Health state self-diagnosis device and method for compressed air foam fire extinguishing system

Abstract

The invention discloses a health state self-diagnosis device and method of a compressed air foam fire extinguishing system, and belongs to the technical field of fire fighting, and the health state self-diagnosis device of the compressed air foam fire extinguishing system comprises a fire extinguishing device, a control device assembly and a collector assembly, a collector assembly and a control device assembly are both arranged in the fire extinguishing device, the collector assembly is responsible for monitoring key parameters of the fire extinguishing device in real time, the collector assembly is in circuit connection with the control device assembly, the control device assembly comprises a data collection system and a processing system, the data collection system is in circuit connection with the collector assembly, and the processing system is in circuit connection with the collector assembly. The data collection system is used for collecting real-time data of all elements of the detection device, the data collection system is connected with the processing system, and the processing system processes the collected data, monitors abnormal conditions of the device, gives an alarm in time, and can perform real-time feedback correction and risk assessment.

Description

Technical Field

[0001] The present invention relates to the field of fire fighting technology, and in particular to a health status self-diagnosis device and method for a compressed air foam fire extinguishing system. Background Art

[0002] The compressed air foam fire extinguishing system is an important fire extinguishing system used in substations. It plays an important role in substation fire protection. However, the compressed air foam fire extinguishing system has many components and complex composition. The debugging and maintenance require a high level of technology. Failures may occur during its standby or operation, leading to serious accident consequences. Therefore, it is particularly important to promptly detect and deal with the safety hazards of the compressed air foam fire extinguishing system. The current compressed air foam fire extinguishing system lacks effective detection devices. If voltage and current fluctuations, pressure and flow are too high or too low, equipment temperature is too high, liquid level is too low, and the water pump is empty, if not handled in time, it may cause equipment shutdown, affecting fire extinguishing or even damage.

[0003] Most compressed air foam fire extinguishing systems do not pay attention to the physical parameters during the actual equipment debugging and maintenance process, and cannot deeply detect the fault problems that are prone to occur during the actual operation of the compressed air foam fire extinguishing system. There is no intelligent monitoring of the compressed air foam fire extinguishing system, and the level of intelligent monitoring is low. Summary of the invention

[0004] In order to solve the deficiencies in the prior art, the present invention provides a self-diagnosis device and method for the health status of a compressed air foam fire extinguishing system. The present invention solves the problems that the compressed air foam fire extinguishing system has a complex structure and is inconvenient for maintenance, and that system hidden dangers and faults are difficult to identify. The present invention can effectively monitor the health status of the system, discover hidden dangers and deal with them early to avoid accidents.

[0005] The present invention adopts the following technical solution.

[0006] The first aspect of the present invention discloses a health status self-diagnosis device for a compressed air foam fire extinguishing system, comprising: a fire extinguishing device, a control device component and a collector component, wherein the collector component and the control device component are both arranged inside the fire extinguishing device, the collector component is responsible for real-time monitoring of key parameters of the fire extinguishing device, the collector component and the control device component are circuit-connected, the control device component comprises: a data collection system and a processing system, the data collection system and the collector component are circuit-connected, and are used to collect real-time data of all components of the detection device, the data collection system and the processing system are connected, the processing system processes the collected data, monitors abnormal conditions of the device, promptly alarms, and is capable of real-time feedback correction and risk assessment.

[0007] Preferably, the collector assembly includes: electrical components, pressure components, flow components, temperature components, load components, speed components and sound components, all components of the control device assembly and the collector assembly are circuit-connected, the fire extinguishing device includes: an air compressor, a compressed air pipeline, a water tank, a water pipeline, a foam liquid tank, a foam pipeline, a water pump, a mixing pipeline, a foam pump and an injection port, the water tank and the water pump are connected through a water pipeline, the foam liquid tank and the foam pump are connected through a foam pipeline, the water pipeline and the foam pipeline are connected and the connection between the two is connected to the mixing pipeline, a nozzle is arranged on one side of the mixing pipeline, the air compressor is connected to the mixing pipeline through a compressed air pipeline,

[0008] Preferably, the electrical components include: a voltmeter and an ammeter. The air compressor, water pump and foam pump are all provided with a voltmeter and an ammeter for monitoring the voltage and current of the air compressor, water pump and foam pump. The pressure component includes: a pressure gauge. The pressure gauge is arranged on the compressed air pipeline, water pipeline, foam pipeline and mixing pipeline to monitor the pressure of all pipelines. The flow component includes: an air flow meter and an electromagnetic flow meter. The electromagnetic flow meter is respectively arranged on the water pipeline and the foam pipeline, and the air flow meter is arranged on the compressed air pipeline. The temperature component includes: a temperature sensor. The temperature sensor is arranged in the air compressor, water pump and foam pump to record the temperature and time change curve. The speed component includes: a speed sensor. The speed sensor is arranged in the air compressor, water pump and foam pump to monitor the speed of the air compressor, water pump and foam pump. The sound component includes: a sound collector. The sound collector is arranged in the air compressor, water pump and foam pump to collect sound information.

[0009] Preferably, the control device component further comprises: a label system, which analyzes the historical operation data of each group of elements of the collector component in advance and sets thresholds.

[0010] The second aspect of the present invention discloses a method for a health status self-diagnosis device for a compressed air foam fire extinguishing system, based on the health status self-diagnosis device for a compressed air foam fire extinguishing system, comprising the following steps:

[0011] A. Label setting: The label system on the control device assembly analyzes the historical operating data of each group of components in advance and sets the threshold value, sets the data range of each group of components when they are operating normally to be marked as "normal" and sets the data range outside the range to be "fault";

[0012] B. Data collection: Various components of the collector assembly transmit data to the data collection system of the control device assembly, and the data collection system transmits the collected data to the processing system of the control device assembly;

[0013] C. Data preprocessing: The processing system performs data standardization and normalization, and the data of the sound element is transformed into a spectrum through Fourier transformation;

[0014] D. Data input to the model: The pre-processed data of the electrical components, pressure components, flow components and speed components are transmitted to the Dense fully connected layer; the temperature time series formed by the temperature component is transmitted to the LSTM or GRU layer; the sound component is transformed into a spectrum through Fourier transform and input into the CNN convolution layer;

[0015] E. Feature extraction and fusion: After the data of each group of components arrives at the corresponding branch layer, they are fused after feature extraction to build a larger feature vector; the features are further processed through the fully connected layer;

[0016] F. Classification and regression reasoning: The fused feature vector enters the output layer, and the data of each component outputs the normal or faulty data of the equipment through the classification task. At the same time, the data of the pressure component, flow component and temperature component outputs the numerical prediction data through the regression task;

[0017] G. Fault detection and alarm system: detects abnormal or fault conditions through output data, triggers alarms in time, and provides processing suggestions based on historical data and current status;

[0018] H. Data training and verification: Real-time data is used for model reasoning and compared with historical data to evaluate the model effect;

[0019] I. Data self-correction: Adapt to new data through online learning through data training to improve prediction accuracy;

[0020] J. Safety prediction and risk assessment: Use time series data to predict future operating trends, generate safety assessment reports and provide fault prevention suggestions, classify risk levels, and take preventive measures.

[0021] Preferably, step C comprises:

[0022] Step C1: Data cleaning, removing outliers, noise or missing data points,

[0023] Step C2: Standardize or normalize the data, convert data of different ranges into the same scale, and implement data synchronization;

[0024] Step C3: filtering is then performed to remove noise interference;

[0025] Step C4: At the same time, the sound sensor of the sound element uses Fourier transform to convert the time domain signal into a frequency domain signal to obtain a spectrum, and divides the audio signal into segments to generate a spectrum diagram.

[0026] Preferably, in step F, the output layer uses a Softmax activation function to perform classification tasks, all component data are output through the classification task and compared with the threshold set by the label system to determine whether the equipment is normal or faulty, and at the same time, the output layer also uses a linear activation function to perform regression tasks to predict pressure, temperature-time, and flow numerical data, and compares the predicted value with the actual value to determine whether it exceeds the threshold set by the label system to determine whether the equipment is faulty.

[0027] Preferably, the loss function of the classification task adopts cross entropy loss, and the loss function of the regression task adopts mean square error loss.

[0028] Preferably, step G comprises:

[0029] Step G1: Data division: divide the data into training set, validation set and test set, and 70% of the data is used for training, 15% for validation and 15% for testing;

[0030] Step G2: Data enhancement, the sound data needs to be enhanced by spectrogram;

[0031] Step G3: Data training, training the neural network model, monitoring the changes in the loss function during training to avoid overfitting;

[0032] Step G4: Evaluate the classification task of the neural network model, compare the results with the labeling system, calculate the classification evaluation index, and use the evaluation function of scikit-learn to calculate the accuracy, precision, recall, and F1-score to evaluate the fault classification effect of the classification model of the neural network model;

[0033] Step G5: At the same time, the regression task of the neural network model is evaluated by comparing the predicted value with the true value, using the mean square error or mean absolute error to evaluate, and then evaluating the performance of the regression model in the neural network model.

[0034] Preferably, step I prevents overfitting by using L2 regularization or Dropout on the neural network model on the training set, and then adjusts the number of layers and neurons of the model, optimizes the model architecture, and uses adaptive weights for sensor data of different components, so that the model automatically learns to correct each sensor data.

[0035] The beneficial effect of the present invention is that, compared with the prior art, the present invention starts from real-time data, combines the debugging experience of the equipment, and through comprehensive analysis of multi-element sensor data, as well as reasoning and detection of neural network models, the device can effectively predict and detect abnormal conditions of compressed air foam fire extinguishers. By further combining the safety prediction module, it is possible to effectively pay attention to the abnormal state and failure of the equipment, realize comprehensive monitoring, collection, processing and feedback of the compressed air foam system, detect potential failures in advance, and take appropriate protective measures to ensure the safe operation of the equipment and avoid losses. This comprehensive system not only improves the fault detection capability of the equipment, helps to quickly discover hidden dangers and ensure the healthy and normal operation of the equipment, but also enhances the reliability of the equipment and can detect and handle failures in a timely manner. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 It is a structural diagram of the health status self-diagnosis device of the compressed air foam fire extinguishing system in the present invention;

[0037] Figure 2 It is a method flow chart of the health status self-diagnosis device of the compressed air foam fire extinguishing system in the present invention.

[0038] In the figure: 1. Voltmeter; 2. Ammeter; 3. Speed ​​sensor; 4. Temperature sensor; 5. Sound collector; 6. Pressure gauge; 7. Air flow meter; 8. Electric ball valve; 9. Electromagnetic flow meter; 10. Electric regulating valve. DETAILED DESCRIPTION

[0039] In order to make the purpose, technical scheme and advantages of the present invention clearer, the technical scheme of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. The embodiments described in this application are only part of the embodiments of the present invention, not all of them. Based on the spirit of the present invention, other embodiments obtained by ordinary technicians in this field without creative work are all within the scope of protection of the present invention.

[0040] like Figure 1As shown, the present invention provides a health status self-diagnosis device and method for a compressed air foam fire extinguishing system. The device includes: a fire extinguishing device, a control device component and a collector component. The control device component and the collector component are both arranged inside the fire extinguishing device. The collector component includes: electrical components, pressure components, flow components, temperature components, load components, speed components and sound components. All components of the control device component and the collector component are circuit-connected. The fire extinguishing device includes: an air compressor, a compressed air pipeline, a water tank, a water pipeline, a foam liquid tank, a foam pipeline, a water pump, a mixing pipeline, a foam pump and an injection port. The water tank and the water pump are connected through a water pipeline, the foam liquid tank and the foam pump are connected through a foam pipeline, the water pipeline and the foam pipeline are connected and the connection between the two is connected to the mixing pipeline, a nozzle is arranged on one side of the mixing pipeline, the air compressor is connected to the mixing pipeline through a compressed air pipeline, and the electrical components in the collector component include: a voltmeter 1 and An ammeter 2, an air compressor, a water pump and a foam pump are all provided with a voltmeter 1 and an ammeter 2 for monitoring the voltage and current of the air compressor, the water pump and the foam pump, a pressure element comprises a pressure gauge 6, the pressure gauge 6 is arranged on the compressed air pipeline, the water pipeline, the foam pipeline and the mixing pipeline, and monitors the pressure of all pipelines, a flow element comprises an air flow meter 7 and an electromagnetic flow meter 9, the electromagnetic flow meter 9 is respectively arranged on the water pipeline and the foam pipeline, and the air flow meter 7 is arranged on the compressed air pipeline, a temperature element comprises a temperature sensor 4, the temperature sensor 4 is arranged in the air compressor, the water pump and the foam pump, and is used to record the temperature and time change curve, a speed element comprises a speed sensor 3, the speed sensor 3 is arranged in the air compressor, the water pump and the foam pump, and is used to monitor the speed of the air compressor, the water pump and the foam pump, a sound element comprises a sound collector 5, the sound collector 5 is arranged in the air compressor, the water pump and the foam pump, and is used to collect sound information.

[0041] The control device assembly includes: a data collection system, a label system and a processing system. The control device assembly also has a user interface that displays the system status and allows operators to set parameters and perform system diagnosis.

[0042] The labeling system analyzes the historical operating data of each group of components in advance and sets thresholds, setting the data range when each group of components is operating normally to be marked as "normal" and setting the data outside the range to be "faulty";

[0043] The data collection system is connected to all components of the collector assembly to collect real-time data from all components of the detection device. The data collection system is connected to the processing system.

[0044] The data collection system transmits the collected data to the processing system for data preprocessing. First, the data is cleaned to remove outliers, noise or missing data points, and then the data is standardized or normalized to convert data of different ranges to the same scale, implement data synchronization, and then filter to remove noise interference, thereby completing data preprocessing. At the same time, the sound sensor of the sound element uses Fourier transform to convert the time domain signal into the frequency domain signal to obtain the spectrum, and the audio signal is segmented to generate a spectrum diagram as the input feature of the subsequent CNN (convolutional neural network) model.

[0045] After that, the processing system inputs the pre-processed data of all components into the corresponding processing layer for feature extraction. The processing system uses multiple parallel neural network model branches to process different types of numerical data. The numerical data generated by electrical components, pressure components, load components, flow components and speed components are all input into the Dense layer (fully connected layer) for feature extraction, while the temperature-time curve data generated by the temperature component is time series data input into the LSTM (Long Short-Term Memory Network) or GRU (Gated Recurrent Unit) to capture time dependencies, and the sound component is input into the CNN convolution layer through the Fourier transformed spectrum diagram for feature extraction;

[0046] After feature extraction, the data of each group of components are fused using the Concatenate function to construct a larger feature vector;

[0047] The fused feature vector enters the output layer through the fully connected layer Dense. The output layer uses the Softmax activation function for classification tasks. All component data are output through the classification task and compared with the threshold set by the label system to determine whether the equipment is normal or faulty to determine whether all components are normal or faulty. The loss function of the classification task uses cross entropy loss. At the same time, the output layer also uses a linear activation function for regression tasks to predict numerical data such as pressure, temperature-time, and flow. The loss function of the regression task uses mean square error loss.

[0048] At the same time, classification tasks require data training and verification to ensure the performance of classification tasks. The performance of neural network models is ensured through data partitioning, data enhancement and data training. Data partitioning is to divide data into training set, validation set and test set. Usually 70% of the data is used for training, 15% for validation, and 15% for testing. Data enhancement is for sound data, which may require spectrum enhancement, such as random shift and noise injection. A small amount of random perturbation can be performed on other sensor data to enhance the generalization ability of the model. Data training: Train the neural network model and monitor the changes in the loss function during training to avoid overfitting.

[0049] Moreover, the device adopts the Adam optimizer to train the model, and the learning rate can be adjusted through experiments and can generally be set to 0.001.

[0050] During data training, in order to ensure the accuracy of judging whether a component is normal or faulty through classification tasks, the results of the classification task in the neural network model are compared with the threshold of the label system, and the classification evaluation index is calculated. The evaluation function is used to calculate the accuracy, precision, recall, and F1-score (the harmonic mean of precision and recall), and then the fault classification effect of the classification model in the neural network model is evaluated.

[0051] At the same time, the evaluation of the regression task of the neural network model is to compare the predicted value with the true value, using the mean square error (MSE) or mean absolute error (MAE) to evaluate, and then evaluate the performance of the regression model in the neural network model. In addition, the neural network model uses L2 regularization (weight decay) or Dropout on the training set to prevent overfitting, and then adjusts the number of layers and neurons of the model to optimize the model architecture. Adaptive weights are used for sensor data of different components to allow the model to automatically learn the importance of each sensor data.

[0052] At the same time, in a specific embodiment, the device deploys a neural network model on the control device component to monitor the sensor data of each component in real time, and perform health status prediction and fault diagnosis. In order to improve the safety of the equipment during operation, a special safety prediction module is designed, which is mainly used to predict whether the equipment is at risk of imminent failure and take corresponding measures in advance. The safety prediction model is constructed through the following aspects:

[0053] 1. Historical data and safety boundary setting:

[0054] Extract the normal range and threshold of equipment operation from a large amount of equipment operation history data. Use statistical methods to analyze the range values ​​of various sensors under normal conditions and define the threshold of equipment operation. The model uses the current operation status data of the equipment and combines it with historical operation data to determine whether it is within the safe range.

[0055] 2. Safety prediction based on time series: Use time series prediction algorithms (such as LSTM) to predict key parameters of the equipment (such as voltage, pressure, temperature, etc.) and analyze their changing trends in the future. If the model predicts that certain parameters may exceed the safety range, an early warning will be issued. Use anomaly detection algorithms (such as autoencoders or isolation forests) to detect abnormal fluctuations in sensor data and determine whether the equipment has potential failures.

[0056] 3. Safety assessment based on fault classification: For different types of faults, the model can generate a safety assessment report by analyzing the frequency, severity and potential impact of the fault on the equipment. For example, it can predict overheating or damage caused by idling of the water pump and take timely measures.

[0057] 4. Fault level and handling suggestions:

[0058] The possible failures of the equipment are classified into different levels according to their severity, such as low, medium and high risks, and different treatment plans are set for each level. The model combines multiple types of sensor data to evaluate the current safety status of the equipment and give specific countermeasures. For example, if it is predicted that a water pump may run dry, the model can recommend reducing the workload or shutting down for inspection.

[0059] like Figure 2 As shown, the present invention provides a method for self-diagnosis of the health status of a compressed air foam fire extinguishing system, and the specific steps include:

[0060] A. Label setting: The label system on the control device assembly analyzes the historical operating data of each group of components in advance and sets the threshold value, sets the data range of each group of components when they are operating normally to be marked as "normal" and sets the data range outside the range to be "fault";

[0061] B. Data collection: Various components of the collector assembly transmit data to the data collection system of the control device assembly, and the data collection system transmits the collected data to the processing system of the control device assembly;

[0062] C. Data preprocessing: The processing system performs data standardization and normalization, and the data of the sound element is transformed into a spectrum through Fourier transformation;

[0063] Step C1: Data cleaning, removing outliers, noise or missing data points,

[0064] Step C2: Standardize or normalize the data, convert data of different ranges into the same scale, and implement data synchronization;

[0065] Step C3: filtering is then performed to remove noise interference;

[0066] Step C4: At the same time, the sound sensor of the sound element uses Fourier transform to convert the time domain signal into a frequency domain signal to obtain a spectrum, and divides the audio signal into segments to generate a spectrum diagram.

[0067] D. Data input to the model: The pre-processed data of electrical components, pressure components, flow components and speed components are transmitted to the Dense fully connected layer; the temperature time series formed by the temperature component is transmitted to the LSTM or GRU layer; the Fourier transform of the sound component forms a spectrum graph and is input to the CNN convolution layer;

[0068] E. Feature extraction and fusion: After the data of each group of components arrives at the corresponding branch layer, they are fused after feature extraction to build a larger feature vector; the features are further processed through the fully connected layer;

[0069] F. Classification and regression reasoning: The fused feature vector enters the output layer, and all component data are compared with the threshold set by the label system through the classification task output to determine whether the equipment is normal or faulty. At the same time, the data of pressure components, flow components, and temperature components are output as numerical prediction data through the regression task;

[0070] G. Fault detection and alarm system: detects abnormal or fault conditions through output data, triggers alarms in time, and provides processing suggestions based on historical data and current status;

[0071] Step G1: Data division, divide the data into training set, validation set and test set. Usually 70% of the data is used for training, 15% for validation and 15% for testing.

[0072] Step G2: Data enhancement: the sound data needs to be enhanced with a spectrogram.

[0073] Step G3: Data training, training the neural network model, monitoring the changes in the loss function during training to avoid overfitting;

[0074] Step G4: Evaluate the classification task of the neural network model, compare the results with the labeling system, calculate the classification evaluation index, use the evaluation function to calculate the accuracy, precision, recall, and F1-score, and then evaluate the fault classification effect of the classification model of the neural network model;

[0075] Step G5: At the same time, the regression task of the neural network model is evaluated by comparing the predicted value with the true value, using the mean square error or mean absolute error to evaluate, and then evaluating the performance of the regression model in the neural network model.

[0076] H. Data training and verification: Real-time data is used for model reasoning and compared with historical data to evaluate the model effect;

[0077] I. Data self-correction: Adapt to new data through online learning through data training to improve prediction accuracy;

[0078] J. Safety prediction and risk assessment: Use time series data to predict future operating trends, generate safety assessment reports and provide fault prevention suggestions, classify risk levels (low, medium, high), and take preventive measures.

[0079] The beneficial effect of the present invention is that, compared with the prior art, the present invention starts from real-time data, combines the debugging experience of the equipment, and through comprehensive analysis of multi-element sensor data, as well as reasoning and detection of neural network models, the device can effectively predict and detect abnormal conditions of the equipment. By further combining the safety prediction module, it is possible to effectively pay attention to the abnormal state and failure of the equipment, realize comprehensive monitoring, collection, processing and feedback of the compressed air foam system, detect potential failures in advance, and take appropriate protective measures to ensure the safe operation of the equipment and avoid losses. This comprehensive system not only improves the fault detection capability of the equipment, helps to quickly discover hidden dangers and ensure the healthy and normal operation of the equipment, but also enhances the reliability of the equipment and can detect and handle failures in a timely manner.

[0080] 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, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents, and any modifications or equivalent replacements that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A health status self-diagnosis device for a compressed air foam fire extinguishing system, comprising: The fire extinguishing device, control device assembly and collector assembly are characterized by: The collector component and the control device component are both arranged inside the fire extinguishing device. The collector component is responsible for real-time monitoring of key parameters of the fire extinguishing device. The collector component and the control device component are circuit-connected. The control device component includes: a data collection system and a processing system. The data collection system and the collector component are circuit-connected and used to collect real-time data of all components of the detection device. The data collection system and the processing system are connected. The processing system processes the collected data, monitors abnormal conditions of the device, promptly alarms, and can provide real-time feedback, corrections, and risk assessments.

2. A health status self-diagnosis device for a compressed air foam fire extinguishing system according to claim 1, characterized in that: The collector assembly includes: electrical components, pressure components, flow components, temperature components, load components, speed components and sound components. All components of the control device assembly and the collector assembly are circuit-connected. The fire extinguishing device includes: an air compressor, a compressed air pipeline, a water tank, a water pipeline, a foam liquid tank, a foam pipeline, a water pump, a mixing pipeline, a foam pump and an injection port. The water tank and the water pump are connected through a water pipeline, the foam liquid tank and the foam pump are connected through a foam pipeline, the water pipeline and the foam pipeline are connected and the connection between the two is connected to the mixing pipeline, a injection port is arranged on one side of the mixing pipeline, and the air compressor is connected to the mixing pipeline through a compressed air pipeline.

3. A health status self-diagnosis device for a compressed air foam fire extinguishing system according to claim 2, characterized in that: The electrical components include: a voltmeter (1) and an ammeter (2). The air compressor, the water pump and the foam pump are all provided with the voltmeter (1) and the ammeter (2) for monitoring the voltage and current of the air compressor, the water pump and the foam pump. The pressure component includes: a pressure gauge (6). The pressure gauge (6) is provided on the compressed air pipeline, the water pipeline, the foam pipeline and the mixing pipeline to monitor the pressure of all pipelines. The flow component includes: an air flow meter (7) and an electromagnetic flow meter (9). The electromagnetic flow meter (9) is provided on the water pipeline and the foam pipeline respectively. The air flow meter (7) is provided on the compressed air pipeline. The temperature component includes: a temperature sensor (4). The temperature sensor (4) is provided in the air compressor, the water pump and the foam pump to record the temperature and time change curve. The speed component includes: a speed sensor (3). The speed sensor (3) is provided in the air compressor, the water pump and the foam pump to monitor the speed of the air compressor, the water pump and the foam pump. The sound component includes: a sound collector (5). The sound collector (5) is provided in the air compressor, the water pump and the foam pump to collect sound information.

4. The health status self-diagnosis device of a compressed air foam fire extinguishing system according to claim 1, characterized in that: The control device assembly also includes: a label system, which analyzes the historical operation data of each group of components of the collector assembly in advance and sets thresholds.

5. A method for self-diagnosis of the health status of a compressed air foam fire extinguishing system, based on a device for self-diagnosis of the health status of a compressed air foam fire extinguishing system as claimed in any one of claims 1 to 4, characterized in that: A. Label setting: The label system on the control device component analyzes the historical operating data of each group of components in advance and sets the threshold value, sets the data range of each group of components when they are operating normally to be marked as "normal" and sets the data range outside the range to be "fault"; B. Data collection: Various components of the collector assembly transmit data to the data collection system of the control device assembly, and the data collection system transmits the collected data to the processing system of the control device assembly; C. Data preprocessing: The processing system performs data standardization and normalization, and the data of the sound element is transformed into a spectrum through Fourier transformation; D. Data input to the model: The pre-processed data of electrical components, pressure components, flow components and speed components are transferred to the Dense fully connected layer; The temperature time series formed by the temperature element is transmitted to the LSTM or GRU layer; The sound component is transformed into a spectrogram through Fourier transform and input into the convolutional layer of CNN; E. Feature extraction and fusion: After the data of each group of components arrives at the corresponding branch layer, they are fused after feature extraction and fused into a vector; the features are further processed through the fully connected layer; F. Classification and regression reasoning: The fused feature vector enters the output layer, and the data of each component outputs the normal or faulty data of the equipment through the classification task. At the same time, the data of the pressure component, flow component and temperature component outputs the numerical prediction data through the regression task; G. Fault detection and alarm system: detects abnormal or fault conditions through output data, triggers alarms in time, and provides processing suggestions based on historical data and current status; H. Data training and verification: Real-time data is used for model reasoning and compared with historical data to evaluate the model effect; I. Data self-correction: Adapt to new data through online learning through data training to improve prediction accuracy; J. Safety prediction and risk assessment: Use time series data to predict future operating trends, generate safety assessment reports and provide fault prevention suggestions, classify risk levels, and take preventive measures.

6. A method for self-diagnosis of the health status of a compressed air foam fire extinguishing system according to claim 5, characterized in that: The step C comprises: Step C1: Data cleaning, removing outliers, noise or missing data points, Step C2: Standardize or normalize the data, convert data of different ranges into the same scale, and implement data synchronization; Step C3: filtering is then performed to remove noise interference; Step C4: At the same time, the sound sensor of the sound element uses Fourier transform to convert the time domain signal into a frequency domain signal to obtain a spectrum, and divides the audio signal into segments to generate a spectrum diagram.

7. A method for self-diagnosis of the health status of a compressed air foam fire extinguishing system according to claim 5, characterized in that: The loss function of the classification task in step F adopts cross entropy loss, and the loss function of the regression task adopts mean square error loss.

8. The method for self-diagnosis of the health status of a compressed air foam fire extinguishing system according to claim 5, characterized in that: In step F, the output layer uses the Softmax activation function to perform classification tasks. All component data are output through the classification task and compared with the threshold set by the label system to determine whether the equipment is normal or faulty. At the same time, the output layer also uses a linear activation function to perform regression tasks to predict pressure, temperature-time, and flow numerical data. The predicted value is compared with the actual value of the label system and exceeds the set threshold to determine whether the equipment is faulty.

9. A method for self-diagnosis of the health status of a compressed air foam fire extinguishing system according to claim 5, characterized in that: The step G comprises: Step G1: Data division: divide the data into training set, validation set and test set, and 70% of the data is used for training, 15% for validation and 15% for testing; Step G2: Data enhancement, the sound data needs to be enhanced by spectrogram; Step G3: Data training, training the neural network model, monitoring the changes in the loss function during training to avoid overfitting; Step G4: Evaluate the classification task of the neural network model, compare the results with the labeling system, calculate the classification evaluation index, use the evaluation function to calculate the accuracy, precision, recall, and F1-score, and then evaluate the fault classification effect of the classification model of the neural network model; Step G5: At the same time, the regression task of the neural network model is evaluated by comparing the predicted value with the true value, using the mean square error or mean absolute error to evaluate, and then evaluating the performance of the regression model in the neural network model.

10. A method for self-diagnosis of the health status of a compressed air foam fire extinguishing system according to claim 4, characterized in that: In step I, L2 regularization or Dropout is used on the neural network model in the training set to prevent overfitting, thereby adjusting the number of layers and neurons of the model, optimizing the model architecture, and using adaptive weights for sensor data of different components, so that the model can automatically learn to correct each sensor data.

Images