Intelligent Fault Diagnosis Method and Device for Heat Exchange Modules

Abstract

This invention discloses an intelligent fault diagnosis device for heat exchange modules, belonging to the field of fault diagnosis technology. It includes modules for data acquisition, preprocessing, feature extraction, fault diagnosis modeling, and result output. The data acquisition module uses a multi-sensor network to accurately collect operational data at key locations. The preprocessing module removes noise and outliers, smooths and normalizes the data. The feature extraction module extracts multi-level, multi-dimensional features, including time-domain, frequency-domain, and time-frequency-domain features. The fault diagnosis model employs a hybrid model combining capsule networks and long short-term memory networks. The result output module outputs results in a multimodal manner, with 3D visualization and multi-channel notification. This invention provides accurate diagnosis, reducing false alarms and missed alarms; efficient diagnosis, enabling real-time fault detection; adaptive and intelligent learning capabilities, adapting to different operating conditions; multimodal output facilitates operation and maintenance; and fault prediction functionality ensures stable equipment operation.

Description

Technical Field

[0001] This invention relates to the field of fault diagnosis technology, and in particular to an intelligent fault diagnosis method and device for heat exchange modules. Background Technology

[0002] In industrial production and daily life, heat exchange modules are widely used in air conditioning, refrigeration equipment, power systems, and other fields, and their stable operation is crucial. The following section begins by explaining the shortcomings of existing technologies, then elaborates on the improvements brought about by this patent.

[0003] As a core component for heat exchange, heat exchange modules play a crucial role in numerous fields such as industrial production, building HVAC, and transportation. In modern industrial production, heat recovery in chemical processes and steam condensation in power plants rely on heat exchange modules; in the civilian sector, air conditioning systems, water heaters, and other equipment are also indispensable. With the increasing demands for energy efficiency and equipment stability across industries, the performance and reliability of heat exchange modules directly impact the overall system's operating efficiency, energy consumption, and product quality. Failure of a heat exchange module can not only lead to energy waste and decreased production efficiency but may also cause equipment damage or even safety accidents. Therefore, accurate and timely fault diagnosis of heat exchange modules is essential to ensuring their normal operation.

[0004] Currently, fault diagnosis of heat exchange modules mainly relies on traditional methods and simple automation technologies. Traditional methods, such as manual inspection, depend on the experience of technicians to judge the equipment status through observation, touch, and sound. This approach is inefficient, highly subjective, and struggles to detect early potential faults and subtle internal problems, and it cannot achieve real-time monitoring. Existing automated diagnostic technologies are mostly based on single sensor data or simple threshold judgments, such as monitoring only temperature or pressure data and issuing an alarm when the value exceeds a set threshold. These methods ignore the correlation between multiple parameters during the operation of the heat exchange module, resulting in low accuracy in diagnosing complex faults, an inability to distinguish the characteristics of different types of faults, and a high risk of false alarms and missed alarms. Furthermore, existing fault diagnosis systems lack intelligent learning and adaptive capabilities, making it difficult to adapt to different operating conditions and environmental changes, and unable to self-optimize as the equipment's operating status changes and over time.

[0005] With the development of industrial intelligence, traditional fault diagnosis technologies are no longer sufficient to meet the growing operational and management needs of heat exchange modules. On the one hand, the structure and operating conditions of modern heat exchange modules are becoming increasingly complex, placing higher demands on the accuracy and timeliness of fault diagnosis. On the other hand, the widespread adoption of the equipment lifecycle management concept requires fault diagnosis systems to possess functions such as fault prediction and intelligent analysis. Therefore, there is an urgent need to develop an intelligent fault diagnosis method and device for heat exchange modules with multi-parameter fusion analysis, intelligent learning, and adaptive capabilities to address the shortcomings of existing technologies and improve the operational reliability and management level of heat exchange modules. Summary of the Invention

[0006] The present invention proposes an intelligent fault diagnosis method and device for heat exchange modules to solve the problems mentioned in the prior art.

[0007] To achieve the above objectives, the present invention adopts the following technical solution: an intelligent fault diagnosis device for a heat exchange module, comprising:

[0008] Data acquisition module: A distributed sensor network is used to deploy different types of sensors at the heat exchanger inlet and outlet, pump body, and valves of the heat exchange module to acquire fluid flow in real time; each sensor transmits the acquired temperature, pressure, flow, and vibration frequency to the data acquisition terminal through wired or wireless communication protocols; the acquisition terminal is equipped with a data buffer function to cope with communication failures;

[0009] Data preprocessing module: After receiving the collected raw running data, it first uses a median filtering algorithm to remove noise and isolated outliers; for continuous outlier data segments, it employs a statistical analysis-based anomaly detection method, calculating the interquartile range (IQR) of the data to identify outliers. and Data within the specified range is identified as anomaly and corrected; then, an adaptive moving average filtering algorithm is used to smooth the data. The algorithm automatically adjusts the size N of the moving window based on data fluctuations, as shown in the formula: ,in and This is an adjustable coefficient. The standard deviation of the data. The mean of the data is used; then, the Z-score normalization method is applied to normalize the data, and the formula is: Where x is the original data, The mean of the data. The standard deviation of the data;

[0010] Feature extraction module: Extracts features from preprocessed data, uses Principal Component Analysis (PCA) for dimensionality reduction, and employs Independent Component Analysis (ICA) to extract independent component features. ICA separates independent components by maximizing the non-Gaussianity of the data, as shown in the formula. Where W is the separation matrix, s is the source signal, and J is the non-Gaussianity metric function; simultaneously, the time-domain features, frequency-domain features, and time-frequency-domain features of the data are extracted; the time-frequency-domain feature extraction adopts Hilbert-Huang transform (HHT) to decompose the data into intrinsic mode functions (IMFs), and analyzes the instantaneous frequency and amplitude characteristics of each IMF;

[0011] Fault diagnosis model module: A hybrid model combining CapsNet (a deep learning-based capsule network) and LSTM (Long Short-Term Memory) is used for fault diagnosis. Extracted features are input into the hybrid model, where the input layer performs initial feature encoding. The model's loss function is a dynamic routing loss function combined with cross-entropy loss and reconstruction loss, as shown in the formula below. ,in For category labels, Let be the output vector of the capsule. and For the threshold, Here are the weighting coefficients for the reconstruction loss, and x is the input data. To reconstruct the data;

[0012] The results output module displays the fault diagnosis results through charts and text on the screen, and also provides a 3D visualization model to show the location and scope of the fault. The results output module also has a notification function, which sends fault information to operators via SMS and email.

[0013] Furthermore, it also includes:

[0014] Real-time monitoring module: Monitors the operating status of the heat exchange module in real time, and uses a multi-scale analysis method to analyze the collected data. Wavelet decomposition is used to decompose the data into different scale spaces, and corresponding early warning thresholds are set at each scale. For the data characteristics at different scales, fuzzy logic reasoning is used to make a comprehensive judgment to determine whether an early warning signal needs to be issued. The setting of the early warning threshold is based on historical data statistical analysis, and also takes into account the operating conditions of the heat exchange module and environmental factors.

[0015] Model update module: This module is used to periodically update the fault diagnosis model. It adopts a transfer learning-based approach. When there are differences between the newly collected operational data and the model training data, or when a new type of fault occurs, the pre-trained model parameters are used as initial values ​​to fine-tune the model on the new data. At the same time, an active learning mechanism is introduced. Data samples are selected and manually labeled, and then these labeled data are added to the training set to retrain the model.

[0016] Furthermore, the data acquisition module is equipped with a self-calibration function, which periodically self-calibrates the sensor by comparing it with a standard reference value and fitting a calibration curve using the least squares method. The formula is as follows: ,in The sensor's measured value, The standard reference value is a, and the calibration coefficients are b. At the same time, the data acquisition module monitors the health status of the sensor and issues a replacement prompt when a sensor malfunction is detected.

[0017] Furthermore, the feature extraction module also incorporates machine learning feature selection methods to remove redundant and irrelevant features, retaining features that contribute to fault diagnosis and improving the training efficiency of the model.

[0018] Furthermore, the fault diagnosis model module also uses federated learning to train the model. Data from heat exchange modules distributed in different geographical locations are used to train a global model together without sharing the original data, through encrypted communication and model parameter exchange.

[0019] Furthermore, the result output module is also equipped with a fault prediction function. By analyzing historical fault data and current operating data, it uses a time series prediction model to predict the types and times of future faults, providing early warnings and decision support for equipment maintenance and management.

[0020] A method for using an intelligent fault diagnosis system for the aforementioned heat exchange module includes:

[0021] Data acquisition steps: Use a distributed sensor network to collect operating data at the heat exchange module, including temperature, pressure, flow rate, and vibration frequency. Each sensor transmits the data to the acquisition terminal via wired or wireless communication protocols.

[0022] Data preprocessing steps: First, the median filtering algorithm is used to remove noise and isolated outliers from the collected raw running data. For continuous outlier data segments, an anomaly detection method based on statistical analysis is used for correction. Then, an adaptive moving average filtering algorithm is used for smoothing. The algorithm automatically adjusts the sliding window size N according to data fluctuations. Finally, the Z-score normalization method is used to normalize the data.

[0023] Feature extraction steps: Features are extracted from the preprocessed data. Principal component analysis (PCA) and independent component analysis (ICA) are used for dimensionality reduction. ICA separates independent components by maximizing the non-Gaussianity of the data. Simultaneously, time-domain features, frequency-domain features, and time-frequency-domain features are extracted. Time-frequency-domain features are extracted using Hilbert-Huang transform (HHT).

[0024] Fault diagnosis steps: The extracted features are input into a hybrid model combining CapsNet and LSTM deep learning for fault diagnosis. The model uses a dynamic routing loss function, which is determined by combining cross-entropy loss and reconstruction loss formulas. The fault type, fault probability, and fault location are determined based on the model output.

[0025] Results output steps: Fault diagnosis results are displayed through charts and text on the screen, and a 3D visualization model shows the fault location and scope of impact; at the same time, operators are notified via SMS, email, and other means.

[0026] Furthermore, it also includes:

[0027] Real-time monitoring steps: During the data acquisition process, a multi-scale analysis method is used to perform wavelet decomposition on the data, and warning thresholds are set at different scales. Fuzzy logic reasoning is used to comprehensively judge the data characteristics at different scales, and the warning thresholds are dynamically adjusted according to the operating conditions of the heat exchange module and environmental factors. When the data exceeds the warning threshold, a warning signal is issued.

[0028] Model update steps: Regularly evaluate the fault diagnosis model. When new data is abnormal compared to the training data or new fault types appear, use a transfer learning-based method to fine-tune the pre-trained model parameters. Introduce an active learning mechanism, select data samples, manually label them, and add them to the training set to retrain the model.

[0029] Compared with existing technologies, the beneficial effects of this invention are:

[0030] In terms of diagnostic accuracy, multi-dimensional data is collected through a distributed multi-sensor network, combined with multi-level feature extraction techniques, such as principal component analysis and independent component analysis, to mine data features. Then, a hybrid model combining capsule network and long short-term memory network is used for fault diagnosis, which can accurately identify various complex faults, effectively reduce false alarms and false negatives, and significantly improve diagnostic accuracy compared with traditional methods.

[0031] In terms of diagnostic efficiency, the data preprocessing module uses an adaptive algorithm to process data quickly, and the fault diagnosis model has efficient feature learning and reasoning capabilities, enabling real-time monitoring and rapid diagnosis of the heat exchange module's operating status. It can complete a large amount of data processing and analysis in a short time, promptly detect potential faults, and avoid equipment downtime and production losses caused by untimely fault detection.

[0032] The patented device and method also possess powerful adaptive and intelligent learning capabilities. The real-time monitoring module dynamically adjusts warning thresholds based on operating conditions and environmental changes; the model update module continuously optimizes the fault diagnosis model through transfer learning and active learning mechanisms, enabling it to adapt to different equipment and operating conditions. Furthermore, the results output module presents diagnostic results in a multimodal manner, providing not only intuitive charts and text information but also 3D visualization of the fault location. Combined with multi-channel notifications and tiered alerts, this facilitates timely monitoring of equipment status by maintenance personnel. Simultaneously, the fault prediction function allows for advance planning of equipment maintenance, reducing maintenance costs and improving the stability and reliability of equipment operation, bringing significant economic benefits and safety assurance to enterprises. Attached Figure Description

[0033] Figure 1 This is a schematic block diagram of the intelligent fault diagnosis system for the heat exchange module proposed in this invention.

[0034] Figure 2 This is a schematic block diagram of the intelligent fault diagnosis method for heat exchange modules proposed in this invention.

[0035] Figure 3 This is a schematic block diagram comparing the diagnostic accuracy of different fault types in the intelligent fault diagnosis method for the heat exchange module proposed in this invention.

[0036] Figure 4 This is a schematic block diagram illustrating the variation of average fault diagnosis time with data volume in the intelligent fault diagnosis method for heat exchange modules proposed in this invention. Detailed Implementation

[0037] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0038] In the description of this invention, it should be understood that the terms "center," "longitudinal," "lateral," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," "clockwise," and "counterclockwise," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0039] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, features defined with "first" and "second" may explicitly or implicitly include one or more of the stated features. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified. Furthermore, the terms "installed," "connected," and "linked" should be interpreted broadly; for example, they may refer to a fixed connection, a detachable connection, or an integral connection; they may refer to a mechanical connection or an electrical connection; they may refer to a direct connection or an indirect connection through an intermediate medium; and they may refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances. The invention will now be described in further detail with reference to the accompanying drawings.

[0040] Reference Figure 1 and Figure 4 Detailed Implementation of an Intelligent Fault Diagnosis Method and Device for Heat Exchange Modules

[0041] Data Acquisition Module: Sensors are installed at the heat exchanger inlet and outlet, circulating pump, and valves of the heat exchange module. The thermistor temperature sensor has a temperature measurement range of -40℃ to 200℃ and an accuracy of ±0.1℃, acquiring data once every 0.01 seconds. The piezoresistive pressure sensor has a measurement range of 0-10MPa and an accuracy of ±0.05%FS, acquiring data 10 times per second. The electromagnetic flow sensor has a measurement accuracy of ±0.2% and acquires data 5 times per second. The sensors transmit data to the data acquisition terminal via the Modbus protocol. The terminal is equipped with a 128GB solid-state drive, which can cache 72 hours of data. When communication is interrupted, the terminal automatically stores the data and retransmits it after communication is restored.

[0042] Data preprocessing module: After the raw data enters the preprocessing module, a median filtering algorithm with a window size of 3 is first used to remove isolated noise points. For continuous outliers, the quartiles Q1 and Q3 are calculated, based on... It will exceed and Data within the specified range is considered anomaly and corrected. Subsequently, the data standard deviation is used as the basis for correction. and mean According to the formula Calculate the window size N for the adaptive moving average filter to smooth the data. Finally, use the formula... Z-score normalization is applied to the data to map it to a standard distribution interval.

[0043] Feature extraction module: For the preprocessed data, Principal component analysis (PCA) is first applied to calculate the covariance matrix. Principal components with a cumulative contribution rate of 90% were selected. Independent component analysis (ICA) was then performed to maximize the non-Gaussianity metric function. Independent components are separated. Simultaneously, kurtosis and skewness in the time domain, power spectral density and frequency bandwidth in the frequency domain are extracted, and time-frequency domain features are obtained through Hilbert-Huang transform (HHT). The data is decomposed into multiple intrinsic mode functions (IMFs) to analyze instantaneous frequency and amplitude.

[0044] The fault diagnosis model module inputs extracted features into a hybrid model of CapsNet and Long Short-Term Memory (LSTM). The CapsNet's input layer initially encodes the features, and a dynamic routing algorithm combines these features. The LSTM part processes time-series information through forget gates, input gates, and output gates. Model training employs an improved dynamic routing loss function. ,in By optimizing model parameters through backpropagation algorithm, the diagnosis of fault type, probability and location can be achieved.

[0045] Results Output Module: Diagnostic results are displayed as a 3D visualization model showing the fault location, presented in a browser using WebGL technology. A report containing bar charts and line graphs is also generated, showing the fault probability distribution and time trend. Fault information is notified through multiple channels including WeChat, email, and SMS; severe faults trigger audible and visual alarms. A time series prediction model is used to predict the fault risk for the next 7 days based on data from the past 30 days.

[0046] Real-time monitoring module: Performs 5-level wavelet decomposition on the collected data, and sets parameters at each scale based on the mean and standard deviation of historical data. The principle of setting early warning thresholds is as follows: Fuzzy logic reasoning is used, taking the characteristics of parameters such as temperature, pressure, and flow rate at different scales as input, and making a comprehensive judgment through membership functions and fuzzy rule bases. When the fuzzy reasoning result exceeds the set threshold, an early warning is issued.

[0047] Model update module: The model is evaluated weekly. When the distribution difference between the new data and the training data exceeds a threshold (e.g., KL divergence greater than 0.3) or a new fault type appears, transfer learning is performed using the pre-trained model parameters. Through active learning algorithms, uncertainty indicators (e.g., entropy) of the data samples are calculated, and the top 10% of samples are manually labeled and added to the training set to retrain the model.

[0048] The data acquisition module has a self-calibration function: it performs monthly self-calibration of the sensor by measuring data from known standard values ​​(such as standard pressure blocks or constant temperature baths). Compared with standard value By least squares method Fit the calibration curve and update the sensor's calibration coefficients a and b. Simultaneously monitor the sensor's drift, prompting for replacement when the drift exceeds 10% of the accuracy. Feature extraction module optimization: Employ the Recursive Feature Elimination (RFE) algorithm, removing the feature with the smallest decrease in model prediction accuracy each time, repeating this process until the number of remaining features reaches 60% of the original features. Then, calculate feature importance using a random forest, selecting the top 50 most important features to construct the optimized feature set.

[0049] Federated learning for the fault diagnosis model module: Data from heat exchanger modules in multiple plants are used to train the model locally, with only the model parameter gradients uploaded to the central server. The central server aggregates the parameter gradients, updates the global model, and then distributes it to each plant. Each plant uses its local data to fine-tune the model, achieving collaborative training while protecting data privacy.

[0050] II. Detailed Implementation Methods and Steps: Data Acquisition Steps:

[0051] Following the deployment method of the data acquisition module described above, sensors are installed at key locations on the heat exchange module. The sensor acquisition frequency and communication protocol are set, and the data is transmitted to the acquisition terminal and buffered for 72 hours. Data preprocessing steps: The raw data is sequentially subjected to median filtering, outlier correction, adaptive moving average filtering, and smoothing. Finally, Z-score normalization is applied to obtain the preprocessed data.

[0052] Feature extraction steps: From the preprocessed data, dimensionality reduction is performed using PCA and ICA to extract time-domain, frequency-domain, and time-frequency-domain features. The feature set is then optimized using RFE and random forest algorithms.

[0053] Fault diagnosis steps: Input the optimized features into the CapsNet and LSTM hybrid model, train the model using the improved dynamic routing loss function, and determine the fault type, probability, and location based on the model output.

[0054] Results output steps: Diagnostic results are output in the form of 3D visualization, chart reports, multi-channel notifications and hierarchical reminders, and fault prediction is performed using time series models.

[0055] Real-time monitoring steps: Perform wavelet decomposition on the collected data, set early warning thresholds at different scales, and make a comprehensive judgment through fuzzy logic reasoning. When the conditions are met, an early warning is issued.

[0056] Model update steps: Regularly evaluate the model, and when the update conditions are met, use transfer learning and active learning to update the model parameters to improve the model's diagnostic capabilities.

[0057] III. Characterization and Interpretation of Beneficial Effect Data .

[0058] This patent significantly outperforms traditional solutions in fault diagnosis accuracy. Data shows that traditional solutions, relying on single-parameter judgment and simple models, struggle to identify complex faults. In contrast, this patent, through multi-sensor data fusion, multi-level feature extraction, and advanced hybrid models, accurately captures fault characteristics. The average fault diagnosis time is significantly reduced due to efficient data processing and a rapid inference model, enabling timely fault detection. Lower false alarm and false negative rates reduce unnecessary downtime and the risk of missed faults. Equipment maintenance costs are reduced by 35% thanks to early fault prediction and accurate diagnosis, making maintenance more planned and avoiding the high costs associated with over-maintenance and unexpected failures. This fully demonstrates the significant advantages of this patent in practical applications.

[0059] The above are merely preferred embodiments 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 intelligent fault diagnosis device for a heat exchange module, characterized in that, include: Data acquisition module: A distributed sensor network is used to deploy different types of sensors at the heat exchanger inlet and outlet, pump body, and valves of the heat exchange module to acquire fluid flow in real time; each sensor transmits the acquired temperature, pressure, flow rate, and vibration frequency to the data acquisition terminal through wired or wireless communication protocols; The data acquisition terminal is equipped with a data caching function to cope with communication failures; Data preprocessing module: After receiving the collected raw running data, it first uses the median filtering algorithm to remove noise and isolated outliers; For consecutive outlier data segments, a statistical analysis-based anomaly detection method is used. This method calculates the interquartile range (IQR) of the data to identify outliers. and Data within the specified range is identified as anomaly and corrected; then, an adaptive moving average filtering algorithm is used to smooth the data. The algorithm automatically adjusts the size N of the moving window based on data fluctuations, as shown in the formula: ,in and This is an adjustable coefficient. The standard deviation of the data. The mean of the data is used; then, the Z-score normalization method is applied to normalize the data, and the formula is: Where x is the original data, The mean of the data. The standard deviation of the data; Feature extraction module: Extracts features from preprocessed data, uses principal component analysis (PCA) to reduce the dimensionality of the data, and also uses independent component analysis (ICA) to extract independent component features from the data; ICA separates independent components by maximizing the non-Gaussianity of the data; it also extracts time-domain, frequency-domain, and time-frequency-domain features of the data; the time-frequency-domain feature extraction uses the Hilbert-Huang Transform (HHT). Fault diagnosis model module: A hybrid model combining CapsNet (a deep learning-based capsule network) and LSTM (Long Short-Term Memory) is used for fault diagnosis. Extracted features are input into the hybrid model, where the input layer performs initial feature encoding. The model's loss function is a dynamic routing loss function combined with cross-entropy loss and reconstruction loss, as shown in the formula below. ,in For category labels, Let be the output vector of the capsule. and For the threshold, Here are the weighting coefficients for the reconstruction loss, and x is the input data. To reconstruct the data; The data acquisition module is equipped with a self-calibration function, which periodically performs self-calibration on the sensor. By comparing the calibration curve with a standard reference value, the least squares method is used to fit the calibration curve, as shown in the formula. ,in The sensor's measured value, a and b are the standard reference values ​​and the calibration coefficients; at the same time, the data acquisition module monitors the health status of the sensor and issues a replacement prompt when a sensor malfunction is detected. The fault diagnosis model module also uses federated learning to train the model. Data from heat exchange modules distributed in different geographical locations are used to train a global model together without sharing the original data, through encrypted communication and model parameter exchange. The output module is equipped with a fault prediction function. By analyzing historical fault data and current operating data, it uses a time series prediction model to predict the types and times of future faults, providing early warnings and decision support for equipment maintenance and management. Real-time monitoring module: During the data acquisition process, a multi-scale analysis method is used to perform wavelet decomposition on the data, and early warning thresholds are set at different scales. Fuzzy logic reasoning is used to comprehensively judge the data characteristics at different scales, and the warning threshold is dynamically adjusted according to the operating conditions of the heat exchange module and environmental factors. When the data exceeds the warning threshold, a warning signal is issued.

2. The intelligent fault diagnosis device for the heat exchange module according to claim 1, characterized in that, The feature extraction module also incorporates machine learning feature selection methods to remove redundant and irrelevant features, retaining features that contribute to fault diagnosis and improving the training efficiency of the model.

3. The intelligent fault diagnosis device for the heat exchange module according to claim 1, characterized in that, Also includes: The results output module displays the fault diagnosis results through charts and text on the screen, and also provides a 3D visualization model to show the location and scope of the fault. The results output module also has a notification function, which sends fault information to operators via SMS and email.

4. The intelligent fault diagnosis device for the heat exchange module according to claim 1, characterized in that, Also includes: Model update module: Regularly evaluate the fault diagnosis model. When new data is different from the training data or new fault types appear, use the transfer learning method to fine-tune the pre-trained model parameters. Introduce an active learning mechanism, select data samples, manually label them, and add them to the training set to retrain the model.

5. A method for applying the intelligent fault diagnosis device for the heat exchange module according to any one of claims 1-4, characterized in that, include: The steps include data acquisition, data preprocessing, feature extraction, fault diagnosis, result output, real-time monitoring, and model update.

Images