Air chamber pressure stabilizing and adjusting method and system of high-precision constant-current bubble water level gauge

Abstract

The application provides a gas chamber pressure stabilizing adjustment method and system of a high-precision constant-current bubble water level gauge, and belongs to the technical field of gas pressure control. The method comprises the following steps: obtaining multi-source data; performing feature extraction and fusion on the multi-source data to obtain a multi-source feature vector; inputting the multi-source feature vector into a pre-trained gas pressure fault diagnosis model to obtain a gas pressure fault category; based on the gas pressure fault category, matching a corresponding pressure stabilizing adjustment strategy in a pre-set adjustment strategy library; based on the pressure stabilizing adjustment strategy, controlling a target execution mechanism to adjust the pressure of the gas chamber of the bubble water level gauge to obtain a target gas pressure; within a pre-set observation period, verifying the stability of the target gas pressure to obtain a verification result; if the verification result is verified, maintaining the current adjustment strategy and performing steady-state monitoring; if the verification result is not verified, executing an optimization iteration mechanism. The application enhances the adaptability of the bubble water level gauge in a complex environment.

Description

Technical Field

[0001] This application relates to the field of air pressure control technology, and in particular to a method and system for regulating the air chamber pressure of a high-precision constant flow bubble level gauge. Background Technology

[0002] Bubble level gauges are widely used in hydrological monitoring and water conservancy projects due to their strong anti-interference capabilities and adaptability to harsh environments. Their measurement accuracy depends on the stability of the air chamber pressure. However, in practical applications, factors such as fluctuations in ambient temperature and humidity, changes in air source pressure, and equipment aging can cause abnormal air chamber pressure, leading to measurement deviations or even data failure. Traditional fault diagnosis relies on single-parameter monitoring, resulting in high false alarm rates and delayed responses. Furthermore, fault handling often requires manual intervention, leading to low measurement efficiency. Existing pressure stabilization strategies lack specificity and are difficult to adapt to complex fault scenarios, often resulting in persistent pressure instability even after multiple adjustments.

[0003] To address the aforementioned issues, achieve accurate diagnosis and efficient adjustment of air pressure faults, and ensure the reliability of water level monitoring data, there is an urgent need for a high-precision constant flow bubble level gauge air chamber pressure stabilization and adjustment method and system. Summary of the Invention

[0004] To address the aforementioned technical problems, this application provides a method and system for regulating the pressure of the air chamber of a high-precision constant flow bubble level gauge.

[0005] A first aspect of this application provides a method for regulating the pressure of the air chamber of a high-precision constant-flow bubble level gauge, comprising: Acquire multi-source data, including environmental data of the bubble level gauge and state parameters of the bubble level gauge; The multi-source data is subjected to feature extraction and fusion to obtain a multi-source feature vector; The multi-source feature vectors are input into a pre-trained barometric pressure fault diagnosis model to obtain the barometric pressure fault category; Based on the gas pressure fault category, a corresponding pressure stabilization and regulation strategy is matched with the preset regulation strategy library; Based on the pressure stabilization and regulation strategy, the target actuator is controlled to adjust the air chamber pressure of the bubble level gauge to obtain the target air pressure; During the preset observation period, the stability of the target air pressure is verified, and the verification results are obtained. If the verification result is successful, the current adjustment strategy is maintained and steady-state monitoring is performed; If the verification result is that the verification fails, then an optimization iteration mechanism is executed.

[0006] A second aspect of this application provides a chamber pressure stabilization and regulation system for a high-precision constant-flow bubble level gauge, comprising: The data acquisition module is used to acquire multi-source data, including environmental data of the bubble level gauge and state parameters of the bubble level gauge. The feature fusion module is used to extract and fuse features from the multi-source data to obtain a multi-source feature vector. The fault diagnosis module is used to input the multi-source feature vector into the pre-trained barometric pressure fault diagnosis model to obtain the barometric pressure fault category. The strategy matching module is used to match the corresponding pressure stabilization and regulation strategy with the preset regulation strategy library based on the pressure fault category. The pressure regulation module is used to control the target actuator based on the pressure regulation strategy to regulate the air chamber pressure of the bubble level gauge to obtain the target air pressure; The pressure verification module is used to verify the stability of the target air pressure within a preset observation period and obtain the verification result. If the verification result is that the verification is successful, the current adjustment strategy is maintained and steady-state monitoring is performed. If the verification result is that the verification is unsuccessful, an optimization iteration mechanism is executed.

[0007] A third aspect of this application provides an electronic device, including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor executes the computer program to implement the steps of the above-described high-precision constant flow bubble level gauge chamber pressure regulation method.

[0008] In a fourth aspect of this application, a computer-readable storage medium is provided, the computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the above-described high-precision constant-flow bubble level gauge chamber pressure regulation method.

[0009] The beneficial effects of the high-precision constant-flow bubble level gauge chamber pressure stabilization method and system provided in this application are as follows: This application breaks through the limitations of traditional single-parameter monitoring by integrating environmental data and state parameters of the bubble level gauge. The extracted multi-source feature vectors comprehensively reflect the pressure correlation information, which helps the pre-trained pressure fault diagnosis model identify fault categories and reduces the risk of misjudgment. Based on fault category matching, a dedicated pressure stabilization strategy is used to replace manual intervention, improving response efficiency and regulation targeting. The adjusted target pressure is verified and iteratively optimized during the observation period to form a closed-loop control, which not only avoids repeated adjustments but also ensures the long-term stability of the target pressure in the chamber. This application improves the accuracy and continuity of water level measurement data while reducing the operation and maintenance cost of the bubble level gauge and enhancing its adaptability in complex environments. Attached Figure Description

[0010] Figure 1 A schematic flowchart of a method for regulating the pressure of a high-precision constant-flow bubble level gauge in accordance with an embodiment of this application; Figure 2 A structural block diagram of a chamber pressure regulating system for a high-precision constant flow bubble level gauge provided in an embodiment of this application; Figure 3 This is a schematic block diagram of an electronic device provided in an embodiment of this application. Detailed Implementation

[0011] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0012] To make the purpose, technical solution, and advantages of this application clearer, the following will be described in conjunction with the appendix. Figure 1-3 The following is an explanation using specific examples.

[0013] Please refer to Figure 1 , Figure 1 This is a flowchart illustrating a method for regulating the pressure of a high-precision constant-flow bubble level gauge chamber according to an embodiment of this application. The method includes: S101: Acquire multi-source data, including environmental data and status parameters of the bubble level gauge.

[0014] In this embodiment, the environmental data of the bubble level gauge includes multiple dimensions, including external environmental parameters of the water level monitoring point, such as ambient temperature, humidity, and atmospheric pressure. These data affect the air pressure loss during the bubble transmission process.

[0015] The status parameters of the bubble level gauge include the air chamber pressure, gas flow rate in the air pipe, bubble generation frequency, sensor power supply voltage, and signal strength of the data transmission module. To ensure the accuracy of data acquisition, a time-division synchronous acquisition mechanism is used. For example, acquisition intervals are set for parameters that change rapidly, such as air chamber pressure and gas flow rate; while relatively stable parameters such as ambient temperature and power supply voltage are acquired at 5-second intervals. Simultaneously, outlier removal, missing value completion, and data standardization are performed on the acquired multi-source data to convert data of different dimensions to a unified range. After data acquisition, the data is uploaded to the data center via industrial Ethernet or 4G / 5G communication, and simultaneously cached and backed up locally.

[0016] S102: Extract and fuse features from multi-source data to obtain multi-source feature vectors.

[0017] In this embodiment, after extracting the basic time-domain features such as mean, variance, and peak value of the continuous time-series data such as the air chamber pressure and atmospheric pressure of the bubble level gauge, it is also necessary to mine the frequency-domain features such as the main frequency and the proportion of spectral energy through Fourier transform to capture the periodic patterns and abnormal sudden change signals of pressure fluctuations.

[0018] For example, data preprocessing is first performed, using a sliding window technique to divide the time-series data into blocks, with the window size set to 5-10 minutes. The window size is adjusted according to the monitoring accuracy requirements. Missing values ​​within the window are filled using linear interpolation, and impulse noise is removed using the 3σ criterion. Time-domain feature extraction involves simultaneously calculating basic indicators such as mean, variance, peak value, trough value, and maximum frequency for each window. The mean and variance reflect the overall pressure level and fluctuation range, the peak and trough values ​​capture extreme pressure states, and the maximum frequency counts the number of pressure peaks occurring per unit time, providing a basis for judging pressure stability.

[0019] Frequency domain feature extraction is accomplished using the Fast Fourier Transform (FFT) algorithm. Specifically, the preprocessed window data undergoes an FFT to map time-domain features to frequency-domain features. Then, spectral analysis is used to extract features such as the dominant frequency (the frequency component with the highest energy in the spectrum), the spectral energy ratio (the ratio of the energy of each frequency component to the total energy), and spectral entropy. The dominant frequency represents the periodicity of pressure fluctuations, while the spectral energy ratio and spectral entropy can identify spectral distortions caused by abrupt signal changes. Finally, the time-domain and frequency-domain features are fused into a unified feature vector.

[0020] After single-source feature extraction, multi-dimensional features need to be integrated through a combination of weighted fusion and attention mechanisms. Specifically, a feature importance assessment model is constructed based on historical fault diagnosis data, and the random forest algorithm is used to calculate the contribution weight of each feature to the fault identification result. For example, the weight of the air chamber pressure fluctuation feature is greater than that of the environmental humidity feature, making key information dominate in the fusion. Subsequently, an attention mechanism is used to perceive the feature correlation of the data scene, increasing the attention of features strongly correlated with air pressure faults, such as air chamber pressure peak deviation and gas flow mutation, while weakening the interference of secondary features such as slow changes in environmental temperature. Finally, the features and their weights are fused and concatenated to obtain a multi-source feature vector.

[0021] S103: Input the multi-source feature vectors into the pre-trained barometric pressure fault diagnosis model to obtain the barometric pressure fault category.

[0022] In this embodiment, the air pressure fault diagnosis model uses a deep convolutional neural network as its core architecture. After training with labeled samples composed of historical data and adjusting with real-time data, it forms a stable ability to identify the characteristics of various air pressure faults and can distinguish the feature differences of different fault types such as air chamber leakage, sensor drift, and airway blockage.

[0023] Before feature input, the multi-source feature vectors undergo format adaptation and standardization. For example, the fused feature vectors are adjusted to a tensor format of [batch_size, feature_dim, 1], and the feature values ​​are compressed to the [0, 1] interval using Min-Max normalization to eliminate the interference of differences in feature dimensions on the inference of the barometric pressure fault diagnosis model. Next, the multi-source feature vectors are fed into the input layer of the barometric pressure fault diagnosis model, where they undergo feature enhancement and dimensionality compression through convolutional and pooling layers. The convolutional layers extract local correlation information from the feature vectors using 3×1 convolutional kernels, such as the coupling features between chamber pressure fluctuations and gas flow rate changes; the pooling layers retain key features and reduce computational load. Subsequently, a fully connected layer is used to implement nonlinear mapping. Specifically, the high-dimensional feature vectors are transformed into low-dimensional vectors corresponding to the fault categories, and the probability distribution of various barometric pressure faults is output through a softmax activation function. The barometric pressure fault diagnosis model is trained using the cross-entropy loss function and the Adam optimizer, with an initial learning rate of 0.001, a batch size of 32, and 100 epochs of iterative training. Early stopping is performed using a validation set to prevent overfitting. For example, given a feature vector, after the barometric pressure fault diagnosis model outputs the probabilities of slight air leakage in the air chamber (0.92), sensor drift (0.05), and tracheal obstruction (0.03), the slight air leakage in the air chamber, with the highest probability value, is determined as the final fault category.

[0024] S104: Based on the gas pressure fault category, match the corresponding pressure stabilization and regulation strategy in the preset regulation strategy library.

[0025] In this embodiment, the preset adjustment strategy library is not a static set of rules, but a knowledge base built based on more than a thousand sets of fault handling cases, equipment operating parameters and environmental impact data. It is constructed using a four-dimensional storage structure of fault category-fault degree-environmental parameters-adjustment strategy.

[0026] The regulation strategy library pre-sets differentiated regulation logic for different gas pressure fault categories. For example, for the same type of gas chamber leakage fault, the pressure stabilization regulation strategy for a minor leakage with a leakage rate of less than 0.01 MPa / min is to reduce the start / stop threshold of the air pump by 0.02 MPa, shorten the pressure monitoring interval to 1 second, and maintain pressure stability through high-frequency pressure compensation. For a moderate leakage with a leakage rate greater than or equal to 0.01 MPa / min but less than 0.05 MPa / min, the strategy is to increase the output pressure of the air pump by 15% and activate dual pressure monitoring. For a severe leakage with a leakage rate greater than or equal to 0.05 MPa / min, an emergency strategy is triggered to shut down the main gas circuit and activate the backup gas chamber, while simultaneously notifying maintenance personnel. In addition to the characteristics of the fault itself, the regulation strategy library also has environmental adaptive rules. For example, in high-temperature environments where the temperature is greater than the second temperature threshold, the gas expansion coefficient increases, and the regulation parameters corresponding to the same fault will reduce the pressure compensation by 10%.

[0027] This embodiment employs a two-stage matching mechanism of indexing and optimization. The first stage uses fault category tags for indexing. For example, it matches the diagnostic result of sensor drift +0.03MPa with the corresponding tag in the adjustment strategy library to quickly identify the pressure stabilization strategy, which involves calibrating the sensor zero point and lowering the target pressure of the air pump by 0.03MPa. The second stage is the optimization phase. Based on real-time data obtained from the data acquisition phase, such as atmospheric pressure, air chamber temperature, and air pump operating load, the parameters of the pressure stabilization strategy are corrected. For example, taking a minor airway blockage fault in a low-temperature environment as an example, the pressure stabilization strategy would be to start the air pump for low-pressure purging. After optimization, this would be adjusted to increase the purging pressure by 20%, extend the purging time to 5 seconds, and repeat the process every 30 seconds to offset the reduced gas flow caused by low temperature.

[0028] S105: Based on the pressure stabilization and regulation strategy, the target actuator is controlled to regulate the air chamber pressure of the bubble level gauge to obtain the target air pressure.

[0029] In this embodiment, the abstract pressure stabilization strategy is transformed into precise mechanical actions, achieving stable air chamber pressure through the coordinated operation of multiple actuators. The target actuator is not a single device, but a coordinated execution system with an air pump, an electromagnetic exhaust valve, and a proportional pressure regulating valve as the core, supplemented by pressure sensors and flow sensors. The air pump is responsible for actively replenishing air and increasing pressure, the exhaust valve undertakes the function of depressurizing, and the proportional pressure regulating valve realizes the regulation of the air chamber pressure of the bubble level gauge. The three work together to form a full-range regulation capability of pressure increase-stabilization-depressurization.

[0030] The adjustment process operates in a closed-loop mode of strategy analysis, instruction generation, action execution, and real-time feedback. First, the received pressure stabilization strategy is structurally analyzed to extract key control parameters. For example, for the adjustment strategy addressing slight air leakage in the air chamber, indicators such as the target air pressure (0.3 MPa), the air pump start threshold (0.28 MPa), the air replenishment time (0.5 s), and the allowable pressure fluctuation range (±0.01 MPa) are decomposed and converted into electrical signal commands for each actuator. These commands are then sent to the target actuators. The target actuators, based on the electrical signal commands and their own functional divisions, coordinate their actions to adjust the air chamber pressure of the bubble level gauge to obtain the target air pressure.

[0031] In addition, the bubble level gauge is protected during the adjustment process according to the protection mechanism. Specifically, there are three aspects: First, pressure over-range protection: when the air chamber pressure is detected to be more than 1.5 times the target air pressure, regardless of the current adjustment stage, the power supply to the air pump is immediately cut off and the exhaust valve is opened to release pressure at full speed until the pressure returns to the safe range. Second, actuator fault detection: the operating current of the air pump and valve is monitored by a current sensor. If an abnormal increase in current (blocking) or no current (shutdown) occurs, the system immediately switches to the backup actuator and triggers an alarm. Third, extreme environment adaptation: in special environments such as high temperature and low temperature, the PID parameters are automatically corrected. For example, in low temperature environments, the gas density increases, and the air pump running time will be appropriately extended to compensate for pressure loss. When the pressure sensor collects three consecutive data points showing that the air chamber pressure is stable within the target air pressure ±0.01MPa range, the adjustment is considered complete. Low temperature refers to a temperature environment with a temperature lower than the first temperature threshold.

[0032] S106: Within the preset observation period, the stability of the target air pressure is verified and the verification result is obtained; if the verification result is successful, the current adjustment strategy is maintained and steady-state monitoring is carried out; if the verification result is unsuccessful, the optimization iteration mechanism is executed.

[0033] In this embodiment, the duration of the observation period is set according to the principle of fault association + scenario adaptation. The principle of fault association + scenario adaptation is to set the duration based on the severity of the air pressure fault of the bubble level gauge, the complexity of the adjustment strategy, and the intensity of environmental interference. For example, for easily correctable faults such as slight sensor drift, the observation period is set to 30 seconds after adjustment; while for complex scenarios such as moderate air leakage in the air chamber and low temperature environment, the observation period needs to be extended to 2 minutes. The observation period needs to be sufficiently longer than the period during which the air pressure is affected by environmental fluctuations.

[0034] A high-frequency pressure sensor was used, with a sampling interval of 50ms, to collect the target pressure value. Based on the collected pressure values, a pressure change curve for the observation period was constructed. Then, tiered verification was conducted from two core dimensions: stability and adaptability. Specifically, stability verification focused on the basic stability of the pressure, achieved by calculating three indicators: the standard deviation, range, and fluctuation frequency of the pressure data during the observation period. For example, if the standard deviation was less than 0.02MPa, the range was less than ±2% of the target pressure, and there were no sudden changes greater than ±0.01MPa for 10 consecutive seconds, the stability indicator was considered to be met. Adaptability verification simulated interference scenarios in actual operation, using a water pressure sensor to capture water level changes and monitoring the air pressure anti-interference capability of the bubble level gauge. If the pressure recovered to the target range within 2 seconds after the interference occurred and the fluctuation amplitude was less than ±0.005MPa, the adaptability indicator was considered to be met. Only when both indicators met the requirements could the verification result be considered passed. At this point, the current pressure stabilization and regulation strategy was maintained. Simultaneously, a steady-state monitoring mode is implemented, adjusting the air pressure sampling interval from 50ms to 1s to reduce equipment energy consumption while ensuring monitoring accuracy. If either of the above two indicators fails to meet the requirements, the verification result is deemed unsuccessful, triggering an optimization iteration mechanism. This optimization iteration mechanism is based on the core logic of problem localization, parameter optimization, and secondary adjustment.

[0035] For example, the entire adjustment process is first retrieved, including raw multi-source data, multi-source feature vectors, fault diagnosis results, pressure regulation strategies and their parameters, etc. A comparative analysis method is used to pinpoint the root cause of the problem. For instance, if the air pressure value is consistently lower than the target value while the air pump current is normal, the root cause is that airway resistance was not considered during strategy matching; if air pressure fluctuations are synchronized with changes in ambient temperature, the root cause is insufficient compensation of environmental parameters. Subsequently, corresponding optimization schemes are generated based on different root causes. Specifically, for problems involving diagnostic deviations, data is fed back to the fault diagnosis model for parameter adjustment; for problems involving pressure regulation strategy parameters, the strategy matching unit recalculates the adjustment parameters, such as increasing the air pump output pressure by 10% or adjusting the PID integral coefficient. After optimization, a new round of adjustment and verification process is initiated. If the iteration process exceeds the preset execution threshold and verification is still not passed, a tiered alarm will be triggered. The tiered alarms include: Level 1 alarm information will be pushed to the operation and maintenance management system. The alarm information includes the fault occurrence time, equipment number, water level gauge air chamber location, preceding fault type, iteration optimization record, current air pressure data curve, and details of unmet indicators, such as a standard deviation of 0.03MPa in stability verification and an interference recovery time of 3s in adaptability verification. Based on the triggering of Level 1 alarm, Level 2 alarm will be triggered simultaneously. The reminder will be strengthened by sending SMS messages to the preset operation and maintenance manager and on-duty personnel, and by pushing real-time messages to the operation and maintenance APP. The SMS content includes the water level gauge air pressure adjustment alarm icon, and marks the equipment number, location, alarm time, unmet indicators, and emergency contact person and telephone number. If the alarm is not confirmed and handled within 30 minutes, the reminder will be pushed repeatedly to the backup operation and maintenance personnel to avoid omission. When a graded alarm is triggered, an emergency pressure stabilization mode is activated. This mode locks in the current optimal adjustment parameters, such as reducing the air pump response delay from 100ms to 50ms to improve response sensitivity, and temporarily adjusting the allowable air pressure fluctuation range from ±0.005MPa to ±0.01MPa to ensure the continuity and effectiveness of the water level gauge data. This emergency mode will continue to run until maintenance personnel have completed on-site handling and manually reset the system.

[0036] As can be seen from the above, this application, by integrating environmental data and state parameters from the bubble level gauge, overcomes the limitations of traditional single-parameter monitoring. The extracted multi-source feature vectors comprehensively reflect the pressure correlation information, which helps the pre-trained pressure fault diagnosis model identify fault categories and reduces the risk of misjudgment. Based on fault category matching, a dedicated pressure stabilization strategy is used to replace manual intervention and improve response efficiency and adjustment targeting. The adjusted target pressure is verified and iteratively optimized during the observation period to form a closed-loop control, which not only avoids repeated adjustments but also ensures the long-term stability of the target pressure in the air chamber. This application improves the accuracy and continuity of water level measurement data while reducing the operation and maintenance costs of the bubble level gauge and enhancing its adaptability in complex environments.

[0037] In one embodiment of this application, multi-source feature vectors are input into a pre-trained barometric pressure fault diagnosis model to obtain barometric pressure fault categories, including: Multi-source feature vectors are input into the input layer of the barometric pressure fault diagnosis model; By performing nonlinear transformation and feature abstraction on multi-source feature vectors through the hidden layer of the air pressure fault diagnosis model, deep-level fault features are obtained. The output layer of the barometric pressure fault diagnosis model outputs the probability distribution of barometric pressure fault categories based on deep-level fault characteristics. Based on probability distribution, the type of air pressure failure is determined.

[0038] In this embodiment, a dual standardization process is required before input: First, format adaptation, converting the fused multi-source feature vectors into a model-recognizable three-dimensional tensor format of [batch_size, feature_dim, 1], where feature_dim is the feature dimension, and the dimension adjustment is determined by the number of features such as air chamber pressure and gas flow rate. 1 represents single-channel input, ensuring that the data can be effectively received by the input layer neurons. Second, data distribution adaptation, compressing all feature values ​​to the [0,1] interval using the Min-Max normalization algorithm, eliminating numerical differences between features of different dimensions such as air chamber pressure and bubble frequency, and avoiding the gas pressure fault diagnosis model from being overly sensitive to large numerical features. Next, the feature vectors are fed into the input layer of the gas pressure fault diagnosis model in parallel transmission, and the input layer neurons perform initial data transmission through a linear activation function.

[0039] The hidden layer of the gas pressure fault diagnosis model is the core region for feature deepening processing. Through nonlinear transformation and feature abstraction, it transforms the input shallow multi-source feature vectors into deep-level fault features that can characterize fault modes. This hidden layer adopts a composite architecture of convolutional layer + pooling layer + fully connected layer. Specifically, the convolutional layer is configured with 3×1 one-dimensional convolutional kernels to perform sliding convolution operations on temporal feature vectors, capturing the coupled correlation features between gas chamber pressure fluctuations and gas flow rate changes, such as the key pattern of synchronicity between pressure drop and flow rate increase when the gas chamber leaks. Through parallel operations of multiple sets of convolutional kernels, multi-dimensional local feature maps are generated. The pooling layer adopts a max pooling strategy, which compresses the feature dimension by 50% while retaining key feature information, reducing the computational load of the gas pressure fault diagnosis model and avoiding overfitting. The fully connected layer introduces the ReLU nonlinear activation function to break the linear correlation between features, achieving deep integration of the output features of the convolutional layer. For example, local features such as pressure fluctuation period and flow rate change amplitude are abstracted into high-level fault features such as gas chamber sealing performance and sensor response status.

[0040] The core function of the output layer of the barometric pressure fault diagnosis model is to output the probability distribution of barometric pressure fault categories based on deep-level fault characteristics, providing a quantitative basis for the final fault determination. Specifically, the output layer adopts a fully connected structure, with the number of neurons perfectly matching the number of preset barometric pressure fault categories. These preset categories include core faults such as air chamber leakage, sensor drift, and tracheal obstruction, as well as normal states. The feature vectors output by the fully connected layer are converted into probability values ​​using a softmax activation function. This activation function ensures that the sum of the probabilities of all fault categories is 1, and the probability value of each fault category is within the range [0,1], representing the confidence level of the barometric pressure fault diagnosis model for each type of fault. For example, for a certain detection scenario, the output layer outputs the probability distribution results for slight air chamber leakage (0.91), sensor drift (0.04), tracheal obstruction (0.03), and normal state (0.02). The barometric pressure fault category is determined based on the probability distribution. The probability distribution results determine the confidence level of the barometric pressure fault diagnosis model in identifying that fault category.

[0041] As can be seen from the above, this embodiment achieves accurate transformation from multi-source feature vectors to fault categories through the coordinated operation of the input layer, hidden layer, and output layer of the barometric pressure fault diagnosis model. The nonlinear transformation and feature abstraction capabilities of the hidden layer can deeply mine the fault correlation information hidden in the data, breaking through the bottleneck of traditional linear analysis methods. The output layer, based on probability distribution, provides a basis for the scientific determination of fault categories, avoids errors in subjective judgment, and improves the accuracy and reliability of the diagnostic results.

[0042] In one embodiment of this application, determining the type of barometric pressure fault based on a probability distribution includes: If the probability value of a single fault category in the probability distribution is greater than or equal to the preset judgment threshold, then the single fault category will be regarded as the air pressure fault category. If the highest probability value in the probability distribution is less than the judgment threshold, and there are two or more fault categories with probability values ​​greater than or equal to the preset auxiliary threshold, and this probability distribution feature continues to exist in subsequent consecutive detection cycles, then it is judged as a composite fault category. If the probability values ​​of all fault categories in the probability distribution are less than the auxiliary threshold, then the fault is determined to be an unknown fault.

[0043] In this embodiment, if the probability value of a single fault category in the probability distribution is greater than or equal to a preset judgment threshold, then the single fault category is regarded as a pressure fault category. The judgment threshold needs to be adjusted according to the severity of the fault category and the accuracy of the diagnostic model. For example, for fault categories such as air chamber leakage that directly affect measurement accuracy, the judgment threshold is set to 60%, while for indirect faults such as sensor data transmission delay, the judgment threshold is set to 50%.

[0044] If the highest probability value in the probability distribution is less than the judgment threshold, and there are two or more fault categories with probability values ​​greater than or equal to the preset auxiliary threshold, and this probability distribution characteristic persists in subsequent consecutive detection cycles, then it is determined to be a composite fault category. The auxiliary threshold is usually set to 50%-60% of the judgment threshold to avoid misjudgment; the number of consecutive detection cycles needs to be determined according to the development pattern of the fault category. For example, for common fault categories, three detection cycles are set, with each cycle lasting 10 seconds. This can eliminate probability fluctuations caused by instantaneous interference and promptly capture real composite faults, such as interrelated fault combinations like abnormal tracheal resistance + insufficient ambient temperature compensation.

[0045] If the probability values ​​of all fault categories in the probability distribution are less than the auxiliary threshold, the fault is classified as unknown. This situation may be caused by novel faults not included in the diagnostic model or extreme environmental disturbances, and is therefore classified as an unknown fault.

[0046] As can be seen from the above, this embodiment achieves a clear distinction between single faults, compound faults, and unknown faults by setting judgment thresholds and auxiliary thresholds. The explicit judgment method for single faults improves the matching rate of adjustment strategies; the dual judgment criteria of probability characteristics and duration period for compound faults avoid misjudgments caused by instantaneous data fluctuations and solve the problem of difficult diagnosis in scenarios with multiple faults overlapping; and the identification mechanism for unknown faults reserves an interface for subsequent model optimization, breaking the limitation of traditional diagnostic methods' reliance on known faults and improving adaptability to complex and unknown working conditions.

[0047] In one embodiment of this application, after determining that all fault categories have probability values ​​less than an auxiliary threshold, the method further includes: Based on the air pressure fluctuation range of historical normal data of the bubble level gauge, the target safe range of air chamber pressure of the bubble level gauge is set. The control actuator adjusts the air pressure in the air chamber of the bubble level gauge to obtain the adjusted air pressure and determines whether it is within the target safe range. If the regulating air pressure is within the target safe range, the multi-source data corresponding to the regulating air pressure will be marked as normal data and will not be added to the sample set. If the regulated air pressure is not within the target safe range, or if new unknown faults are continuously generated, the multi-source data and regulation effects recorded during the regulation process will be marked as an unknown fault sample set and temporarily stored. When the amount of data in the unknown fault sample set reaches a set size or a fixed synchronization period, the unknown fault sample set is used to incrementally learn the air pressure fault diagnosis model.

[0048] In this embodiment, the target safe range for the air chamber pressure is first set based on the historical normal data of the bubble level gauge. Historical normal data refers to a dataset where the bubble level gauge is in a stable operating state, with no fault records and environmental interference within the normal range. The 95% confidence interval of the pressure fluctuation in this dataset is used as the base interval. Then, boundary corrections are made according to the factory accuracy requirements of the bubble level gauge to finally form the target safe range. For example, the historical normal data pressure fluctuation range of a certain model of bubble level gauge is 0.1MPa-0.15MPa. Based on its factory accuracy of ±0.008MPa, the target safe range is set to 0.098MPa-0.152MPa, ensuring that the target safe range includes the normal fluctuation range while reserving a reasonable safety margin. During the pressure adjustment process, the air pump, solenoid valve, and other actuators are controlled to adjust the air chamber pressure of the bubble level gauge, obtaining the adjusted actual pressure value (i.e., the adjusted pressure), and comparing it with the preset target safe range in real time.

[0049] Specifically, during the air pressure regulation process, the air pump and solenoid valve are controlled to work in coordination based on the target safe range, according to the results of previous fault diagnosis and the current environmental parameters. For example, when the high-frequency pressure sensor detects that the air pressure in the air chamber is less than the lower limit of the target safe range, the air pump starts to inflate according to the set output pressure. At this time, the solenoid valve remains closed to ensure the air chamber is sealed. As the air pressure rises to near the target value, the air pump reduces its output power to a stable operating mode. If the air pressure is detected to be greater than the upper limit of the target safe range, the air pump stops working, and the solenoid valve opens quickly to release pressure. When the air pressure drops to within the target safe range, the solenoid valve closes quickly. At the same time, the PID regulation algorithm is used to correct the start and stop frequency of the air pump and the opening and closing duration of the solenoid valve in real time, so that the air pressure in the air chamber is stabilized within the target range.

[0050] Secondly, based on the comparison results between the regulated air pressure and the target safe range, a differentiated data labeling strategy is implemented: if the regulated air pressure is stable within the target safe range, for example, if it is stable within the target safe range for 5 consecutive collection cycles, then the current adjustment effect is determined to meet the requirements, and the multi-source data corresponding to the regulated air pressure is marked as normal data, and such data is not included in the sample; if the regulated air pressure does not fall into the target safe range, or if new unknown faults are continuously generated after adjustment, that is, if unknown faults are determined for 2 consecutive detection cycles, then the multi-source data and adjustment effect data recorded throughout the adjustment process, such as the air pressure change curve before and after adjustment, the actuator, the fault judgment basis, etc., are uniformly marked as unknown fault samples and temporarily stored in the local cache database.

[0051] In this embodiment, the size threshold of the unknown fault sample set is set based on the optimal data volume for a single learning iteration of the barometric pressure fault diagnosis model, with a recommended value of 50-100 samples. This avoids overfitting due to too few samples or decreased learning efficiency due to too many samples. The fixed synchronization period is determined based on the equipment operating frequency, typically set to 7 days, ensuring that the barometric pressure fault diagnosis model can be updated regularly even if the number of unknown fault samples is insufficient in the short term. When any triggering condition is met, the locally stored unknown fault sample set is encrypted and uploaded to the barometric pressure fault diagnosis model for training. Subsequently, after removing duplicate data and labeling key features from the unknown fault samples, an incremental learning algorithm is used to update the parameters of the barometric pressure fault diagnosis model. During the learning process, the core parameters of the model that have been trained to maturity need to be frozen, and only the fault identification boundary and feature weights are adjusted. For example, the adjustment of the fault identification boundary is based on the feature distribution of the unknown fault samples and is achieved using cluster analysis + boundary migration. Specifically, the K-means clustering algorithm is used to cluster the labeled unknown fault samples and historical fault samples based on their features, thus identifying the clustering regions of unknown fault samples in the feature space. For example, for novel faults such as sensor drift at high temperatures, features such as air pressure fluctuation frequency and temperature correlation coefficient will form independent clusters. Then, the minimum feature distance between this cluster and existing fault categories is calculated. Using 1.2 times the distance value as a boundary safety redundancy, the original fault identification boundary is migrated towards the new cluster, forming a new identification boundary. Simultaneously, to avoid category confusion caused by excessive boundary migration, cross-validation is used to verify the adjustment effect, ensuring that the identification accuracy of historical fault samples under the new boundary remains above 95%.

[0052] Feature weight adjustments follow the principle of contribution quantification and weighting, optimizing feature importance based on the core requirements of air pressure fault diagnosis. A random forest algorithm is used to calculate the contribution of each feature to the identification of unknown faults, selecting the top three contributing key features, such as air pressure fluctuation amplitude, ambient temperature change rate, and air pump current fluctuation value. Their original weights are increased by 20%-30%; for example, if the original weight of air pressure fluctuation amplitude was 0.25, it is increased to 0.32 after adjustment. For secondary features with a contribution of less than 0.05, such as equipment operating time, the original weight is maintained or appropriately reduced to avoid invalid features consuming the resources of the air pressure fault diagnosis model. After weight adjustment, parameters are fine-tuned using a gradient descent algorithm. These adjusted parameters include the weight coefficients of the feature mapping layer, the input weight matrix of the fault classification layer, and the local parameters of the activation function. For example, for the key feature of pressure surge magnitude, its weight coefficient in the feature mapping layer is adjusted to match the increased feature importance, and the corresponding weight matrix elements in the classification layer are adjusted simultaneously. The specific values ​​of these parameters are then adjusted using the gradient descent algorithm to reduce the loss function value of the barometric pressure fault diagnosis model below a preset threshold under the new feature weight configuration, ultimately achieving high fault identification accuracy. This adjustment improves the barometric pressure fault diagnosis model's ability to identify new types of barometric pressure faults while avoiding compromising its original diagnostic accuracy.

[0053] From the above, it can be concluded that this embodiment sets a safe range by using historical data to ensure that the air chamber pressure can quickly return to a safe range when an unknown fault occurs, thus guaranteeing the basic operation of the bubble level gauge. The differentiated labeling and storage strategy for adjustment data enables the efficient accumulation of useful samples and avoids the waste of data resources. Furthermore, the incremental learning mechanism based on sample size and time period enables the air pressure fault diagnosis model to continuously absorb new data experience, constantly expand the fault identification range, optimize model performance, and give the pressure regulation method the ability to self-evolve.

[0054] In one embodiment of this application, based on the type of air pressure fault, a corresponding pressure stabilization strategy is matched with a preset regulation strategy library, further comprising: Based on the deviation between the multi-source feature vector and the normal operating condition baseline vector, as well as the rate of change of the state parameters, a weighted fusion calculation is performed to determine the comprehensive severity level of the air pressure fault category. Based on the type of air pressure failure and the overall severity level, the corresponding pressure stabilization strategy is matched from the preset adjustment strategy library.

[0055] In this embodiment, a weighted fusion calculation is performed based on the deviation between the multi-source feature vector and the normal operating condition baseline vector, as well as the rate of change of the state parameters, to determine the comprehensive severity level of the air pressure fault category. The normal operating condition baseline vector is generated from historical normal operating data of the equipment through statistical modeling. The deviation is calculated using a cosine similarity algorithm; a deviation closer to 1 indicates a more significant fault characteristic. The rate of change of state parameters focuses on the real-time changing trends of key parameters such as air pressure and air pump power. The sliding window method is used to calculate the change amplitude per unit time; a larger rate of change indicates a more rapid fault development.

[0056] When calculating the weighted fusion of the deviation between the multi-source feature vector and the normal operating condition baseline vector, as well as the rate of change of the state parameters, the deviation and the rate of change need to be normalized to obtain the standardized deviation and the standardized rate of change. Then, the standardized deviation and the standardized rate of change are weighted and fused based on the preset first weight coefficient and the second weight coefficient to obtain the comprehensive severity score. Finally, the comprehensive severity score is mapped to the preset severity score range to obtain the comprehensive severity level.

[0057] This embodiment's adjustment strategy library constructs a two-dimensional mapping relationship based on fault type and severity level, formulating differentiated solutions for single faults and compound faults. For example, a single fault with slight sensor drift is matched with a strategy of slightly correcting sensor calibration parameters + a short observation period of 30 seconds; a moderate air leakage in the air chamber is matched with a strategy of increasing the air pump output pressure by 15% + adjusting the PID proportional coefficient + a 2-minute observation period; a compound fault with abnormal tracheal resistance + insufficient ambient temperature compensation is matched with a strategy of clearing the tracheal passage + optimizing the temperature compensation algorithm + high-frequency start-stop adjustment of the air pump + a 5-minute long observation period. During strategy matching, mature strategies with a historical matching success rate greater than 90% are prioritized. If multiple suitable strategies exist, the optimal solution is selected and triggered by calculating the energy consumption cost and adjustment accuracy after strategy execution.

[0058] As can be seen from the above, this embodiment achieves the judgment of the degree of fault impact by fusing feature deviation and parameter change rate. This dual matching standard of category and level changes the drawback of the traditional one-size-fits-all adjustment strategy, and can match differentiated adjustment schemes for faults of different severity. At the same time, the weighted fusion calculation method makes the adjustment strategy more in line with the actual fault scenario, improving the accuracy and economy of voltage regulation.

[0059] In one embodiment of this application, a weighted fusion calculation is performed based on the deviation between the multi-source feature vector and the normal operating condition reference vector, as well as the rate of change of the state parameters, to determine the comprehensive severity level of the barometric pressure fault category, including: The deviation and the rate of change are normalized to obtain the standardized deviation and the standardized rate of change. The standardized deviation and standardized change rate are weighted and fused based on the preset first weight coefficient and second weight coefficient to obtain the comprehensive severity score; wherein the sum of the first weight coefficient and the second weight coefficient is one. The overall severity score is mapped to a preset severity score range to obtain the overall severity level.

[0060] In this embodiment, the overall severity level of a fault is quantitatively assessed through a standardized process of normalization, weighted fusion, and level mapping. Specifically, the differences in the dimensions of different indicators are first eliminated, then weights are assigned according to the fault characteristics, and finally, the results are transformed into a level standard. The specific operations are as follows: The first step is index normalization, which involves converting the deviation of the multi-source feature vector from the normal operating condition baseline vector and the rate of change of the state parameters into standardized indices of a unified scale, resulting in standardized deviation and standardized rate of change. The deviation is calculated using a cosine similarity algorithm. If the original value is already in the [0,1] range, this range is retained to maintain the physical meaning of the difference between fault characteristics and normal conditions. The closer the value is to 1, the more pronounced the fault characteristics. The rate of change of state parameters has different dimensions depending on the parameter type. For example, the unit of air pressure change rate is MPa / s, and the unit of air pump power change rate is kW / s. The min-max normalization method is used to compress it to the [0,1] range, as shown in the formula: Standardized value = (Original value - Minimum value) / (Maximum value - Minimum value) For example, if the historical extreme values ​​of the air pressure change rate of a certain device are 0 and 0.02 MPa / s, and the real-time change rate is 0.01 MPa / s, then its standardized change rate is (0.01-0) / (0.02-0) = 0.5.

[0061] The second step is weighted fusion calculation. Based on preset first and second weighting coefficients, two standardized indicators are linearly fused to obtain a comprehensive severity score, with the sum of the two weighting coefficients fixed at 1. The weight allocation is adjusted according to the priority of fault handling. For example, when prioritizing the identification of the fault's nature, the first weighting coefficient corresponding to the standardized deviation is set to 0.6, and the second weighting coefficient corresponding to the standardized rate of change is set to 0.4, highlighting the impact of the fault characteristics themselves. When the equipment is in sensitive scenarios such as high-load operation, the second weighting coefficient can be increased to 0.5 to strengthen the consideration of the fault's development speed. The formula for calculating the comprehensive severity score is: The overall severity score = standardized deviation × first weighting coefficient + standardized rate of change × second weighting coefficient.

[0062] The third step is grade mapping, which matches the overall severity score with a preset score range to ultimately determine the overall severity level. Based on the impact of the bubble level gauge's pressure malfunction, the preset range is divided into three levels: a score less than 0.3 corresponds to a minor level, where the fault characteristics are not significant and develop slowly, such as a standardized deviation of 0.2 and a rate of change of 0.1 caused by slight sensor drift, resulting in an overall score of 0.16; a score greater than or equal to 0.3 and less than or equal to 0.7 corresponds to a moderate level, where the fault has manifested but is still controllable, such as a standardized deviation of 0.5 and a rate of change of 0.4 corresponding to slight air leakage in the air chamber, resulting in an overall score of 0.46; a score greater than 0.7 corresponds to a severe level, where the fault characteristics are prominent and develop rapidly, requiring emergency handling, such as a standardized deviation of 0.8 and a rate of change of 0.9 for severe air leakage in the air chamber, resulting in an overall score of 0.84.

[0063] From the above, it can be concluded that this embodiment eliminates the evaluation bias caused by the different dimensions of deviation and rate of change through normalization processing, ensuring the comparability of the two indicators; the weighted fusion method with preset weight coefficients realizes differentiated consideration of different fault influencing factors, which conforms to the objective law that the importance of each parameter is different in actual working conditions; the method of mapping the calculation results to a fixed score interval makes the division of severity level clear and avoids strategy matching errors caused by fuzzy judgment.

[0064] In one embodiment of this application, a method for regulating the pressure of the air chamber of a high-precision constant-flow bubble level gauge further includes: Based on the fuzzy logic control strategy, the standardized deviation and standardized change rate are used as input variables; The reasoning is performed based on the preset fuzzy rule base to obtain the adjustment strategies for the first weight coefficient and the second weight coefficient. The fuzzy rule base includes the weight coefficient adjustment strategies corresponding to different combinations of input variables. By defuzzing the calculation, the adjustment strategy is transformed to obtain the adjustment instructions; Based on the adjustment command, the corresponding weight coefficients are adjusted using the preset first step length or second step length.

[0065] In this embodiment, a fuzzy logic control strategy is used to achieve precise adjustment to adapt to fault characteristics. The adjustment process takes the standardized deviation and standardized change rate as the core inputs and completes weight optimization through a closed-loop process of fuzzification, rule reasoning, defuzzification, and adjustment execution.

[0066] Specifically, the two precise values, standardized deviation and standardized rate of change, are first converted into fuzzy linguistic variables. Based on the requirements of barometric pressure fault assessment, the fuzzy linguistic values ​​of both are divided into three levels: low, medium, and high. Corresponding membership functions are then established. For example, a triangular membership function is used to map precise values ​​to fuzziness. For instance, the low level of standardized deviation corresponds to the interval [0, 0.2, 0.4], the medium level to [0.3, 0.5, 0.7], and the high level to [0.6, 0.8, 1.0]. When the standardized deviation is 0.5, the fuzziness belonging to the medium level is 1, and the fuzziness belonging to the low and high levels is 0. This method quantifies the fuzzy characteristics of the input variables.

[0067] The fuzzy rule base is built based on historical experience in handling barometric pressure faults and a large amount of experimental data. It includes adjustment strategies for all combinations of standardized deviation (representing fault severity) and standardized rate of change (representing the dynamic development trend of the fault). Each combination includes corresponding first and second weight coefficient configurations and a core adjustment objective. For example, if both the standardized deviation and the standardized rate of change are high, the corresponding first and second weight coefficients are 0.55 and 0.45, respectively, with the adjustment objective being to prioritize ensuring adjustment sensitivity for rapid response. If both the standardized deviation and the standardized rate of change are high, the corresponding first and second weight coefficients are 0.65 and 0.35, respectively, with the adjustment objective being to highlight the essential identification of the fault to avoid misadjustment. The specific core rule configuration is shown in Table 1, the standardized deviation-rate of change weight configuration table. Table 1. Weighting of Standardized Deviation-Rate of Change

[0068] Secondly, the adjustment strategy is transformed into specific adjustment instructions through defuzzification calculation. For example, the centroid method is used as the defuzzification algorithm. This method obtains the weight coefficient value by calculating the centroid coordinates of the fuzzy output set, combining advantages in computational efficiency and accuracy. For instance, if the fuzzy set of the inference output has a first weight of approximately 0.55, and its fuzzy distribution interval is [0.50, 0.55, 0.60], corresponding to the triangular membership function distribution of the medium-level fuzzy linguistic values, the centroid method is used to calculate the centroid abscissa of the region enclosed by the membership function curve and the horizontal axis within this region, thus obtaining the first weight coefficient.

[0069] The specific calculation process is as follows: Let the membership function be expressed as f(x) in the interval [a,b,c]. When x∈[a,b], f(x)=(xa) / (ba), When x∈[b,c], f(x)=(cx) / (cb). The centroid coordinates x0 = ∫(x f(x))dx / ∫f(x)dx, Substituting [a=0.50, b=0.55, c=0.60], we get x0=0.55, which is the first weighting coefficient. Similarly, we can obtain the second weighting coefficient. Based on the first and second weighting coefficients, an adjustment instruction is generated.

[0070] Finally, based on the adjustment command, the corresponding weight coefficients are adjusted using a preset first step length or second step length. The step length is determined based on the fluctuation range of the input variable. For example, when the first weight coefficient is 0.6, the target value of the adjustment command obtained after defuzzification is 0.55, and the fluctuation value of the input variable is 0.08 in two consecutive detection cycles, it indicates that the fault state is stable, and the first step length is used to adjust the weight coefficient. If the target value is 0.5 and the fluctuation of the input variable is still smooth, the weight coefficient is adjusted in two steps of 0.05. After each adjustment, the stability of the air pressure in the bubble level gauge is monitored to ensure a smooth transition of the weight. If the fluctuation value of the input variable suddenly increases to 0.12, the second step length, which is greater than the first step length, is used to adjust the weight coefficient. Next, the standardized deviation and standardized change rate are linearly fused based on the adjusted weight coefficient to obtain the comprehensive severity score. The selection principle for the first and second step lengths is as follows: if the fluctuation values ​​of the standardized deviation and the standardized change rate are both less than 0.1 in two consecutive detection cycles, it indicates that the fault state is stable. The first step length of 0.05 is used for adjustment to avoid evaluation bias caused by sudden changes in weights. If the fluctuation value of any variable is greater than or equal to 0.1, it indicates that the fault state changes drastically. The second step length of 0.1 is used for adjustment to ensure that the weight adjustment can respond to the fault in a timely manner.

[0071] From the above, it can be concluded that this embodiment uses standardized deviation and rate of change as input variables and performs inference based on a fuzzy rule base to capture the impact of changes in operating conditions on the importance of each evaluation index; defuzzification calculation transforms the inference results into specific adjustment instructions, ensuring the executability of weight adjustment; and the step-by-step adjustment method ensures both the timeliness of adjustment and avoids the impact of large fluctuations on the evaluation results, making the comprehensive severity level evaluation more in line with real-time operating conditions and further improving the adaptability of the adjustment strategy.

[0072] In one embodiment of this application, a corresponding pressure stabilization strategy is matched from a preset regulation strategy library based on the type and overall severity level of the pressure failure, including: Based on the type of air pressure failure, identify the relevant target actuator; The degree of parameter adjustment of the target actuator is determined based on the comprehensive severity level of the air pressure failure category; The adjustment parameters are corrected based on environmental data.

[0073] In this embodiment, the air chamber pressure stability of the bubble level gauge depends on the synergistic effect of air pipe transmission, air chamber sealing, and constant flow pump air supply. Three types of faults—air pipe blockage, air chamber leakage, and abnormal constant flow pump power—all affect the air chamber pressure. The occurrence mechanisms and affected components of different pressure fault categories are correlated, and the target actuator is determined based on this correlation. For example, when the fault is determined to be air pipe blockage, the core problem is impurities or condensate buildup in the air pipe, and the corresponding target actuator is the pulse purge solenoid valve. If the fault category is air chamber leakage, the focus should be on the air chamber sealing assembly and the air pressure control pump. The target actuator for the fault category of abnormal constant flow pump power is the constant flow pump itself. This targeted matching of the target actuator avoids the wasted energy of ineffective actuators and improves the efficiency of air chamber pressure regulation in the bubble level gauge.

[0074] In this embodiment, the adjustment parameters include one or more of the following: adjustment range, action time, and action frequency. If the gas pressure fault category is pipeline blockage, the relevant target actuators are a constant flow pump and a pulse purging solenoid valve. When the overall severity level is mild, the adjustment strategy is to increase the power of the constant flow pump. When the overall severity level is moderate, the adjustment strategy is to increase the power of the constant flow pump and trigger low-frequency pulse purging. When the overall severity level is severe, the adjustment strategy is to maximize the power of the constant flow pump and trigger high-frequency strong pulse purging.

[0075] If the gas pressure fault category is gas chamber leakage, the relevant target actuators are the constant flow pump and the backup gas circuit valve. When the overall severity level is minor, the adjustment strategy is to slightly increase the power of the constant flow pump to compensate for the leakage. When the overall severity level is moderate, the adjustment strategy is to significantly increase the power of the constant flow pump and operate it intermittently. When the overall severity level is severe, the adjustment strategy is to activate the backup gas circuit and trigger a severe leak alarm.

[0076] If the pressure fault category is sensor drift, the relevant target actuator is the system controller. The adjustment strategy involves initiating a software calibration algorithm and compensating for the sensor readings based on the deviation. The pressure regulation strategy includes the target actuator, the degree of parameter adjustment to the target actuator, and the correction of its parameters.

[0077] From the above, it can be concluded that this embodiment locks the target actuator according to the fault category, avoiding resource waste and adjustment interference caused by the action of irrelevant mechanisms; it determines the adjustment degree based on the severity level, realizing on-demand adjustment and reducing the wear and tear on the equipment caused by over-adjustment; and the parameter correction based on environmental data makes up for the shortcomings of traditional strategies that ignore environmental influences, enabling the adjustment parameters to adapt to environmental changes such as temperature and humidity, improving the stability of the bubble level gauge's adjustment effect under different environmental conditions, while also improving the bubble level gauge's environmental adaptability and adjustment accuracy.

[0078] In one embodiment of this application, the adjustment parameters are corrected based on environmental data, including: If the ambient temperature is less than the preset first temperature threshold, the single action time of the pulse purging solenoid valve is increased in a positive correlation with the negative deviation between the ambient temperature and the first temperature, and its action frequency is reduced in a negative correlation. If the ambient temperature is greater than the preset second temperature threshold, the adjustment range of the constant flow pump power will be reduced in a negative correlation manner based on the positive deviation between the ambient temperature and the second temperature threshold. If the ambient humidity is greater than the preset humidity threshold, a preventive pulse purging task will be performed based on the idle period of the bubble level gauge after the voltage regulation strategy is executed.

[0079] In this embodiment, ambient temperature and humidity data are key variables affecting the pressure regulation effect of the bubble level gauge chamber. The actuator action parameters are adjusted based on the temperature and humidity data.

[0080] For example, when the ambient temperature is lower than the preset first temperature threshold, it is considered a low-temperature environment. Low temperatures cause water vapor in the trachea to condense into ice, increasing the risk of blockage. In this case, the operating parameters of the pulse purging solenoid valve need to be adjusted bidirectionally. The adjustment logic is based on the negative temperature deviation value, which is the difference between the first temperature threshold and the actual ambient temperature. The larger the negative deviation value, the higher the probability of tracheal icing. For example, on the one hand, the single-action time is increased in a positive correlation manner. When the negative deviation value is 3℃, the single purging time is extended from the standard 0.5 seconds to 0.8 seconds; when the negative deviation value reaches 5℃, it is extended to 1.2 seconds, ensuring the blockage removal effect by strengthening the single purging force. On the other hand, the action frequency is reduced in a negative correlation manner. For example, the standard frequency is once every 30 seconds, adjusted to once every 45 seconds when the negative deviation is 3℃, and once every 60 seconds when the negative deviation is 5℃, forming a balanced adjustment mode of strengthening the single-action effect and reducing the action frequency. The first temperature threshold is set to 5℃, but it can also be set according to the equipment operation manual.

[0081] When the ambient temperature exceeds the preset second temperature threshold, it is considered a high-temperature environment. High temperatures reduce the heat dissipation efficiency of the constant flow pump's internal components, and maintaining the original adjustment range could easily lead to overheating and damage to the pump. In this case, the power adjustment range of the constant flow pump needs to be negatively corrected based on the positive temperature deviation (the difference between the actual temperature and the second temperature threshold). Specifically, when the positive deviation is 5℃, the original adjustment range is reduced by 30%; when the positive deviation reaches 10℃, the reduction is increased to 50%. The second temperature threshold is typically set to 40℃, but can also be set according to the high-temperature operating limits of the constant flow pump.

[0082] When the ambient humidity exceeds the preset humidity threshold, a preventative pulse purging task is added after the pressure stabilization strategy has been completed. The task execution relies on the equipment's idle time period identification module to ensure that normal water level measurement is not affected—the idle time period typically refers to the middle of the interval between two water level data acquisitions (generally 1-2 minutes). The purging parameters need to be different from the fault handling mode. The single purging time is set to 1.5 times the standard fault purging time (e.g., if the standard is 0.5 seconds, adjust to 0.75 seconds), and the purging frequency is once every two measurement cycles. Through gentle and regular purging actions, condensate adhering to the inner wall of the pipe is promptly removed, preventing its accumulation and blockage or impact on the accuracy of air pressure transmission. This constructs a proactive prevention and fault reduction protection system. The humidity threshold is determined comprehensively based on the anti-condensation requirements of the bubble level gauge's air path, the typical humidity range of the equipment's operating environment, and the air pressure accuracy requirements of constant flow bubble transmission. It can also be determined by referring to the recommended values ​​in the equipment's technical manual.

[0083] From the above, it can be concluded that this embodiment, by adjusting the solenoid valve's action with increased time and reduced frequency in low-temperature environments, ensures both the purging effect and avoids the wear and tear caused by frequent equipment start-ups and shutdowns; in high-temperature environments, reducing the constant flow pump power adjustment range prevents large pressure fluctuations due to temperature sensitivity; and in high-humidity environments, performing preventative purging proactively avoids the risk of water vapor condensation blocking the gas path. These targeted correction measures enable the pressure stabilization strategy to allow the bubble level gauge to actively adapt to different environmental conditions, further improving the stability of the bubble level gauge's gas chamber pressure and extending the equipment's service life.

[0084] Corresponding to the air chamber pressure regulation method of the high-precision constant flow bubble level gauge in the above embodiment, Figure 2 This is a structural block diagram of a chamber pressure stabilization and regulation system for a high-precision constant-flow bubble level gauge provided in one embodiment of this application. For ease of explanation, only the parts relevant to the embodiment of this application are shown. References Figure 2 The air chamber pressure regulation system 20 of the high-precision constant flow bubble level gauge includes: a data acquisition module 21, a feature fusion module 22, a fault diagnosis module 23, a strategy matching module 24, a pressure regulation module 25, and a pressure verification module 26.

[0085] in, Data acquisition module 21 is used to acquire multi-source data, including environmental data of the bubble level gauge and status parameters of the bubble level gauge; Feature fusion module 22 is used to extract and fuse features from multi-source data to obtain multi-source feature vectors; The fault diagnosis module 23 is used to input multi-source feature vectors into the pre-trained barometric pressure fault diagnosis model to obtain the barometric pressure fault category. The strategy matching module 24 is used to match the corresponding pressure stabilization and regulation strategy with the preset regulation strategy library based on the pressure fault category. The pressure regulation module 25 is used to control the target actuator based on the pressure stabilization regulation strategy to regulate the air chamber pressure of the bubble level gauge and obtain the target air pressure. The pressure verification module 26 is used to verify the stability of the target air pressure within a preset observation period and obtain the verification result. If the verification result is that the verification is successful, the current adjustment strategy is maintained and steady-state monitoring is performed. If the verification result is that the verification is unsuccessful, the optimization iteration mechanism is executed.

[0086] See Figure 3 , Figure 3 This is a schematic block diagram of an electronic device provided according to an embodiment of this application. Figure 3 The electronic device 300 in this embodiment may include one or more processors 301, one or more input devices 302, one or more output devices 303, and one or more memories 304. The processors 301, input devices 302, output devices 303, and memories 304 communicate with each other via a communication bus 305. The memories 304 store computer programs, including program instructions. The processors 301 execute the program instructions stored in the memories 304. Specifically, the processors 301 are configured to invoke the program instructions to perform the functions of the modules in the aforementioned device embodiments, for example... Figure 2 The functions of the data acquisition module 21, feature fusion module 22, fault diagnosis module 23, strategy matching module 24, pressure regulation module 25, and pressure verification module 26 are shown.

[0087] It should be understood that, in the embodiments of this application, the processor 301 may be a central processing unit (CPU), or it may be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor.

[0088] Input device 302 may include a touchpad, a fingerprint sensor (for collecting the user's fingerprint information and fingerprint orientation information), a microphone, etc., and output device 303 may include a display (LCD, etc.), a speaker, etc.

[0089] The memory 304 may include read-only memory and random access memory, and provides instructions and data to the processor 301. A portion of the memory 304 may also include non-volatile random access memory. For example, the memory 304 may also store device type information.

[0090] In specific implementations, the processor 301, input device 302, and output device 303 described in the embodiments of this application can execute the implementation methods described in any embodiment of the high-precision constant current bubble level gauge air chamber pressure regulation method provided in the embodiments of this application, or they can execute the implementation methods of the electronic devices described in the embodiments of this application, which will not be repeated here.

[0091] In another embodiment of this application, a computer-readable storage medium is provided. This computer-readable storage medium stores a computer program, which includes program instructions. When executed by a processor, the program instructions implement all or part of the processes in the methods described above. Alternatively, the computer program can instruct related hardware to complete the process. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include any entity or device capable of carrying computer program code, a recording medium, a USB flash drive, a portable hard drive, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium, etc.

[0092] The computer-readable storage medium can be an internal storage unit of the electronic device in any of the foregoing embodiments, such as a hard disk or memory of the electronic device. The computer-readable storage medium can also be an external storage device of the electronic device, such as a plug-in hard disk, smart media card (SMC), secure digital card (SD), flash card, etc., equipped on the electronic device. Furthermore, the computer-readable storage medium can include both internal and external storage units of the electronic device. The computer-readable storage medium is used to store computer programs and other programs and data required by the electronic device. The computer-readable storage medium can also be used to temporarily store data that has been output or will be output.

[0093] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this application.

[0094] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the electronic devices and units described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0095] In the several embodiments provided in this application, it should be understood that the disclosed electronic devices and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces or units, or it may be an electrical, mechanical, or other form of connection.

[0096] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of the embodiments of this application, depending on actual needs.

[0097] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0098] The above are merely specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and these modifications or substitutions should all be covered within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for regulating the pressure of the air chamber in a high-precision constant-flow bubble level gauge, characterized in that, include: Acquire multi-source data, including environmental data of the bubble level gauge and state parameters of the bubble level gauge; The multi-source data is subjected to feature extraction and fusion to obtain a multi-source feature vector; The multi-source feature vectors are input into a pre-trained barometric pressure fault diagnosis model to obtain the barometric pressure fault category; Based on the gas pressure fault category, a corresponding pressure stabilization and regulation strategy is matched with the preset regulation strategy library; Based on the pressure stabilization and regulation strategy, the target actuator is controlled to adjust the air chamber pressure of the bubble level gauge to obtain the target air pressure; During the preset observation period, the stability of the target air pressure is verified, and the verification results are obtained. If the verification result is successful, the current adjustment strategy is maintained and steady-state monitoring is performed; If the verification result is that the verification fails, then an optimization iteration mechanism is executed.

2. The method for regulating the air chamber pressure of a high-precision constant-flow bubble level gauge according to claim 1, characterized in that, The step of inputting the multi-source feature vectors into a pre-trained barometric pressure fault diagnosis model to obtain barometric pressure fault categories includes: The multi-source feature vector is input into the input layer of the barometric pressure fault diagnosis model; By performing nonlinear transformation and feature abstraction on the multi-source feature vector through the hidden layer of the air pressure fault diagnosis model, deep-level fault features are obtained. The output layer of the barometric pressure fault diagnosis model outputs the probability distribution of the barometric pressure fault category based on the deep fault characteristics. Based on the probability distribution, the type of air pressure fault is determined.

3. The method for regulating the air chamber pressure of a high-precision constant-flow bubble level gauge according to claim 2, characterized in that, The determination of the barometric pressure fault category based on the probability distribution includes: If the probability value of a single fault category in the probability distribution is greater than or equal to a preset judgment threshold, then the single fault category is taken as the air pressure fault category. If the highest probability value in the probability distribution is less than the judgment threshold, and there are two or more fault categories with probability values ​​greater than or equal to the preset auxiliary threshold, and this probability distribution feature continues to exist in subsequent consecutive detection cycles, then it is determined to be a composite fault category. If the probability values ​​of all fault categories in the probability distribution are less than the auxiliary threshold, then the fault is determined to be an unknown fault.

4. The method for regulating the pressure of the air chamber of a high-precision constant-flow bubble level gauge according to claim 3, characterized in that, If the probability values ​​of all fault categories are less than the auxiliary threshold, then after determining it as an unknown fault, the following steps are also included: Based on the air pressure fluctuation range of the historical normal data of the bubble level gauge, the target safe range of the air chamber pressure of the bubble level gauge is set. The actuator is controlled to adjust the air pressure in the air chamber of the bubble level gauge to obtain the adjusted air pressure, and it is determined whether it is within the target safe range. If the regulating air pressure is within the target safe range, the multi-source data corresponding to the regulating air pressure will be marked as normal data and will not be added to the sample set. If the regulated air pressure is not within the target safe range, or if new unknown faults are continuously generated, the multi-source data and regulated effects recorded during the regulated process will be marked as an unknown fault sample set and temporarily stored. When the amount of data in the unknown fault sample set reaches a set size or a fixed synchronization period, the unknown fault sample set is used to incrementally learn the air pressure fault diagnosis model.

5. The method for regulating the air chamber pressure of a high-precision constant-flow bubble level gauge according to claim 1, characterized in that, The step of matching the corresponding pressure stabilization strategy with a preset adjustment strategy library based on the pressure fault category also includes: Based on the deviation between the multi-source feature vector and the normal operating condition reference vector, and the rate of change of the state parameters, a weighted fusion calculation is performed to determine the comprehensive severity level of the air pressure fault category. Based on the type of the air pressure fault and the overall severity level, a corresponding pressure stabilization strategy is matched from the preset regulation strategy library.

6. The method for regulating the air chamber pressure of a high-precision constant-flow bubble level gauge according to claim 5, characterized in that, The determination of the comprehensive severity level of the barometric pressure fault category by performing a weighted fusion calculation based on the deviation between the multi-source feature vector and the normal operating condition baseline vector, and the rate of change of the state parameters, includes: The deviation and the rate of change are normalized to obtain the standardized deviation and the standardized rate of change. The standardized deviation and the standardized rate of change are weighted and fused based on a preset first weighting coefficient and a second weighting coefficient to obtain a comprehensive severity score; wherein the sum of the first weighting coefficient and the second weighting coefficient is one. The overall severity score is mapped to a preset severity score range to obtain the overall severity level.

7. The method for regulating the air chamber pressure of a high-precision constant-flow bubble level gauge according to claim 6, characterized in that, Also includes: Based on the fuzzy logic control strategy, the standardized deviation and standardized rate of change are used as input variables; Reasoning is performed based on a preset fuzzy rule base to obtain the adjustment strategies for the first weight coefficient and the second weight coefficient. The fuzzy rule base includes weight coefficient adjustment strategies corresponding to different combinations of input variables. By performing defuzzy calculations, the adjustment strategy is transformed to obtain adjustment instructions; Based on the adjustment instruction, the corresponding weight coefficients are adjusted using a preset first step length or second step length.

8. The method for regulating the pressure of the air chamber of a high-precision constant-flow bubble level gauge according to claim 5, characterized in that, The step of matching a corresponding pressure stabilization strategy from the preset adjustment strategy library based on the type of the pressure fault and the overall severity level includes: Based on the described air pressure fault category, determine the target actuator associated with it; The degree of parameter adjustment of the target actuator is determined based on the comprehensive severity level of the gas pressure failure category. Based on the environmental data, the adjustment parameters are corrected.

9. The method for regulating the air chamber pressure of a high-precision constant-flow bubble level gauge according to claim 8, characterized in that, The step of correcting the adjustment parameters based on the environmental data includes: If the ambient temperature is less than the preset first temperature threshold, the single action time of the pulse purging solenoid valve is increased in a positive correlation with the negative deviation between the ambient temperature and the first temperature, and its action frequency is reduced in a negative correlation. If the ambient temperature is greater than the preset second temperature threshold, the adjustment range of the constant flow pump power is reduced in a negative correlation manner based on the positive deviation between the ambient temperature and the second temperature threshold. If the ambient humidity is greater than the preset humidity threshold, then after the voltage regulation strategy is executed, a preventive pulse purging task is performed based on the idle time period of the bubble level gauge.

10. A chamber pressure stabilization and regulation system for a high-precision constant-flow bubble level gauge, characterized in that, include: The data acquisition module is used to acquire multi-source data, including environmental data of the bubble level gauge and state parameters of the bubble level gauge. The feature fusion module is used to extract and fuse features from the multi-source data to obtain a multi-source feature vector. The fault diagnosis module is used to input the multi-source feature vector into the pre-trained barometric pressure fault diagnosis model to obtain the barometric pressure fault category. The strategy matching module is used to match the corresponding pressure stabilization and regulation strategy with the preset regulation strategy library based on the pressure fault category. The pressure regulation module is used to control the target actuator based on the pressure regulation strategy to regulate the air chamber pressure of the bubble level gauge to obtain the target air pressure; The pressure verification module is used to verify the stability of the target air pressure within a preset observation period and obtain the verification result. If the verification result is that the verification is successful, the current adjustment strategy is maintained and steady-state monitoring is performed. If the verification result is that the verification is unsuccessful, an optimization iteration mechanism is executed.

Images