type through a multi-classification algorithm. The fault classifier is trained using an ensemble learning algorithm. Based on the fault type, the fault cause is traced at the first level to obtain the fault occurrence factor, and the fault occurrence factor is traced at the second level to obtain the fault impact factor. The fault classifier adopts a hybrid model of gradient boosting decision tree and long short-term memory network. The gradient boosting decision tree is used to process structured feature data, and the long short-term memory network is used to process temporal feature data. The output results of the gradient boosting decision tree and the long short-term memory network are integrated by a weighted fusion module to obtain the final current fault type.

[0016] The specific process for obtaining the fault occurrence factor is as follows: S201. Obtain the fault feature vector, parse the fault location based on the fault feature vector, and obtain the current fault type corresponding to the fault feature vector; S202. Construct a fault propagation tree based on the qualitative reasoning results of the causal graph. Take the fault location as the root node and trace the possible impact path of the fault occurrence upwards. Mark the fault probability nodes on the possible impact path and calculate the fault factor probability on each fault probability node. S203. Use a Bayesian network to quantify the probability dependencies between fault probability nodes to obtain the fault factor probability corresponding to the optimized fault probability node. According to the preset fault probability judgment threshold, mark the fault probability nodes that exceed the fault probability judgment threshold as fault occurrence factors. After judging each fault probability node one by one, integrate and output the fault occurrence factor.

[0017] The specific process for obtaining the fault impact factor is as follows: S301. Obtain the fault occurrence factors to determine the time, environmental conditions and fault manifestations of the fault, and define the fault scope according to the current fault type to clarify the system components, functional modules and business processes affected by the fault. S302. Set up a fault simulation environment based on the time of the fault occurrence, environmental conditions, and fault manifestations. At the same time, construct a causal chain from the cause of the fault to the fault phenomenon based on the system components, functional modules, and business processes affected by the fault, and clarify the logical relationship between each fault occurrence factor. S303. Set multiple sets of fault occurrence conditions in the fault simulation environment, obtain simulation operation data, extract simulation feature vectors through the fault feature extraction model, calculate the similarity between the simulation feature vectors and the fault feature vectors, and when the similarity is greater than or equal to the similarity threshold, the current fault occurrence condition is the fault influence factor.

[0018] Step 3: Locate the fault based on the fault occurrence factor and match the corresponding repair strategy from the knowledge base. The knowledge base contains the mapping relationship between fault types and repair scripts and execution constraints. The effectiveness of the repair strategy is verified in a sandbox environment through an automated execution engine. When the verification is successful, the repair operation in the repair strategy is executed in the production environment. The sandbox environment verification includes the following steps: S401. Build a virtualized test environment isolated from the production environment; S402. Simulate fault scenarios and execute repair scripts in a test environment; S403. Evaluate the repair effect through a preset set of verification indicators, which includes service availability, performance recovery rate and resource utilization. When all data in the verification indicator set are greater than the preset standard indicator set, the repair strategy is verified.

[0019] Step 4: Obtain relevant historical system operation data based on fault impact factors, and obtain updated real-time system operation data after performing repair operations. Compare the historical system operation data with the real-time system operation data to calculate the system optimization coefficient. Determine the positive trend of the repair strategy based on the preset optimization threshold and update it to the knowledge base.

[0020] The specific process for calculating the system optimization coefficients and determining the positive trend of the repair strategy is as follows: S501. Obtain relevant historical system operation data based on fault impact factors. The historical system operation data includes log files, monitoring data, and user feedback. Select relevant historical items for fault impact factors and mark them as data to be verified, Gi. S502. After performing the repair operation, obtain the updated real-time system operation data, and select the verifiable data Hi of the same type as the data to be verified in the real-time system operation data. S503. Calculate the system optimization coefficients according to the following formula. , where i = 1, 2, 3, ..., n, n is the number of data items to be verified and verifiable data, and the system optimization coefficient is used to evaluate the degree of optimization of the operating system after performing the repair operation; S504. Obtain the preset optimization threshold. If the system optimization coefficient is greater than or equal to the optimization threshold, the repair strategy is a positive trend. If the system optimization coefficient is less than the optimization threshold, the repair strategy will be a negative trend.

[0021] This invention utilizes a pre-trained fault feature extraction model to perform feature analysis on system operation data, generating fault feature vectors. These vectors are then input into a fault classifier, where a multi-classification algorithm identifies the current fault type. Based on the fault type, a primary source tracing is performed to obtain fault occurrence factors. Secondary source tracing is then performed on these factors to obtain fault impact factors. Based on these fault occurrence factors, the fault location is determined, and corresponding repair strategies are matched from a knowledge base. Historical system operation data with correlations is obtained based on the fault impact factors. After performing repair operations, updated real-time system operation data is acquired to calculate system optimization coefficients. A positive trend of the repair strategy is determined based on a preset optimization threshold, and the data is updated in the knowledge base to achieve rapid and efficient automatic repair.

[0022] The size of the interval and threshold is set to facilitate comparison. The size of the threshold depends on the amount of sample data and the number of bases set by those skilled in the art for each set of sample data; as long as it does not affect the ratio between the parameter and the quantized value.

[0023] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation. The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. An automatic fault repair method for operating systems based on artificial intelligence, characterized in that, Includes the following steps: Step 1: Acquire the operating system running data of the IT operation and maintenance system in real time through the multi-source data acquisition module. The operating system running data includes log files, performance indicators, network traffic and configuration parameters. Use a pre-trained fault feature extraction model to perform feature analysis on the operating system running data to generate fault feature vectors. The fault feature extraction model is built based on a deep neural network and trained using historical fault samples. Step 2: Input the fault feature vector into the fault classifier, identify the current fault type through a multi-classification algorithm. The fault classifier is trained using an ensemble learning algorithm. Based on the fault type, the fault cause is traced at the first level to obtain the fault occurrence factor, and the fault occurrence factor is traced at the second level to obtain the fault impact factor. Step 3: Locate the fault based on the fault occurrence factor and match the corresponding repair strategy from the knowledge base. The knowledge base contains the mapping relationship between fault types and repair scripts and execution constraints. The effectiveness of the repair strategy is verified in a sandbox environment through an automated execution engine. When the verification is successful, the repair operation in the repair strategy is executed in the production environment. Step 4: Obtain relevant historical system operation data based on fault impact factors, and obtain updated real-time system operation data after performing repair operations. Compare the historical system operation data with the real-time system operation data to calculate the system optimization coefficient. Determine the positive trend of the repair strategy based on the preset optimization threshold and update it to the knowledge base.

2. The automatic operating system fault repair method based on artificial intelligence according to claim 1, characterized in that, The specific process for generating fault feature vectors is as follows: S101. Obtain historical datasets under standard operating conditions and historical datasets under various fault conditions; and use an autoencoder structure to reduce the dimensionality of the data in the historical datasets, extract key features, and integrate them into training samples. According to the ratio of 7:2:1, the generated training samples are divided into training set, validation set and test set, and a fault feature extraction model is built based on a deep neural network. S102. Download the weight file and load it onto the corresponding network to initialize the migration network parameters; S103. Modify the last fully connected layer of the network, keep the input unchanged, set the output to the fault feature vector, initialize the weights of the last layer, use the gradient descent algorithm for learning, and use fixed step size decay to optimize the training parameters. Retrain the entire network to obtain the fault feature extraction model. S104. During the training process, small batches of samples are randomly and non-repeatedly drawn from the training set for training. One training cycle is completed after all training samples are drawn. The iteration cycle continues until the training is completed, and the fault feature extraction model is obtained. S105. Obtain real-time operating system data of the IT operations and maintenance system, and input it into the fault feature extraction model to obtain the fault feature vector.

3. The automatic operating system fault repair method based on artificial intelligence according to claim 1, characterized in that, The fault classifier adopts a hybrid model of gradient boosting decision tree and long short-term memory network. The gradient boosting decision tree is used to process structured feature data, and the long short-term memory network is used to process temporal feature data. The output results of the gradient boosting decision tree and the long short-term memory network are integrated by a weighted fusion module to obtain the final current fault type.

4. The automatic operating system fault repair method based on artificial intelligence according to claim 1, characterized in that, The specific process for obtaining the fault occurrence factor is as follows: S201. Obtain the fault feature vector, parse the fault location based on the fault feature vector, and obtain the current fault type corresponding to the fault feature vector; S202. Construct a fault propagation tree based on the qualitative reasoning results of the causal graph. Take the fault location as the root node and trace the possible impact path of the fault occurrence upwards. Mark the fault probability nodes on the possible impact path and calculate the fault factor probability on each fault probability node. S203. Use a Bayesian network to quantify the probability dependencies between fault probability nodes to obtain the fault factor probability corresponding to the optimized fault probability node. According to the preset fault probability judgment threshold, mark the fault probability nodes that exceed the fault probability judgment threshold as fault occurrence factors. After judging each fault probability node one by one, integrate and output the fault occurrence factor.

5. The automatic operating system fault repair method based on artificial intelligence according to claim 1, characterized in that, The specific process for obtaining the fault impact factor is as follows: S301. Obtain the fault occurrence factors to determine the time, environmental conditions and fault manifestations of the fault, and define the fault scope according to the current fault type to clarify the system components, functional modules and business processes affected by the fault. S302. Set up a fault simulation environment based on the time of the fault occurrence, environmental conditions, and fault manifestations. At the same time, construct a causal chain from the cause of the fault to the fault phenomenon based on the system components, functional modules, and business processes affected by the fault, and clarify the logical relationship between each fault occurrence factor. S303. Set multiple sets of fault occurrence conditions in the fault simulation environment, obtain simulation operation data, extract simulation feature vectors through the fault feature extraction model, calculate the similarity between the simulation feature vectors and the fault feature vectors, and when the similarity is greater than or equal to the similarity threshold, the current fault occurrence condition is the fault influence factor.

6. The automatic operating system fault repair method based on artificial intelligence according to claim 1, characterized in that, The sandbox environment verification includes the following steps: S401. Build a virtualized test environment isolated from the production environment; S402. Simulate fault scenarios and execute repair scripts in a test environment; S403. Evaluate the repair effect through a preset set of verification indicators, which includes service availability, performance recovery rate and resource utilization. When all data in the verification indicator set are greater than the preset standard indicator set, the repair strategy is verified.

7. The automatic operating system fault repair method based on artificial intelligence according to claim 1, characterized in that, The specific process for calculating the system optimization coefficients and determining the positive trend of the repair strategy is as follows: S501. Obtain relevant historical system operation data based on fault impact factors. The historical system operation data includes log files, monitoring data, and user feedback. Select relevant historical items for fault impact factors and mark them as data to be verified, Gi. S502. After performing the repair operation, obtain the updated real-time system operation data, and select the verifiable data Hi of the same type as the data to be verified in the real-time system operation data. S503. Calculate the system optimization coefficient Ut according to the following formula: , where i = 1, 2, 3, ..., n, n is the number of data items to be verified and verifiable data, and the system optimization coefficient is used to evaluate the degree of optimization of the operating system after performing the repair operation; S504. Obtain the preset optimization threshold. If the system optimization coefficient is greater than or equal to the optimization threshold, the repair strategy is a positive trend. If the system optimization coefficient is less than the optimization threshold, the repair strategy will be a negative trend.

Images