Industrial equipment fault diagnosis system and method based on image recognition

Abstract

The invention relates to the technical field of industrial equipment fault diagnosis and intelligent control, and discloses an industrial equipment fault diagnosis system and method based on image recognition, and the system comprises a data collection module which is used for collecting the multi-modal data of industrial equipment, and the multi-modal data comprise equipment state data, environment data and image data; and the data preprocessing module is connected with the data acquisition module and is used for preprocessing the acquired multi-modal data, including smoothing, denoising and anomaly detection processing of the equipment state data, the environment data and the image data. The environment data, the equipment state data and the image data are subjected to weighted fusion through an attention mechanism, so that feature weights of different data sources can be dynamically adjusted, the actual state of the equipment can be reflected more accurately, a feedback control module is integrated, the running state of the equipment is dynamically adjusted according to a real-time fault diagnosis result, and the fault diagnosis accuracy is improved. And the closed-loop design realizes self-adaptive adjustment of the equipment state, so that the fault risk is effectively reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of industrial equipment fault diagnosis and intelligent control, and in particular to an industrial equipment fault diagnosis system and method based on image recognition. Background Art

[0002] The widespread and complex application of industrial equipment in the production process, the safe and stable operation of equipment is of great significance to industrial production efficiency and safety. However, due to the complex structure of industrial equipment, changeable environmental conditions, and various types and causes of failures, traditional fault diagnosis methods are difficult to meet the real-time and accuracy requirements. Traditional industrial equipment fault diagnosis solutions usually rely on manual experience or models from a single data source. Such solutions can only diagnose common faults to a certain extent, but the data source is single and the fault model is fixed, which makes it impossible to effectively respond to changing equipment conditions and fault types in complex industrial environments.

[0003] Comparative document CN118709076A discloses a method for fault diagnosis of industrial equipment in an electric power system, which includes obtaining multimodal data information of existing industrial equipment in an electric power system and preprocessing to construct a training data set; constructing an initial model for fault diagnosis of industrial equipment in an electric power system and training to obtain a fault diagnosis model for industrial equipment in an electric power system; and using the fault diagnosis model for industrial equipment in an electric power system to perform actual fault diagnosis of industrial equipment in an electric power system.

[0004] The diagnostic result output is based on a fixed model, and there is no real-time adjustment mechanism based on diagnostic feedback, which makes it difficult to adaptively adjust the equipment status, resulting in low adaptability to complex equipment status changes; there is no mention of the design of fault type priority sorting and emergency response, so when a high-risk fault occurs, it is impossible to achieve rapid processing and risk reduction effects; the intelligent alarm and self-learning mechanism are not implemented, and the fault identification model cannot be optimized based on the equipment's historical data and real-time feedback. When a high-risk fault is detected, maintenance personnel cannot be notified in time, increasing the risk of equipment operation.

[0005] Therefore, those skilled in the art provide an industrial equipment fault diagnosis system and method based on image recognition to solve the problems raised in the above background technology. Summary of the invention

[0006] In view of the deficiencies in the prior art, the present invention provides an industrial equipment fault diagnosis system and method based on image recognition to solve the problems raised in the above background technology.

[0007] To achieve the above objectives, the present invention is implemented through the following technical solutions: an industrial equipment fault diagnosis system based on image recognition, comprising: Data acquisition module, used to collect multimodal data of industrial equipment, including equipment status data, environmental data and image data; The data preprocessing module is connected to the data acquisition module to preprocess the collected multimodal data, including smoothing, denoising and anomaly detection of equipment status data, environmental data and image data; The feature extraction and image recognition module is connected to the data preprocessing module and is used to extract features from the preprocessed image data to perform image recognition and fault detection, and generate fault locations and category labels; A multimodal data fusion module connects the feature extraction and image recognition modules, receives the image features of the device, performs feature-level fusion on the image features of the receiving device, generates fused multimodal feature data, and transmits the fused multimodal feature data to the fault diagnosis and classification module to support accurate fault diagnosis; The fault diagnosis and classification module is connected to the multimodal data fusion module to perform fault diagnosis based on the fused multimodal feature data, generate fault type and impact report, and further refine the fault location and category information; The dynamic scheduling and control module is connected to the fault diagnosis and classification module to adjust the equipment operating status and power transmission path in real time according to the diagnosis results to ensure the reliability and safety of the system.

[0008] Preferably, the data acquisition module includes: A sensor system to collect temperature, voltage and current data of the equipment; Meteorological data interface, used to collect real-time environmental information; Image acquisition equipment, used to acquire visible light images and infrared thermal imaging images of the equipment.

[0009] Preferably, the data preprocessing module includes: The smoothing unit is used to perform sliding average processing on the environmental data and equipment status data collected by the data acquisition module to reduce noise interference and smooth the data. The calculation formula for the smoothed data is: , in, represents the smoothed data, is the sliding window size, is the data at time i, and t represents the current time point; The anomaly detection unit performs anomaly detection on the smoothed data based on the three sigma rule. The judgment conditions for anomaly detection are: , in, is the data value at time t, is the mean of the data, is the standard deviation of the data.

[0010] Preferably, the image recognition and fault detection module includes an image recognition model based on YOLOv4, which divides the image into multiple grids, predicts the bounding box and category label of the fault for each grid, and optimizes based on the following loss function: , Among them, S is the size of the grid, B is the number of prediction boxes, and The weight of the loss term, and is the actual position coordinate, and is the predicted location coordinate, For the actual category, is the predicted category.

[0011] Preferably, the multimodal data fusion module includes: A feature extraction unit, used for extracting features from environmental data, device status data and image data; The feature fusion unit uses the attention mechanism to perform weighted fusion of data features from different sources.

[0012] Preferably, the fault diagnosis and classification module includes: A fault classification unit, used to identify the fault type based on the fused multi-modal data; The priority sorting unit is used to prioritize different types of faults and determine the severity and impact scope of the faults.

[0013] Preferably, the dynamic scheduling and control module includes: A feedback control unit, used for real-time feedback control of the operating status of the equipment based on the diagnosis results; Dynamic programming model with the goal of minimizing equipment load and power loss.

[0014] Preferably, the feature extraction unit extracts key features from the environment data, the device status data and the image data to form a feature vector. The specific formula is as follows: Equipment status data feature extraction: Equipment status data is extracted through statistical features such as mean and standard deviation. The original data of the equipment status at time t is , then the state eigenvector It is expressed as: , in, express The former is the mean feature, and the latter is the standard deviation feature.

[0015] Preferably, the feature fusion unit uses an attention mechanism to analyze environmental features. , Equipment status characteristics and image features Perform weighted fusion to generate fused feature vector ; Feature weight calculation: Use the attention mechanism to generate the weight of each feature. Suppose the weight of the environment data feature is , the device status feature weight is , the feature weight of image data is , then the weight calculation formula is: , , , in, , , are trainable weight vectors for environment data, device status data, and image data, , , is the weight value of each feature; Feature fusion: According to the weight of each feature, the environmental data, equipment status data and image data features are weighted and fused to obtain the final multimodal fusion feature : , in, is the fused feature vector.

[0016] An industrial equipment fault diagnosis method based on image recognition comprises the following steps: Step 1: Collect multimodal data of industrial equipment through a data acquisition module; Step 2: Smoothing, denoising and anomaly detection of multimodal data; Step 3: Perform fault detection on the processed image data to generate fault locations and category labels; Step 4: Perform feature-level fusion on the device status data, environment data and image data to generate fusion features; Step 5: Perform fault diagnosis based on fusion features and generate a fault type and impact report; Step 6: Adjust the equipment operating status and power transmission path in real time according to the fault diagnosis results.

[0017] The present invention provides an industrial equipment fault diagnosis system and method based on image recognition. It has the following beneficial effects: 1. The present invention performs weighted fusion of environmental data, equipment status data and image data through the attention mechanism, so that the feature weights of different data sources can be dynamically adjusted to more accurately reflect the actual status of the equipment, and integrates a feedback control module to dynamically adjust the equipment operation status according to the real-time fault diagnosis results. Its closed-loop design realizes adaptive adjustment of the equipment status and effectively reduces the risk of failure.

[0018] 2. The present invention adopts a rolling optimization algorithm to dynamically schedule power transmission paths and equipment loads, optimizes paths under real-time fluctuating equipment operating conditions, ensures the stability and reliability of power transmission, and designs a self-learning mechanism that can optimize fault identification and diagnosis models based on the equipment's historical operating data and real-time feedback, so that the system can adapt to changes in different environments and equipment states, and improve adaptability to new fault types.

[0019] 3. The invention manages different fault types in a hierarchical manner through a priority sorting module, and can give priority to and trigger an emergency response when a serious fault is detected. Its hierarchical response mechanism helps to reduce the risk of equipment operation, and uses a dynamic programming model to optimize the power transmission path and equipment load to minimize power loss and prevent equipment overload. At the same time, the intelligent alarm module automatically sends an alarm when a high-risk fault is detected, helping maintenance personnel to deal with problems in a timely manner and improving the safety and maintenance efficiency of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 It is a system framework diagram of the present invention; Figure 2 It is a flow chart of the present invention. DETAILED DESCRIPTION

[0021] In order to make the technical personnel in the technical field understand the scheme of the present invention, the technical scheme in the embodiment of the present invention will be clearly and completely described below in combination with the drawings in the embodiment of the present invention. Obviously, the described embodiment is a partial embodiment of the present invention, not a complete embodiment. Based on the embodiment of the present invention, other embodiments obtained by ordinary technicians in the field without creative work should fall within the scope of protection of the present invention.

[0022] The present invention is described in detail below in conjunction with the accompanying drawings: Example: Please see attached Figure 1 and attached Figure 2 , an embodiment of the present invention provides an industrial equipment fault diagnosis system based on image recognition, comprising: Data acquisition module, used to collect multimodal data of industrial equipment, including equipment status data, environmental data and image data; The data preprocessing module is connected to the data acquisition module to preprocess the collected multimodal data, including smoothing, denoising and anomaly detection of equipment status data, environmental data and image data; The feature extraction and image recognition module is connected to the data preprocessing module and is used to extract features from the preprocessed image data to perform image recognition and fault detection, and generate fault locations and category labels; A multimodal data fusion module connects the feature extraction and image recognition modules, receives the image features of the device, performs feature-level fusion on the image features of the receiving device, generates fused multimodal feature data, and transmits the fused multimodal feature data to the fault diagnosis and classification module to support accurate fault diagnosis; The fault diagnosis and classification module is connected to the multimodal data fusion module to perform fault diagnosis based on the fused multimodal feature data, generate fault type and impact report, and further refine the fault location and category information; The dynamic scheduling and control module is connected to the fault diagnosis and classification module to adjust the equipment operating status and power transmission path in real time according to the diagnosis results to ensure the reliability and safety of the system.

[0023] The benefits of the data acquisition module are that it can more fully reflect the operating status of the equipment, which helps to improve the accuracy and diagnostic precision of the system; The benefits of the data preprocessing module are to effectively reduce the interference of data noise and abnormal data on the diagnosis results, so that the system can obtain cleaner and more stable data input, thereby improving the reliability of subsequent fault detection and diagnosis; The benefits of the image recognition and fault detection module are that it can quickly and accurately identify the appearance faults of the equipment, realize non-contact and non-invasive detection of the equipment, and improve the detection efficiency and applicability of the system; The benefits of the multimodal data fusion module enable the system to fully utilize the characteristics of different data sources, making the diagnosis results more accurate, ensuring that the system can accurately reflect the device status in a complex and changing environment, and enhancing the robustness of the system; Benefits of the fault diagnosis and classification module: Accurately diagnose and classify different types of faults, generate fault type and impact reports, and help maintenance personnel identify high-risk faults more quickly, take appropriate emergency measures, and improve the efficiency and safety of equipment operation and maintenance; The benefits of the dynamic scheduling and control module are to achieve adaptive adjustment, avoid overload or equipment loss due to faults, further ensure the stable operation of the system, and improve the service life and energy efficiency of the equipment.

[0024] The data acquisition module includes: A sensor system to collect temperature, voltage and current data of the equipment; Meteorological data interface, used to collect real-time environmental information; Image acquisition equipment, used to acquire visible light images and infrared thermal imaging images of the equipment.

[0025] Benefits of the sensor system Accurately monitor the operating status and load conditions of the equipment, prevent potential equipment failures, and help achieve accurate status monitoring and fault prediction; The benefits of the meteorological data interface can better understand the working conditions of the equipment, optimize fault diagnosis and equipment scheduling strategies, and enhance the environmental adaptability and diagnostic accuracy of the system; The benefits of image acquisition equipment include achieving contactless detection of equipment, improving the comprehensiveness and accuracy of fault identification, and providing multi-dimensional information support for timely detection and prevention of faults.

[0026] The data preprocessing module includes: The smoothing unit is used to perform sliding average processing on the environmental data and equipment status data collected by the data acquisition module to reduce noise interference and smooth the data. The calculation formula for the smoothed data is: , in, represents the smoothed data, is the sliding window size, is the data at time i, and t represents the current time point; The anomaly detection unit performs anomaly detection on the smoothed data based on the three sigma rule. The judgment conditions for anomaly detection are: , in, is the data value at time t, is the mean of the data, is the standard deviation of the data.

[0027] The benefits of smoothing units are to reduce short-term fluctuations in data, remove noise from environmental data and equipment status data, make data more stable, and the system can diagnose based on stable signals, reduce misjudgments caused by data fluctuations, and improve the accuracy and reliability of diagnosis, ensuring that subsequent analysis is more accurate; The anomaly detection unit can quickly and effectively detect extreme data points, such as abnormal conditions of temperature, pressure or current of equipment. Real-time detection can help the system identify potential faults in advance, reduce system misjudgments due to data anomalies, and provide a solid data foundation for fault warning and early maintenance.

[0028] The image recognition and fault detection module includes an image recognition model based on YOLOv4. The model divides the image into multiple grids, predicts the bounding box and category label of the fault for each grid, and optimizes based on the following loss function: , Among them, S is the size of the grid, B is the number of prediction boxes, and The weight of the loss term, and is the actual position coordinate, and is the predicted location coordinate, For the actual category, is the predicted category.

[0029] The benefits of the YOLOv4 loss function balance the error between the actual position and the predicted position, ensuring that the model focuses on the position accuracy of the bounding box during training, so that the model can more accurately locate the specific area of ​​the equipment failure and improve the accuracy of fault detection. The model can accurately classify different fault types and more effectively distinguish fault categories, support subsequent fault classification and priority management, increase the system's fault detection coverage, identify faults in multiple locations on the surface of complex equipment, enhance the comprehensiveness of the system's detection, and optimize the model's performance in different detection scenarios, further enhancing the model's focus on precise positioning.

[0030] The multimodal data fusion module includes: A feature extraction unit, used for extracting features from environmental data, device status data and image data; The feature fusion unit uses the attention mechanism to perform weighted fusion of data features from different sources.

[0031] The feature extraction unit can extract key features from various data sources, so that the system can obtain multi-dimensional information such as environmental parameters, equipment operating status and visual features, providing comprehensive data support for fault diagnosis. The system can better capture various possibilities of equipment abnormalities and improve the sensitivity and accuracy of fault detection; The benefit of the feature fusion unit is that the system can adaptively assign weights according to the importance and relevance of the data source. The system can more accurately focus on the most relevant features for the current fault detection, effectively avoid errors caused by data redundancy or irrelevant features, and enhance the adaptability and reliability of the fault detection system.

[0032] The fault diagnosis and classification module includes: A fault classification unit, used to identify the fault type based on the fused multi-modal data; The priority sorting unit is used to prioritize different types of faults and determine the severity and impact scope of the faults.

[0033] The benefits of the fault classification unit are that it combines the environment, status and visual information to analyze fault characteristics in multiple dimensions and identify different fault types. Through the identification of the classification unit, the system takes specific countermeasures according to different types of faults, effectively improving the accuracy and comprehensiveness of fault diagnosis and providing a clear direction for equipment maintenance and troubleshooting. The benefits of the priority sorting unit are to ensure that high-risk faults are handled first. Through the grading mechanism, the system gives priority to harmful faults when multiple faults occur, thereby effectively reducing the risk of equipment operation. It can generate impact reports to help maintenance personnel more accurately understand the potential impact of faults and optimize operation and maintenance decision-making efficiency and response speed.

[0034] The dynamic scheduling and control module includes: A feedback control unit, used for real-time feedback control of the operating status of the equipment based on the diagnosis results; The dynamic programming model aims to minimize equipment load and power loss, and its objective function is: , in, The set of lines that connect the device. For line The transmission power, , , is the loss coefficient.

[0035] The benefits of the feedback control unit are that it can dynamically adjust the operating status of the equipment according to the real-time fault diagnosis results, and timely feedback and adjust the operating parameters of the equipment to keep the equipment in a safe and stable state. Through feedback control, the system can achieve adaptive adjustment to prevent fault escalation or chain reactions caused by abnormal parameters, reduce equipment downtime and maintenance frequency, and enhance equipment reliability and fault recovery capabilities; Benefits of the dynamic programming model The dynamic programming model aims to minimize equipment load and power loss. By optimizing the power transmission path and load distribution, it can reduce the energy loss and overload risk of the equipment. Through the model, the system can dynamically plan the optimal transmission path when the equipment operating environment changes, while maintaining the efficient operation of the power system. Its optimization plan can effectively extend the life of the equipment, reduce energy consumption, and further improve the operating efficiency of the equipment and the overall energy efficiency of the system.

[0036] The feature extraction unit extracts key features from environmental data, equipment status data, and image data to form a feature vector. The specific formula is as follows: Equipment status data feature extraction: Equipment status data is extracted through statistical features such as mean and standard deviation. The original data of the equipment status at time t is , then the state eigenvector It is expressed as: , in, express The mean of , the former is the mean feature, and the latter is the standard deviation feature; The feature fusion unit uses the attention mechanism to analyze the environmental features. , Equipment status characteristics and image features Perform weighted fusion to generate fused feature vector ; Feature weight calculation: Use the attention mechanism to generate the weight of each feature. Suppose the weight of the environment data feature is , the device status feature weight is , the feature weight of image data is , then the weight calculation formula is: , , , in, , , are trainable weight vectors for environment data, device status data, and image data, , , is the weight value of each feature; Feature fusion: According to the weight of each feature, the environmental data, equipment status data and image data features are weighted and fused to obtain the final multimodal fusion feature : , in, is the fused feature vector.

[0037] Benefits of the equipment status data feature extraction formula Through the feature extraction of mean and standard deviation, the overall trend and volatility of equipment status data can be effectively summarized. The mean feature reflects the average operating status of the equipment at the current moment, while the standard deviation feature captures the fluctuation during the operation of the equipment. This feature extraction can help the system identify whether the operating status of the equipment is stable, detect abnormal fluctuations in time, and warn of potential faults, thereby improving the accuracy and timeliness of fault identification; Benefits of feature weighted calculation formula: Through the attention mechanism, the system can assign weights according to the importance of different features in fault diagnosis. The impact of environmental, state and image features on faults varies in different scenarios. The attention mechanism enables the system to adaptively adjust feature weights and focus on the most critical features in the current diagnosis, thereby improving the accuracy and reliability of diagnosis. Benefits of the feature fusion formula The feature fusion formula performs weighted fusion on the feature vectors of environmental data, equipment status data, and image data to generate a comprehensive multimodal feature vector. The fused vector can make full use of information from various data sources to improve the system's comprehensive understanding of equipment status and its ability to identify fault modes. The weighted fused feature vector can accurately reflect the overall status of the equipment, provide rich data support for fault diagnosis and classification, and significantly improve the diagnostic effect and adaptability of the system.

[0038] An industrial equipment fault diagnosis method based on image recognition comprises the following steps: Step 1: Collect multimodal data of industrial equipment through a data acquisition module; Step 2: Smoothing, denoising and anomaly detection of multimodal data; Step 3: Perform fault detection on the processed image data to generate fault locations and category labels; Step 4: Perform feature-level fusion on the device status data, environment data and image data to generate fusion features; Step 5: Perform fault diagnosis based on fusion features and generate a fault type and impact report; Step 6: Adjust the equipment operating status and power transmission path in real time according to the fault diagnosis results.

[0039] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. An industrial equipment fault diagnosis system based on image recognition, characterized in that: include: Data acquisition module, used to collect multimodal data of industrial equipment, including equipment status data, environmental data and image data; The data preprocessing module is connected to the data acquisition module to preprocess the collected multimodal data, including smoothing, denoising and anomaly detection of equipment status data, environmental data and image data; The feature extraction and image recognition module is connected to the data preprocessing module and is used to extract features from the preprocessed image data to perform image recognition and fault detection, and generate fault location and category labels; A multimodal data fusion module connects the feature extraction and image recognition modules, receives the image features of the device, performs feature-level fusion on the image features of the receiving device, generates fused multimodal feature data, and transmits the fused multimodal feature data to the fault diagnosis and classification module to support accurate fault diagnosis; The fault diagnosis and classification module is connected to the multimodal data fusion module to perform fault diagnosis based on the fused multimodal feature data, generate fault type and impact report, and further refine the fault location and category information; The dynamic scheduling and control module is connected to the fault diagnosis and classification module to adjust the equipment operating status and power transmission path in real time according to the diagnosis results to ensure the reliability and safety of the system.

2. The industrial equipment fault diagnosis system based on image recognition according to claim 1 is characterized in that: The data acquisition module comprises: Sensor system to collect temperature, voltage and current data of the equipment; Meteorological data interface, used to collect real-time environmental information; Image acquisition equipment, used to acquire visible light images and infrared thermal imaging images of the equipment.

3. The industrial equipment fault diagnosis system based on image recognition according to claim 1 is characterized in that: The data preprocessing module comprises: The smoothing unit is used to perform sliding average processing on the environmental data and equipment status data collected by the data acquisition module to reduce noise interference and smooth the data. The calculation formula for the smoothed data is: ， in, represents the smoothed data, is the sliding window size, is the data at time i, and t represents the current time point; The anomaly detection unit performs anomaly detection on the smoothed data based on the three sigma rule. The judgment conditions for anomaly detection are: ， in, is the data value at time t, is the mean of the data, is the standard deviation of the data.

4. The industrial equipment fault diagnosis system based on image recognition according to claim 1 is characterized in that: The image recognition and fault detection module includes an image recognition model based on YOLOv4. The model divides the image into multiple grids, predicts the bounding box and category label of the fault for each grid, and optimizes based on the following loss function: ， Among them, S is the size of the grid, B is the number of prediction boxes, and The weight of the loss term, and is the actual position coordinate, and is the predicted location coordinates, For the actual category, is the predicted category.

5. The industrial equipment fault diagnosis system based on image recognition according to claim 1 is characterized in that: The multimodal data fusion module includes: A feature extraction unit, used for extracting features from environmental data, device status data and image data; The feature fusion unit uses the attention mechanism to perform weighted fusion of data features from different sources.

6. The industrial equipment fault diagnosis system based on image recognition according to claim 1 is characterized in that: The fault diagnosis and classification module includes: A fault classification unit, used to identify the fault type based on the fused multi-modal data; The priority sorting unit is used to prioritize different types of faults and determine the severity and impact scope of the faults.

7. The industrial equipment fault diagnosis system based on image recognition according to claim 1 is characterized in that: The dynamic scheduling and control module includes: A feedback control unit, used for real-time feedback control of the operating status of the equipment based on the diagnosis results; Dynamic programming model with the goal of minimizing equipment load and power loss.

8. The industrial equipment fault diagnosis system based on image recognition according to claim 5 is characterized in that: The feature extraction unit extracts key features from the environment data, device status data and image data to form a feature vector. The specific formula is as follows: Equipment status data feature extraction: Equipment status data is extracted through statistical features such as mean and standard deviation. The original data of the equipment status at time t is , then the state eigenvector It is expressed as: ， in, express The former is the mean feature, and the latter is the standard deviation feature.

9. The industrial equipment fault diagnosis system based on image recognition according to claim 5, characterized in that: The feature fusion unit uses the attention mechanism to analyze the environmental features. , Equipment status characteristics and image features Perform weighted fusion to generate fused feature vector ; Feature weight calculation: Use the attention mechanism to generate the weight of each feature. Suppose the weight of the environment data feature is , the device status feature weight is , the feature weight of image data is , then the weight calculation formula is: ， ， ， in, , , are trainable weight vectors for environment data, device status data, and image data, , , is the weight value of each feature; Feature fusion: According to the weight of each feature, the environmental data, equipment status data and image data features are weighted and fused to obtain the final multimodal fusion feature : ， in, is the fused feature vector.

10. An industrial equipment fault diagnosis method based on image recognition, according to an industrial equipment fault diagnosis system based on image recognition according to any one of claims 1 to 9, characterized in that: The steps include: Step 1: Collect multimodal data of industrial equipment through a data acquisition module; Step 2: Smoothing, denoising and anomaly detection of multimodal data; Step 3: Perform fault detection on the processed image data to generate fault locations and category labels; Step 4: Perform feature-level fusion on the device status data, environment data and image data to generate fusion features; Step 5: Perform fault diagnosis based on fusion features and generate a fault type and impact report; Step 6: Adjust the equipment operating status and power transmission path in real time according to the fault diagnosis results.

Images