Equipment fault diagnosis method and equipment based on artificial intelligence, and medium

Abstract

The invention provides a device fault diagnosis method and device based on artificial intelligence and a medium, and the method comprises the steps: obtaining various data in a device operation process, and carrying out the preprocessing of the data; according to the equipment type and the fault diagnosis requirement, a preset artificial intelligence algorithm is adopted to construct a fault diagnosis model; training the constructed model by using historical fault data and normal operation data, including positive learning and negative learning, generating a training subset for training through a Bootstrap resampling method, and adjusting model parameters to improve the diagnosis accuracy; and inputting equipment data acquired in real time into the trained model, obtaining a fault diagnosis identifier or result, and judging whether the equipment has a fault or not and the fault type. According to the technical scheme provided by the invention, the accuracy, timeliness and intelligent level of equipment fault diagnosis can be effectively improved, and a solid guarantee is provided for stable operation of equipment.

Description

Technical Field

[0001] This document relates to the technical field of equipment fault diagnosis, and in particular to an equipment fault diagnosis method, equipment and medium based on artificial intelligence. Background Art

[0002] With the rapid development of modern industry, equipment continues to grow in complexity and intelligence, and the losses caused by equipment failures are becoming increasingly severe. Traditional equipment fault diagnosis methods, such as manual diagnosis based on experience and those relying on simple physical models, are no longer able to meet the current demand for efficient and accurate equipment operation.

[0003] In complex industrial production environments, equipment operating conditions are influenced by numerous factors, generating data that is multi-source, high-dimensional, and dynamically changing. Various data, including mechanical parameters, operating status data, and alarm information, contain key clues to the equipment's health. However, accurately extracting meaningful information from this massive and complex data presents a major challenge in equipment fault diagnosis.

[0004] At the same time, the rapid development of artificial intelligence (AI) technology has brought new opportunities for equipment fault diagnosis. AI algorithms such as machine learning and deep learning possess powerful data processing and pattern recognition capabilities, enabling analysis and modeling of complex equipment data, thereby enabling intelligent diagnosis of equipment failures. By building an AI-based equipment fault diagnosis system, it is possible to accurately monitor equipment operating status in real time, detect potential faults promptly, and issue early warnings, effectively reducing equipment failure rates and minimizing production interruptions and economic losses caused by these failures.

[0005] However, the current practical application of artificial intelligence for equipment fault diagnosis still faces numerous challenges. For example, the accuracy and reliability of data collection are difficult to guarantee, and multi-sensor data fusion technology is still imperfect. The construction of fault diagnosis models lacks specificity and effectiveness, and diagnostic accuracy needs to be improved. Furthermore, aspects such as parameter optimization and learning strategies during model training require further refinement. Therefore, developing an efficient, accurate, and adaptable artificial intelligence-based equipment fault diagnosis method is of great practical significance. Summary of the Invention

[0006] According to an embodiment of the present invention, a method, device and medium for diagnosing equipment faults based on artificial intelligence are provided, aiming to solve the above-mentioned problems.

[0007] According to an embodiment of the present invention, a device fault diagnosis method based on artificial intelligence is provided, comprising:

[0008] S1. Obtain various data during the operation of the equipment and pre-process the data;

[0009] S2. Based on the equipment type and fault diagnosis requirements, a fault diagnosis model is constructed using a preset artificial intelligence algorithm;

[0010] S3. Use historical fault data and normal operation data to train the constructed model, including positive learning and negative learning. Use the Bootstrap resampling method to generate training subsets for training, and adjust model parameters to improve diagnostic accuracy.

[0011] S4. Input the real-time collected device data into the trained model to obtain the fault diagnosis identification or result, and determine whether the device has a fault and the fault type;

[0012] The various types of data during the operation of the equipment include: mechanical structure parameters of the equipment, operating status data, and alarm information text.

[0013] According to an embodiment of the present invention, there is provided an electronic device, including:

[0014] processor; and,

[0015] A memory is arranged to store computer executable instructions, which, when executed, cause the processor to perform the steps of the above method embodiments.

[0016] According to an embodiment of the present invention, a storage medium is provided for storing computer-executable instructions, wherein the computer-executable instructions implement the steps of the above-mentioned method embodiment when executed.

[0017] The embodiment of the present invention adopts a unified time base, data quality assessment, signal processing and feature layer fusion strategy to ensure data time consistency, screen high-quality data, convert and verify data formats, improve data accuracy and reliability, and provide a better data basis for fault diagnosis; dictionary matching and segmentation of alarm information texts of different lengths help to process text information more efficiently, improve information processing speed and accuracy; positive and negative learning of convolutional neural networks, and dynamically adjust the relationship between the number of learning times, or use Bootstrap resampling method to generate training subsets, can effectively improve the model The ability to identify faults and normal states enhances the generalization performance of the model; deeply integrates support vector machines, long short-term memory networks and random forest algorithms to build a hierarchical hybrid model architecture, and sets up interactive and fusion decision-making mechanisms, which enhances the model's ability to capture complex fault modes and improves the accuracy and reliability of fault diagnosis; by analyzing the source of uncertainty, Monte Carlo simulation and statistical methods, the fault diagnosis results can be evaluated and adjusted to make the diagnosis results more credible; regularly updating the model can adapt to changes in equipment status and new fault types; executing corresponding operations when a fault is diagnosed can timely avoid equipment damage and ensure production safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate one or more embodiments of this specification or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments recorded in this specification. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0019] Figure 1 Flowchart of an artificial intelligence-based equipment fault diagnosis method according to an embodiment of the present invention. DETAILED DESCRIPTION

[0020] In order to enable those skilled in the art to better understand the technical solutions in one or more embodiments of this specification, the technical solutions in one or more embodiments of this specification will be clearly and completely described below in conjunction with the drawings in one or more embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of this specification, not all of the embodiments. Based on one or more embodiments of this specification, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of this document.

[0021] Method Example

[0022] According to an embodiment of the present invention, a device fault diagnosis method based on artificial intelligence is provided. Figure 1 This is a flow chart of an artificial intelligence-based device fault diagnosis method according to an embodiment of the present invention. Figure 1 As shown, the device fault diagnosis method based on artificial intelligence in an embodiment of the present invention specifically includes:

[0023] S1. Acquire various data during the operation of the equipment and pre-process the data; the various data during the operation of the equipment include: mechanical structure parameters of the equipment, operation status data, and alarm information text.

[0024] The acquisition of various types of data during the operation of the equipment adopts multi-sensor fusion technology to simultaneously collect sensor data of different parts and types of equipment, specifically including:

[0025] A unified time base is set for different types of sensors. The clocks of all sensors involved in data collection are synchronized using the IEEE1588 precision time protocol. This ensures that the time error of each sensor's data is within the microsecond level, ensuring that the sensor data collected at the same timestamp is temporally consistent. A data cache mechanism using a first-in-first-out data structure is established to temporarily store the collected multi-sensor data in the cache in chronological order for subsequent unified preprocessing operations.

[0026] A corresponding data quality assessment indicator system is established for different types of sensor data. The assessment indicators for temperature sensor data include data stability and the degree of deviation from historical data. The assessment indicators for vibration sensor data include the signal-to-noise ratio and the rationality of the spectrum. The collected data is screened based on the data quality assessment results, and data whose quality does not meet the preset threshold is marked or eliminated. At the same time, for sensors with missing or abnormal data, an alarm mechanism is triggered to prompt sensor inspection or maintenance.

[0027] For analog sensor data, high-precision analog-to-digital conversion technology is used to convert analog signals into digital signals, low-pass filters are used to remove high-frequency noise, and differential amplifier circuits are used to suppress common-mode interference. For digital sensor data, data format unification and verification are carried out. The output data of digital sensors of different manufacturers and models are converted and stored in a unified data format, and cyclic redundancy check methods are used to ensure data accuracy and integrity.

[0028] A feature layer-based fusion strategy is adopted to fuse the features extracted by different types of sensors. An adaptive weighted fusion algorithm is used to dynamically adjust the weight of each sensor data in the fused data according to the reliability and accuracy of different sensors under different working conditions.

[0029] When preprocessing the data, for the alarm information text, the device information with a character count less than or equal to a specific length is stored in a dictionary, and the device information with a character count greater than the specific length is searched in the dictionary for matching and segmentation.

[0030] S2. Based on the equipment type and fault diagnosis requirements, a fault diagnosis model is constructed using a preset artificial intelligence algorithm;

[0031] When using the preset artificial intelligence algorithm to build a fault diagnosis model, multiple different types of artificial intelligence algorithms are deeply integrated, as follows:

[0032] Screening out suitable AI algorithms for equipment fault diagnosis from a variety of AI algorithms, including but not limited to support vector machines, long short-term memory (LSTM) networks, and random forest algorithms. Support vector machines are used to process and classify high-dimensional data, long short-term memory networks are used to process operating status data with time series characteristics, and random forest algorithms are used to comprehensively consider multiple features for decision-making.

[0033] Build a hierarchical hybrid model architecture, with the bottom layer consisting of multiple independent sub-models, each trained with a selected artificial intelligence algorithm;

[0034] In the hybrid model architecture, an inter-model interaction mechanism is set up to enable the output results of each sub-model to influence and complement each other. Specifically, the output results of one sub-model are used as input features of other sub-models to enhance the model's ability to capture complex failure modes.

[0035] A fusion decision-making mechanism is designed to make a comprehensive decision on the output results of each sub-model. According to the accuracy and reliability of each sub-model in diagnosing different fault types, a corresponding weight is assigned to each sub-model. The output results of each sub-model are fused through weighted summation or voting to obtain the final fault diagnosis result, thereby improving the accuracy and reliability of fault diagnosis.

[0036] S3. Use historical fault data and normal operation data to train the constructed model, including positive learning and negative learning. Use the Bootstrap resampling method to generate training subsets for training, and adjust model parameters to improve diagnostic accuracy.

[0037] When performing a preset number of positive learning and multiple negative learning on the convolutional neural network in the training of the constructed model using the historical fault data and the normal operating data, the second set number of negative learning times is a multiple of the first set number of positive learning times, specifically including:

[0038] A first set number of forward learning times is set in the forward learning stage. In each forward learning, the historical fault data and the normal operation data are divided into multiple data batches according to a preset batch division rule. The data of each batch contains a preset number of samples. The Susonghu samples cover different types of equipment operating states and fault conditions. For each data batch, it is input into the convolutional neural network for forward propagation calculation to obtain the output result of the network. Then, according to the difference between the network output result and the true label, the preset loss function is used to calculate the loss value. Based on the calculated loss value, the back propagation algorithm is used to calculate the gradient of each parameter in the network. Then, the optimization algorithm is used to update the parameters of the convolutional neural network according to the gradient information to reduce the loss value and improve the network's recognition ability for faults and normal states.

[0039] In the negative learning phase, a second set number of negative learning times is determined, and the second set number of negative learning times is an integer multiple of the first set number of positive learning times. In each negative learning phase, a negative sample data set is constructed. The negative sample data is generated by performing specific transformations and disturbances on the normal operating data.

[0040] The negative sample data is input into the convolutional neural network for forward propagation, and the loss value of the negative sample is calculated. The backpropagation algorithm is also used to calculate the gradient corresponding to the negative sample, and the optimization algorithm is used to adjust the network parameters;

[0041] Dynamically adjusting the relationship between the multiples of the number of learning times specifically includes: during the training process, dynamically adjusting the relationship between the multiples of the second set number of negative learning times and the first set number of positive learning times according to the training effect and performance indicators of the convolutional neural network.

[0042] S4. Input the real-time collected device data into the trained model to obtain the fault diagnosis identification or result, and determine whether the device has a fault and the fault type;

[0043] After inputting the real-time collected equipment data into the trained model to obtain the fault diagnosis identification or results, the uncertainty assessment and confidence interval calculation of the fault diagnosis results are also performed. The specific steps are as follows:

[0044] Analyze the sources of uncertainty in fault diagnosis results, including measurement errors in sensor data, information loss during data preprocessing, inadequate model training, and dynamic changes in the equipment operating environment. Quantify each source of uncertainty and determine its impact on the fault diagnosis results.

[0045] The Monte Carlo simulation method is used to simulate the fault diagnosis process multiple times. In each simulation, the input data is randomly perturbed according to the probability distribution of the uncertainty source. The perturbed data is then input into the trained model for diagnosis, and multiple simulated diagnosis results are obtained.

[0046] Based on multiple diagnostic results obtained from Monte Carlo simulation, the confidence interval of the fault diagnosis results is calculated using statistical methods;

[0047] Based on the uncertainty assessment and confidence interval calculation results of the fault diagnosis results, feedback and adjustments are made to the fault diagnosis process.

[0048] The method according to the embodiment of the present invention further includes a model updating step: regularly collecting new equipment operation data and fault data, and updating and training the fault diagnosis model to adapt to changes in equipment operation status and newly occurring fault types.

[0049] The method according to an embodiment of the present invention also includes a result processing step: when a fault is diagnosed in the equipment, corresponding operations are performed according to the type of fault, such as performing an emergency shutdown operation on the industrial equipment, and at the same time displaying fault reminder information through a real-time display device, or sending the fault information to the staff through a wireless communication device.

[0050] S1 is specifically explained by taking a key production equipment of a large factory as an example. The equipment is composed of multiple complex mechanical components.

[0051] Mechanical structure parameter acquisition: Obtain the mechanical structure parameters of the equipment, such as component dimensions, assembly relationships, and material properties, through the equipment's design drawings and related technical documentation. At the same time, use high-precision measuring instruments to measure the actual dimensions of the equipment's key components to ensure data accuracy.

[0052] Operating status data collection: Multiple sensors, such as vibration sensors, temperature sensors, and speed sensors, are installed in key locations such as the equipment's motors, bearings, and transmission components. Multi-sensor fusion technology is used to establish a unified time base for different sensor types. All sensor clocks are synchronized using the IEEE 1588 Precision Time Protocol, ensuring that the time error between sensor data acquisitions is within microseconds. A data caching mechanism with a first-in-first-out (FIFO) structure is implemented to temporarily store collected multi-sensor data in chronological order. A data quality assessment system is established for each type of sensor data. For example, for temperature sensor data, its stability and deviation from historical data are evaluated; for vibration sensor data, the signal-to-noise ratio and spectral rationality of the signal are evaluated. Collected data is screened based on the data quality assessment results, and data that does not meet the preset quality threshold is marked or removed. For sensors with missing or abnormal data, an alarm mechanism is triggered to prompt sensor inspection or repair. For analog sensor data, high-precision analog-to-digital conversion technology is used to convert analog signals into digital signals, low-pass filters are used to remove high-frequency noise, and differential amplifier circuits are used to suppress common-mode interference; for digital sensor data, data format unification and verification are performed, and data output by digital sensors of different manufacturers and models are converted and stored in a unified data format, and a cyclic redundancy check method is used to ensure data accuracy and integrity.

[0053] Alarm Information Text Collection: This function collects alarm information generated by device control systems in real time. Device information with a character count of less than or equal to a specified length (e.g., 50 characters) is stored in a dictionary. Device information with a character count greater than the specified length is then searched through the dictionary for matching and segmentation for subsequent analysis and processing.

[0054] S2 uses a preset artificial intelligence algorithm to build a fault diagnosis model based on the device type and fault diagnosis requirements. Based on the device type and common fault diagnosis requirements, it selects an algorithm suitable for equipment fault diagnosis from a variety of artificial intelligence algorithms. Support vector machines are used to process and classify high-dimensional data, long short-term memory networks (LSTMs) process operating status data with time series characteristics, and random forest algorithms make decisions based on multiple features. A hierarchical hybrid model architecture is constructed, with the bottom layer consisting of multiple independent sub-models, each trained using a selected artificial intelligence algorithm. An inter-model interaction mechanism is established within the hybrid model architecture so that the output results of each sub-model can influence and complement each other. For example, the output results of one sub-model can be used as input features for other sub-models to enhance the model's ability to capture complex fault patterns.

[0055] Based on the model built by S2, model training is performed, including:

[0056] Data preparation: Collect the device's historical failure data and normal operation data from the past year. The historical failure data covers various failure scenarios and corresponding detailed data, while the normal operation data records various data records when the device is in a stable operating state.

[0057] Forward learning: Set a first set number of forward learning times, for example, 100 times. Each time forward learning is performed, the historical fault data and normal operation data are divided into multiple data batches according to the preset batch division rules. Each batch of data contains a preset number (such as 100) samples, and the samples cover different types of equipment operating conditions and fault conditions. For each data batch, it is input into the convolutional neural network for forward propagation calculation to obtain the output result of the network. Then, based on the difference between the network output result and the true label, a preset loss function (such as the cross entropy loss function) is used to calculate the loss value. Based on the calculated loss value, the back propagation algorithm is used to calculate the gradient of each parameter in the network. Then, the optimization algorithm (such as the Adam optimization algorithm) is used to update the parameters of the convolutional neural network according to the gradient information to reduce the loss value and improve the network's ability to recognize faults and normal states.

[0058] Negative learning: Determine the second set number of negative learning times, and the second set number of negative learning times is an integer multiple of the first set number of positive learning times, for example, 200 times. In each negative learning, a negative sample data set is constructed, and the negative sample data is generated by performing specific transformations and perturbations on the normal operating data, such as randomly increasing or decreasing certain characteristic values of the normal operating data by a certain proportion. The negative sample data is input into the convolutional neural network for forward propagation, and the loss value of the negative sample is calculated. The backpropagation algorithm is also used to calculate the gradient corresponding to the negative sample, and the optimization algorithm is used to adjust the network parameters. At the same time, during the training process, the relationship between the second set number of negative learning times and the first set number of positive learning times is dynamically adjusted according to the training effect and performance indicators of the convolutional neural network (such as accuracy, recall rate, etc.).

[0059] During real-time equipment operation, sensors continuously collect real-time equipment data. This data is preprocessed according to pre-defined formats and standards and then fed into a trained fault diagnosis model. The model outputs a fault diagnosis indicator or result, such as "normal operation," "bearing fault," or "motor overheating." Based on this output, equipment management personnel can quickly determine whether a fault exists and what type of fault it is. Once a fault is diagnosed, appropriate actions are taken based on the fault type, such as an emergency shutdown of the industrial equipment. Fault alerts are displayed on real-time displays or transmitted to personnel via wireless communication devices. Furthermore, new equipment operating and fault data are regularly collected to update and train the fault diagnosis model to adapt to changes in equipment operating conditions and emerging fault types. Furthermore, sources of uncertainty in the fault diagnosis results are analyzed, including measurement errors in sensor data, information loss during data preprocessing, inadequate model training, and dynamic changes in the equipment operating environment. Each source of uncertainty is quantified and analyzed to determine its impact on the fault diagnosis results. The fault diagnosis process is simulated multiple times using Monte Carlo simulation. In each simulation, the input data is randomly perturbed based on the probability distribution of the uncertainty source. The perturbed data is then fed into the trained model for diagnosis, resulting in multiple simulated diagnostic results. Based on the multiple diagnostic results obtained from the Monte Carlo simulations, statistical methods are used to calculate confidence intervals for the fault diagnosis results. Feedback and adjustments to the fault diagnosis process are provided based on the uncertainty assessment and confidence interval calculation results.

[0060] By adopting the embodiments of the present invention, the following beneficial effects are achieved:

[0061] The system uses a unified time base, data quality assessment, signal processing, and feature layer fusion strategies to ensure data time consistency, screen high-quality data, and convert and verify data formats, improving data accuracy and reliability and providing a higher-quality data foundation for fault diagnosis. Dictionary matching and segmentation of alarm information texts of varying lengths facilitates more efficient text processing, increasing both speed and accuracy. Convolutional neural networks employ positive and negative learning, dynamically adjusting the learning times, or using bootstrap resampling to generate training subsets, effectively improving the model's ability to recognize faults and normal states and enhancing its generalization performance. Deeply integrating support vector machines, long short-term memory networks, and random forest algorithms to construct a hierarchical hybrid model architecture, with interactive and fusion decision-making mechanisms, enhances the model's ability to capture complex fault patterns and improves the accuracy and reliability of fault diagnosis. By analyzing sources of uncertainty, using Monte Carlo simulation, and statistical methods, fault diagnosis results can be evaluated and adjusted, making them more reliable. Regular model updates adapt to changes in equipment status and new fault types. Executing appropriate actions when a fault is diagnosed can prevent equipment damage and ensure production safety.

[0062] Device Example 1

[0063] According to an embodiment of the present invention, there is provided an electronic device, including:

[0064] processor; and,

[0065] A memory is arranged to store computer executable instructions, which, when executed, cause the processor to perform the steps of the above method embodiments.

[0066] Device Example 2

[0067] According to an embodiment of the present invention, a storage medium is provided for storing computer-executable instructions, wherein the computer-executable instructions implement the steps of the above-mentioned method embodiment when executed.

[0068] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. An artificial intelligence-based equipment fault diagnosis method, characterized in that: The following steps are involved: S1. Obtain various data during the operation of the equipment and pre-process the data; S2. Based on the equipment type and fault diagnosis requirements, a fault diagnosis model is constructed using a preset artificial intelligence algorithm; S3. Use historical fault data and normal operation data to train the constructed model, including positive learning and negative learning. Use the Bootstrap resampling method to generate training subsets for training, and adjust model parameters to improve diagnostic accuracy. S4. Input the real-time collected device data into the trained model to obtain the fault diagnosis identification or result, and determine whether the device has a fault and the fault type; The various types of data during the operation of the equipment include: mechanical structure parameters of the equipment, operating status data and alarm information text.

2. The method according to claim 1, characterized in that The acquisition of various types of data during the operation of the equipment adopts multi-sensor fusion technology to simultaneously collect sensor data of different parts and types of equipment, including: A unified time base is set for different types of sensors. The clocks of all sensors involved in data collection are synchronized using the IEEE1588 precision time protocol. This ensures that the time error of each sensor's data is within the microsecond level, ensuring that the sensor data collected at the same timestamp is temporally consistent. A data cache mechanism using a first-in-first-out data structure is established to temporarily store the collected multi-sensor data in the cache in chronological order for subsequent unified preprocessing operations. A corresponding data quality assessment indicator system is established for different types of sensor data. The assessment indicators for temperature sensor data include data stability and the degree of deviation from historical data. The assessment indicators for vibration sensor data include the signal-to-noise ratio and the rationality of the spectrum. The collected data is screened based on the data quality assessment results, and data whose quality does not meet the preset threshold is marked or eliminated. At the same time, for sensors with missing or abnormal data, an alarm mechanism is triggered to prompt sensor inspection or maintenance. For analog sensor data, high-precision analog-to-digital conversion technology is used to convert analog signals into digital signals, low-pass filters are used to remove high-frequency noise, and differential amplifier circuits are used to suppress common-mode interference. For digital sensor data, data format unification and verification are carried out. The output data of digital sensors of different manufacturers and models are converted and stored in a unified data format, and cyclic redundancy check methods are used to ensure data accuracy and integrity. A feature layer-based fusion strategy is adopted to fuse the features extracted by different types of sensors. An adaptive weighted fusion algorithm is used to dynamically adjust the weight of each sensor data in the fused data according to the reliability and accuracy of different sensors under different working conditions.

3. The artificial intelligence-based equipment fault diagnosis method according to claim 1, characterized in that: When preprocessing the data, for the alarm information text, the device information with a character count less than or equal to a specific length is stored in a dictionary, and the device information with a character count greater than the specific length is searched in the dictionary for matching and segmentation.

4. The artificial intelligence-based equipment fault diagnosis method according to claim 1, characterized in that: When performing a preset number of positive learning and multiple negative learning on the convolutional neural network in the training of the constructed model using the historical fault data and the normal operating data, the second set number of negative learning times is a multiple of the first set number of positive learning times, specifically including: A first set number of forward learning times is set in the forward learning stage. In each forward learning, the historical fault data and the normal operation data are divided into multiple data batches according to a preset batch division rule. The data of each batch contains a preset number of samples. The Susonghu samples cover different types of equipment operating states and fault conditions. For each data batch, it is input into the convolutional neural network for forward propagation calculation to obtain the output result of the network. Then, according to the difference between the network output result and the true label, the preset loss function is used to calculate the loss value. Based on the calculated loss value, the back propagation algorithm is used to calculate the gradient of each parameter in the network. Then, the optimization algorithm is used to update the parameters of the convolutional neural network according to the gradient information to reduce the loss value and improve the network's recognition ability for faults and normal states. In the negative learning phase, a second set number of negative learning times is determined, and the second set number of negative learning times is an integer multiple of the first set number of positive learning times. In each negative learning phase, a negative sample data set is constructed. The negative sample data is generated by performing specific transformations and disturbances on the normal operating data. The negative sample data is input into the convolutional neural network for forward propagation, and the loss value of the negative sample is calculated. The backpropagation algorithm is also used to calculate the gradient corresponding to the negative sample, and the optimization algorithm is used to adjust the network parameters; Dynamically adjusting the relationship between the multiples of the number of learning times specifically includes: during the training process, dynamically adjusting the relationship between the multiples of the second set number of negative learning times and the first set number of positive learning times according to the training effect and performance indicators of the convolutional neural network.

5. The method according to claim 1, characterized in that When using the preset artificial intelligence algorithm to build a fault diagnosis model, multiple different types of artificial intelligence algorithms are deeply integrated, as follows: Screening out suitable AI algorithms for equipment fault diagnosis from a variety of AI algorithms, including but not limited to support vector machines, long short-term memory (LSTM) networks, and random forest algorithms. Support vector machines are used to process and classify high-dimensional data, long short-term memory networks are used to process operating status data with time series characteristics, and random forest algorithms are used to comprehensively consider multiple features for decision-making. Build a hierarchical hybrid model architecture, with the bottom layer consisting of multiple independent sub-models, each trained with a selected artificial intelligence algorithm; In the hybrid model architecture, an inter-model interaction mechanism is set up to enable the output results of each sub-model to influence and complement each other. Specifically, the output results of one sub-model are used as input features of other sub-models to enhance the model's ability to capture complex failure modes. A fusion decision-making mechanism is designed to make a comprehensive decision on the output results of each sub-model. According to the accuracy and reliability of each sub-model in diagnosing different fault types, a corresponding weight is assigned to each sub-model. The output results of each sub-model are fused through weighted summation or voting to obtain the final fault diagnosis result, thereby improving the accuracy and reliability of fault diagnosis.

6. The method according to claim 1, characterized in that After inputting the real-time collected equipment data into the trained model to obtain the fault diagnosis identification or results, the uncertainty assessment and confidence interval calculation of the fault diagnosis results are also performed. The specific steps are as follows: Analyze the sources of uncertainty in fault diagnosis results, including measurement errors in sensor data, information loss during data preprocessing, inadequate model training, and dynamic changes in the equipment operating environment. Quantify each source of uncertainty and determine its impact on the fault diagnosis results. The Monte Carlo simulation method is used to simulate the fault diagnosis process multiple times. In each simulation, the input data is randomly perturbed according to the probability distribution of the uncertainty source. The perturbed data is then input into the trained model for diagnosis, and multiple simulated diagnosis results are obtained. Based on multiple diagnostic results obtained from Monte Carlo simulation, the confidence interval of the fault diagnosis results is calculated using statistical methods; Based on the uncertainty assessment and confidence interval calculation results of the fault diagnosis results, feedback and adjustments are made to the fault diagnosis process.

7. The artificial intelligence-based equipment fault diagnosis method according to claim 1, characterized in that: It also includes a model update step: regularly collect new equipment operation data and fault data, and update and train the fault diagnosis model to adapt to changes in equipment operation status and newly emerging fault types.

8. The artificial intelligence-based equipment fault diagnosis method according to claim 1, characterized in that: It also includes a result processing step: when a fault is diagnosed in the equipment, corresponding operations are performed according to the type of fault, such as performing an emergency shutdown operation on the industrial equipment, and displaying fault reminder information through a real-time display device, or sending the fault information to the staff through a wireless communication device.

9. An electronic device comprising: processor; as well as, A memory arranged to store computer-executable instructions, which, when executed, cause the processor to perform the steps of the artificial intelligence-based equipment fault diagnosis method according to any one of claims 1 to 8.

10. A storage medium for storing computer-executable instructions, wherein the computer-executable instructions, when executed, implement the steps of the artificial intelligence-based equipment fault diagnosis method according to any one of claims 1 to 8.

Images