Method and system for condition diagnosis of a pulse valve of a dust collector based on pressure data

Abstract

The present application relates to the technical field of industrial equipment state monitoring, and particularly relates to a dust collector pulse valve state diagnosis method and system based on pressure data. The method comprises: collecting pressure data on a compressed air pipeline where the pulse valve is located and blowing driving signals of each pulse valve in real time; taking the blowing driving signals as a benchmark, obtaining pressure waveform segments of each pulse valve in a blowing event from the pressure data, and calculating actual pressure differences of each pulse valve in the blowing event; establishing a prediction model based on gas dynamics theory and a regression model architecture, the prediction model being used to predict ideal pressure differences of each pulse valve in the blowing event; comparing the actual pressure differences of each pulse valve with the corresponding ideal pressure differences, and determining state diagnosis results of each pulse valve based on the comparison results. By establishing a complete, quantifiable monitoring chain from a control command to a pneumatic execution effect, the performance state of the pulse valve can be effectively, accurately and early diagnosed.

Description

Technical Field

[0001] This invention relates to the field of industrial equipment condition monitoring technology, specifically to a condition diagnosis method and system for dust collector pulse valves based on pressure data. Background Technology

[0002] Dust collectors (baghouse dust collectors) typically consist of multiple independent compartments connected in parallel. Each compartment contains multiple sets of filter bags. The dust collector's cleaning system is connected to an external compressed air system via compressed air pipelines to obtain a high-pressure air source. The pulse valve is the core actuator of the dust collector's cleaning system. When the filter bag surface is excessively dusty or the filter bag resistance is too high, requiring a cleaning operation, the pulse valve controls the nozzle of the blowpipe to instantly open, releasing high-pressure compressed air to reverse-blow the filter bag and remove surface dust. Therefore, the performance of the pulse valve directly affects the cleaning effect. When the pulse valve malfunctions, such as weak blowing, air leakage, or failure to operate, it will lead to increased dust accumulation on the filter bags, increased system resistance, increased energy consumption, and even emissions exceeding standards due to dust penetration.

[0003] Currently, because multiple pulse valves in a dust collector share the same compressed air pipeline, the action of any one pulse valve will cause pressure fluctuations throughout the pipeline, creating complex mutual interference. Therefore, existing technologies for monitoring the status of pulse valves mainly rely on the following two methods:

[0004] 1. Switching signal feedback: However, the switching signal can only monitor whether the solenoid coil of the pulse valve is energized (whether the control circuit is conductive). That is, it is a binary, electrical signal indicating "action / non-action." But the actual mechanical action (valve core / diaphragm movement) and pneumatic performance (intensity of the jet flow) of the valve occur at the physical execution level. Therefore, it is impossible to determine whether the valve core of the pulse valve is opening normally, whether the diaphragm is intact, or whether the jet flow meets the requirements. Faults such as valve core jamming and diaphragm damage occur precisely in the conversion link from electrical signal to mechanical action, causing a "false normal" blind spot in the monitoring system based on switching signal feedback.

[0005] 2. Manual periodic inspection and experience judgment: Maintenance personnel make judgments subjectively by listening to the blowing sound and touching the airway vibration. However, the judgment standards are not uniform and there is a lack of objective quantitative basis. Furthermore, for faults with gradual performance degradation (such as diaphragm fatigue and partial airway blockage), real-time monitoring and early warning cannot be achieved. The faults are often only discovered after the dust removal fails and the problem worsens, resulting in a delayed response.

[0006] It can be seen that existing technologies still struggle to effectively, accurately, and early diagnose the performance status of the pulse valve, which is the most critical factor in determining the dust removal effect. Summary of the Invention

[0007] To address the aforementioned technical problems, this invention proposes a state diagnosis method and system for dust collector pulse valves based on pressure data. The aim is to establish a quantifiable and complete monitoring chain from "control commands" to "pneumatic execution effects," enabling effective, accurate, and early diagnosis of the pulse valve performance status, which is the most critical factor in determining the dust removal effect.

[0008] In a first aspect, this application provides a condition diagnosis method for a dust collector pulse valve based on pressure data, comprising the following steps:

[0009] Real-time acquisition of pressure data on the compressed air pipeline where the pulse valve is located, and acquisition of the jet drive signal of each pulse valve;

[0010] Based on the jet drive signal, pressure waveform segments of each pulse valve during the jet event are extracted from the pressure data;

[0011] Based on the pressure waveform segments of each pulse valve during the jetting event, the actual pressure difference of each pulse valve during the jetting event is calculated.

[0012] A prediction model is established based on gas dynamics theory and regression model architecture. The prediction model is used to predict the ideal pressure difference of each pulse valve in the jetting event.

[0013] The actual pressure difference of each pulse valve is compared with the corresponding ideal pressure difference, and the condition diagnosis result of each pulse valve is determined based on the comparison result.

[0014] In some embodiments, the actual pressure difference between each pulse valve during a jetting event is calculated based on the pressure waveform segments of each pulse valve during the jetting event, including:

[0015] Based on the pressure waveform segment of the pulse valve during the injection event, the steady-state pressure of the pipeline before the injection event and the valley value of the pipeline pressure during the injection event are extracted.

[0016] The difference between the steady-state pressure of the pipeline before the purging event and the trough of the pipeline pressure during the purging event is calculated to obtain the actual pressure difference of the pulse valve during the purging event.

[0017] In some embodiments, a prediction model is established based on gas dynamics theory and a regression model architecture. This prediction model is used to predict the ideal pressure difference between each pulse valve during a jetting event, including:

[0018] A physical model is constructed based on gas dynamics theory. The physical model is used to calculate the first ideal pressure difference of the pulse valve according to the input characteristics, so as to simplify the blowing process into the adiabatic expansion process of the constant volume gas tank.

[0019] A data-driven model is built based on a regression model architecture. The model is trained using historical pulse-jet event data under healthy conditions to learn the nonlinear relationship between the second ideal pressure difference of the pulse valve and the input characteristics.

[0020] Real-time collection of blowdown event data with a status diagnosis result of "healthy" is used as a new sample set of health data.

[0021] Based on the amount of data in the new health data sample set, the physical model and the trained data-driven model are weighted and fused to form a prediction model. The prediction model is used to predict the final ideal pressure difference of the pulse valve in the blow-off event based on the input features.

[0022] In some embodiments, a physical model is constructed based on gas dynamics theory. This physical model is used to calculate the first ideal pressure difference of the pulse valve according to the input characteristics, so as to simplify the injection process into an adiabatic expansion process of a constant-volume gas tank, including:

[0023] A physical calculation formula for the first ideal pressure difference of a pulse valve is constructed based on the law of mass conservation and the gas law.

[0024] Extract the dominant factor features and key geometric features from the physical calculation formulas;

[0025] The remaining parameters in the physical calculation formula are combined into system characteristic constants, and the system characteristic constants are calibrated.

[0026] The physical calculation formula is transformed into a simplified calculation formula based on system characteristic constants, dominant factor characteristics, and key geometric features, which serves as a physical model. The input of the physical model is the dominant factor characteristics and key geometric features, and the output of the physical model is the first ideal pressure difference of the pulse valve.

[0027] In some embodiments, calibrating the system characteristic constants includes:

[0028] Acquire historical pulse valve pulse event data under healthy conditions, including real-time steady-state pressure of the pipeline before the pulse event, pulse valve pulse time, equivalent pipeline distance between pressure sensor and pulse valve, and actual pressure difference of pulse valve during the pulse event.

[0029] The optimal calibration values ​​of the system characteristic constants were obtained by fitting the historical purging event data under healthy conditions using the least squares method.

[0030] In some embodiments, the dominant factor features include the real-time steady-state pressure of the pipeline before the injection event and the injection time of the pulse valve, and the key geometric features include the equivalent pipeline distance between the pressure sensor and the pulse valve. Then, the expression for the physical model is:

[0031]

[0032] Where ΔP_ideal_1 represents the first ideal pressure difference of the pulse valve, K represents the system characteristic constant, P_real_time represents the real-time steady-state pressure of the pipeline before the injection event, T represents the injection time of the pulse valve, and L represents the equivalent distance of the pipeline between the pressure sensor and the pulse valve.

[0033] In some embodiments, a data-driven model is trained using historical pulse-jet event data under healthy conditions to learn the nonlinear relationship between the second ideal pressure difference of the pulse valve and the input characteristics, including:

[0034] Acquire historical pulse valve pulse event data under healthy conditions, including real-time steady-state pressure of the pipeline before the pulse event, pulse valve pulse time, equivalent pipeline distance between pressure sensor and pulse valve, and actual pressure difference of pulse valve during the pulse event.

[0035] The actual pressure difference of the pulse valve during the injection event under healthy conditions is used as the output of the data-driven model. The real-time steady-state pressure of the pipeline before the injection event, the injection time of the pulse valve, and the equivalent distance between the pressure sensor and the pulse valve are used as the input features of the data-driven model. A regression algorithm is used to train the data-driven model to obtain the trained data-driven model.

[0036] In some embodiments, based on the amount of data in a new health data sample set, a weighted fusion of the physical model and the trained data-driven model is performed as a prediction model, including:

[0037] If the cumulative amount of data in the new health data sample set is less than or equal to the preset quantity threshold, the weight of the physical model is assigned a value of 1, and the weight of the data-driven model is assigned a value of 0.

[0038] If the cumulative amount of data in the new health data sample set exceeds a preset threshold, the weight assignment of the physical model is negatively correlated with the current cumulative amount of data, while the weight assignment of the data-driven model is positively correlated with the current cumulative amount of data.

[0039] In some embodiments, the actual pressure difference of each pulse valve is compared with the corresponding ideal pressure difference, and the condition diagnosis result of each pulse valve is determined based on the comparison result, including:

[0040] If the actual pressure difference is less than the ideal pressure difference × (1-ε), then the condition diagnosis result is that the pulse valve injection intensity is insufficient.

[0041] If the actual pressure difference is greater than the ideal pressure difference × (1 + ε), the condition diagnosis result is that the pulse valve is over-sprayed or has serious leakage.

[0042] If the actual pressure difference is less than the preset lower threshold, and the jet drive signal is confirmed to be issued, the status diagnosis result is that the pulse valve does not operate.

[0043] If ideal pressure difference × (1-ε) ≤ actual pressure difference ≤ ideal pressure difference × (1+ε), then the state diagnosis result is a healthy state, where ε represents the preset error coefficient.

[0044] Secondly, this application provides a condition diagnostic system for dust collector pulse valves based on pressure data, including:

[0045] The data acquisition module is used to collect pressure data on the compressed air pipeline where the pulse valve is located in real time, and to acquire the injection drive signal of each pulse valve.

[0046] The pressure waveform capture module is used to capture pressure waveform segments of each pulse valve during the blowing event from the pressure data, based on the blowing drive signal.

[0047] The actual pressure difference calculation module is used to calculate the actual pressure difference of each pulse valve in the injection event based on the pressure waveform segment of each pulse valve in the injection event;

[0048] The theoretical pressure difference prediction module is used to establish a prediction model based on gas dynamics theory and regression model architecture. The prediction model is used to predict the ideal pressure difference of each pulse valve in the jetting event.

[0049] The status diagnosis module is used to compare the actual pressure difference of each pulse valve with the corresponding ideal pressure difference, and determine the status diagnosis result of each pulse valve based on the comparison result.

[0050] The beneficial technical effects of the present invention include at least the following:

[0051] 1. A state diagnosis method and system for dust collector pulse valves based on pressure data is adopted. By creatively combining high-frequency pressure data, event-driven data slicing, quantitative feature extraction, dynamic prediction models, and comparative diagnosis, a complete pulse valve state monitoring system with interlocking links and adaptive feedback is formed, thus solving the problem of effective, accurate, and early diagnosis of pulse valve performance status in existing technologies. Specifically: First, the dual-source real-time acquisition of pressure data from the compressed air pipeline where the pulse valve is located and the jet drive signal provides the original data foundation for the system. However, this alone cannot avoid pressure fluctuation interference in multi-valve parallel systems. Therefore, pressure waveform segments based on the jet drive signal are extracted, and the continuous pressure data is aligned using the precise timestamp of the jet drive signal. The pressure dynamics of a single jet event are isolated from the mixed signals, fundamentally solving the data attribution problem caused by cross-interference of multiple valve actions. This allows subsequent analysis to be precisely targeted at specific valves, achieving a leap from "mixed monitoring" to "event-specific monitoring". Subsequently, the actual pressure difference of the pulse valve during a pulse-jet event is calculated based on isolated pressure waveform segments, transforming the pulse-jet intensity from a subjective perception into an objective quantitative indicator. However, using only the actual pressure difference is susceptible to false alarms due to fluctuations in operating conditions. Therefore, a predictive model based on gas dynamics theory and regression models is further established. This model dynamically predicts the ideal pressure difference of the pulse valve under healthy conditions based on real-time gas source pressure, pipeline characteristics, and other operating parameters, thus providing an adaptive benchmark for the actual pressure difference. This synergy allows the diagnosis to adapt to dynamic factors such as changes in gas source pressure during dust collector operation, significantly improving the robustness and accuracy of the diagnosis. Finally, by comparing the actual pressure difference with the ideal pressure difference, the pulse valve's condition diagnosis results are generated, thereby achieving a shift from post-fault repair to preventative maintenance. This enables effective, accurate, and early diagnosis of the pulse valve's performance status, which is crucial for determining the dust removal effect.

[0052] 2. Compared to existing technologies that can only determine whether a pulse valve "operates", the prediction model in this application provides gas dynamics theory support through the design of a physical model. It simplifies the blowing process into an instantaneous venting process of a constant-volume gas tank. It assumes that during the blowing process, the pulse valve opens completely instantaneously, and compressed air flows through the valve nozzle at the speed of sound or subsonic speed. During the blowing time, the gas in the tank undergoes adiabatic expansion, making its physical principle clear and easy to explain. This ensures that the prediction of the ideal pressure difference of the pulse valve conforms to physical laws and avoids the risk of a "black box". At the same time, a data-driven model is designed to learn the nonlinear relationship of the actual system through historical blowing event data under healthy conditions, correct the simplification error of the physical model, thereby achieving accurate quantification of the blowing intensity and making the diagnostic results objective and repeatable, reducing operation and maintenance costs (such as eliminating the need for frequent manual inspection of multiple pulse valves). Meanwhile, by inputting the real-time steady-state pressure of the pipeline before the purging event, the fluctuation of the air source pressure is captured in real time, so that the output of the prediction model is physically bound to the purging intensity, and the ideal pressure difference of the pulse valve is dynamically adjusted with the working conditions. Thus, in the subsequent diagnosis process, performance degradation (such as diaphragm fatigue causing the actual pressure difference to slowly deviate from the ideal pressure difference) can be captured early, truly realizing predictive maintenance.

[0053] 3. By comparing the actual pressure difference of the pulse valves with the ideal pressure difference predicted based on the real-time operating conditions of the system, common interferences such as pipeline pressure fluctuations and sensor position differences are effectively eliminated, achieving highly targeted diagnostic conclusions and significantly reducing the misdiagnosis rate. Simultaneously, in the process of determining the status diagnosis results of each pulse valve based on the comparison results, relative error analysis enables refined classification of fault modes, helping maintenance personnel to accurately locate the cause of the fault (such as distinguishing between "diaphragm wear" and "insufficient air supply"), thus improving maintenance efficiency.

[0054] Other features and advantages of the present invention will be disclosed in detail in the following detailed description and accompanying drawings. Attached Figure Description

[0055] The invention will be further described below with reference to the accompanying drawings:

[0056] Figure 1 This is a schematic diagram illustrating the application scenario of the dust collector pulse valve according to an embodiment of the present invention.

[0057] Figure 2 This is a flowchart of a dust collector pulse valve status diagnosis method based on pressure data, according to an embodiment of the present invention.

[0058] Figure 3 This is a timing diagram of the pressure data and the blowing drive signal when the pulse valve is blowing in a healthy state according to an embodiment of the present invention.

[0059] Figure 4 This is a timing diagram of the pressure data and the blowing drive signal when the pulse valve is blowing under abnormal conditions according to an embodiment of the present invention.

[0060] Figure 5 This is a schematic diagram of the status diagnosis system for a dust collector pulse valve based on pressure data, according to an embodiment of the present invention. Detailed Implementation

[0061] The technical solutions of the embodiments of the present invention will be explained and described below with reference to the accompanying drawings. However, the following embodiments are only preferred embodiments of the present invention and not all of them. Other embodiments obtained by those skilled in the art based on the embodiments in the implementation methods without creative effort are all within the protection scope of the present invention.

[0062] In the following description, terms such as “inner,” “outer,” “upper,” “lower,” “left,” and “right” are used only to indicate orientation or positional relationship for the convenience of describing the embodiments and simplifying the description, and are not intended to indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the present invention.

[0063] Please see the appendix Figure 1 , Figure 1 A schematic diagram illustrating an application scenario for a dust collector pulse valve is shown. For example... Figure 1 As shown, taking a partial structure of a cement plant kiln tail dust collector as an example, the dust collector consists of 8 independent chambers, serving 39 pulse valves. Each chamber contains multiple sets of filter bags. All chambers are connected to an external air compressor station via a shared compressed air pipeline to obtain a high-pressure air source. The compressed air pipeline is divided into two lines serving two rows of chambers, with four chambers in each row. A high-frequency pressure sensor is installed at each branch node of the pipeline, for a total of two pressure sensors. The diagnostic target is the 39 pulse valves of the dust collector.

[0064] When the system performs the dust removal operation, it follows a strict and orderly timing control logic: First, the dust collector controller (PLC) sends a closing command to the lifting valve of the target compartment according to a preset sequence (such as offline dust removal mode), isolating the compartment from the main flue gas channel and putting it into an independent dust removal state; then, the PLC sends a short-term jet drive signal (usually 100-200 milliseconds) to the specific pulse valve of the compartment. After receiving the jet drive signal, the specific pulse valve opens instantly, releasing the high-pressure compressed air in the main compressed air pipeline through the outlet of the valve, and directly injecting it into the filter bag of the corresponding group through the jet pipe installed inside the compartment; this high-speed airflow forms a strong reverse pressure inside the filter bag, causing the filter bag to expand and shake rapidly, thereby shaking the dust layer attached to its outer surface into the ash hopper. During the dust removal process, the compressed air pipeline acts as the "energy transmission artery," the pulse valve acts as the "switch" for precise control, and the chamber acts as the "closed execution space" where the dust removal action takes place. The three work closely together through the timing control of the dust collector controller to achieve efficient dust removal in separate chambers, online or offline, and periodically, ensuring that the filtration performance of the filter bags is continuously restored without stopping the machine.

[0065] Example 1:

[0066] Please see the appendix Figure 2 , Figure 2 A schematic flowchart of a dust collector pulse valve status diagnosis method based on pressure data, provided in one embodiment of this specification, is shown.

[0067] like Figure 2 As shown, the condition diagnosis method for dust collector pulse valves based on pressure data may include at least the following steps:

[0068] S1 collects pressure data on the compressed air pipeline where the pulse valve is located in real time, and obtains the jet drive signal of each pulse valve.

[0069] In this embodiment, a high-frequency pressure sensor (response frequency ≥ 1kHz, sampling frequency ≥ 500Hz) is installed in the main compressed air pipeline to achieve real-time acquisition of dynamic pressure data on the compressed air pipeline branch where the pulse valve is located. Simultaneously, this embodiment reads the pulse valve's blowing drive signal (i.e., the switching signal) from the dust collector controller.

[0070] S2, based on the jet drive signal, extracts the pressure waveform segments of each pulse valve during the jet event from the pressure data.

[0071] Understandably, the fundamental problem preventing existing technologies (such as regular manual inspections, experience-based judgment, or switch signal feedback) from monitoring the status of a single pulse valve lies in the complexity and non-specificity of the data. Specifically, in a dust collector cleaning system, multiple pulse valves share the same compressed air main pipeline. Therefore, the action of any pulse valve will generate pressure fluctuations in the pipeline, and these fluctuations will superimpose and interfere with each other. Traditional methods collect a "mixed pressure signal" resulting from the combined action of all valves over a period of time, failing to distinguish the "independent contribution" of any single valve.

[0072] To this end, this embodiment establishes a clear and accurate "identity file" for each pulse-jet event through dual-source data acquisition and pressure waveform segment extraction, thereby achieving precise correlation and isolated feature extraction. Specifically, two types of data are acquired simultaneously: one is the pulse-jet drive signal, which is a "command signal" from the control system, recording "which pulse valve" was commanded to act at "what time point," essentially a precise timestamp and event trigger. The other is the pressure data on the compressed air pipeline branch where the pulse valve is located, which is an "effect signal" generated in the physical world after the command is executed, recording the actual pressure change in the pipeline and reflecting the intensity of the actual pulse-jet action. Next, based on the dual-source data, "event alignment" is performed using the drive signal as a reference. That is, the system uses the rising / falling edge of the pulse-jet drive signal issued by the dust collector controller as the time reference, and extracts pressure waveform segments with fixed time windows from the continuous pressure data stream. Because this time window is very short (e.g., 300ms) and opens immediately following the injection drive signal of a specific valve, the pressure changes captured within this time window greatly reduce the probability of interference from other valve actions. They can be attributed to the injection of that valve, thus avoiding cross-interference from multiple valve actions and solving the problem of "attributing the pressure change to which valve," achieving precise correlation of a single injection event. Furthermore, this captured pressure waveform segment contains only the complete pressure dynamic characteristics (pressure drop, trough, and recovery phase) of that valve's injection event, effectively "isolving" the pressure data from the continuous data stream into an independent analysis sample.

[0073] For example, assume the pulse valves A and B have a 500ms interval between their injection drive signals. When the system processes the command for pulse valve A, it creates a 300ms wide time window centered on the actuation time of pulse valve A. This window occurs entirely before pulse valve B actuates, therefore the pressure data within the window purely reflects the action of pulse valve A and is completely unaffected by pulse valve B. The system creates such an independent "analysis sandbox" for each pulse valve's injection.

[0074] It is understandable that this embodiment uses a dual-source acquisition method of high-frequency pressure sensor and jet drive signal, combined with event-driven data slicing, to transform continuous pressure data stream into independent jet event samples, achieving accurate correlation of single jet events. This fundamentally solves the problem of monitoring intermittent and concurrent events in multi-pulse valve systems, laying a data foundation for subsequent accurate quantitative feature extraction and intelligent diagnosis. This is the core premise for this embodiment to achieve a leap from "qualitative" to "quantitative" and from "general" to "precise" diagnosis.

[0075] S3, based on the pressure waveform segments of each pulse valve during the jetting event, calculate the actual pressure difference of each pulse valve during the jetting event.

[0076] Specifically, in this embodiment, based on the pressure waveform segments of each pulse valve during the jetting event, the actual pressure difference of each pulse valve during the jetting event is calculated, including:

[0077] S31, Based on the pressure waveform segment of the pulse valve during the injection event, extract the steady-state pressure of the pipeline before the injection event and the valley value of the pipeline pressure during the injection event;

[0078] S32, calculate the difference between the steady-state pressure of the pipeline before the purging event and the valley value of the pipeline pressure during the purging event, and obtain the actual pressure difference of the pulse valve during the purging event.

[0079] It is understandable that the actual pressure difference of the pulse valve during a purging event is directly proportional to the amount of gas consumed during purging and is measurable. Therefore, this embodiment uses it as the core indicator of purging intensity to achieve a transformation from qualitative to quantitative analysis, providing objective numerical indicators for pulse valve status monitoring and supporting subsequent trend analysis and accurate comparison.

[0080] S4. A prediction model is established based on gas dynamics theory and regression model architecture. The prediction model is used to predict the ideal pressure difference of each pulse valve in the jetting event.

[0081] It is understandable that existing static threshold methods (such as fixed pressure difference thresholds) cannot adapt to changes in dust collector operating conditions (such as fluctuations in gas source pressure and differences in pipeline layout). Therefore, this embodiment creatively introduces a dynamic prediction model based on gas dynamics theory and a regression model architecture. This model predicts the ideal pressure difference of the pulse valve in a healthy state based on real-time operating conditions, providing an adaptive benchmark for quantitative diagnosis. This effectively improves the adaptability of subsequent state diagnosis and significantly reduces the misdiagnosis rate.

[0082] Specifically, in this embodiment, a prediction model is established based on gas dynamics theory and a regression model architecture. The prediction model is used to predict the ideal pressure difference of each pulse valve during the injection event, including:

[0083] S41, a physical model is constructed based on gas dynamics theory. The physical model is used to calculate the first ideal pressure difference of the pulse valve according to the input characteristics, so as to simplify the blowing process into the adiabatic expansion process of the constant volume gas tank.

[0084] Understandably, the physical model in this embodiment simplifies the blowing process as an instantaneous venting process of a constant-volume gas tank. It assumes that during the blowing process, the pulse valve opens completely instantaneously, and compressed air flows through the valve nozzle at sound or subsonic speeds. During the blowing time, the gas inside the tank undergoes adiabatic expansion, making its physical principle clear and easy to explain. Based on gas dynamics principles, the physical model can provide a reasonable prediction value (i.e., the first ideal pressure difference of the pulse valve) that conforms to physical laws, even without historical data, provided only the input features are available, ensuring the initial reliability of the prediction model in the absence of training data.

[0085] Furthermore, in this embodiment, a physical model is constructed based on gas dynamics theory. This physical model is used to calculate the first ideal pressure difference of the pulse valve according to the input characteristics, so as to simplify the injection process into an adiabatic expansion process of a constant-volume gas tank, including:

[0086] S411, the physical calculation formula for the first ideal pressure difference of the pulse valve, based on mass conservation and the gas law, can be expressed as:

[0087]

[0088] Wherein, ΔP_ideal_1 represents the first ideal pressure difference of the pulse valve, and P_real_time represents the real-time steady-state pressure of the pipeline before the injection event. It can be understood that since there is more than one pulse valve device connected to the compressed air pipeline branch, and multiple devices consume the gas in the compressed air pipeline together, P_real_time in this embodiment needs to capture dynamic fluctuations to reflect the current gas source status, and is not the same data value as the steady-state pressure of the pipeline before the injection event in the aforementioned S3.

[0089] Where k represents the specific heat capacity ratio of the gas, for example, for air k=1.4, The equivalent nozzle area can be calculated based on the valve flow diameter of the selected pulse valve model; T represents the pulse valve's blowing time; V represents the equivalent volume of the pipeline between the pressure sensor and the pulse valve, equivalent to a "constant volume gas tank" with a volume of V. The equivalent volume V of the pipeline is usually determined by the pipeline diameter, length, and local resistance components such as valves and joints, and can be obtained through drawing calculations or experimental calibration; R represents the gas constant. The gas temperature is represented by M; the molar mass is represented by M, for example, the average molar mass of air is M=0.029; and the flow coefficient is represented by Ψ, which is an empirical coefficient used to correct the deviation between actual flow and ideal flow.

[0090] S412 extracts the dominant factor features and key geometric features from the physical calculation formulas.

[0091] Understandably, directly calculating the physical formula for S411 in real-time at the industrial site would be computationally intensive and prone to inaccurate results due to parameter uncertainties. Therefore, this embodiment, while ensuring engineering accuracy, transforms the complex physical relationships into a practical physical model that is easy to apply and calculate on-site. The specific implementation method is as follows:

[0092] 1. Identifying the dominant factor characteristics: First, in gas dynamics, the pressure difference is usually directly proportional to the initial pressure P_real_time. This is the most core and direct influencing factor; therefore, P_real_time is the dominant factor characteristic. Second, the longer the pulse valve's injection time T, the more gas is ejected, and the greater the pressure difference. Therefore, within a certain pressure range, the pressure difference can be approximately considered to be directly proportional to the injection time T, meaning the injection time T is the dominant factor characteristic.

[0093] 2. Identifying Key Geometric Features: In the original physical calculation formula, the relationship between the equivalent volume V of the pipeline between the pressure sensor and the pulse valve and the pressure difference is inversely proportional and nonlinear; that is, the larger V is, the smaller the pressure difference under the same injection conditions. The equivalent volume V can be calculated by multiplying the pipeline cross-sectional area by the equivalent distance L between the pressure sensor and the pulse valve. For a given system, the pipeline diameter is fixed, therefore the equivalent volume V is directly proportional to the equivalent distance L. Since V is directly proportional to L, the terms related to 1 / V in the original formula can be transformed into terms related to 1 / L. However, factors such as gas flow resistance mean that the relationship between the pressure difference and 1 / L is not a simple linear inverse proportion. Therefore, this embodiment cleverly uses... This is used to fit this nonlinear relationship. Specifically, this stems from the relationship between pressure loss and pipe length in fluid mechanics. In turbulent flow, the pressure loss along the pipe is proportional to the pipe length. However, jetting is a dynamic and instantaneous process involving the propagation and reflection of pressure waves. Therefore, this embodiment uses... This power-law relationship empirically describes its combined effect as a smooth transition between linear inverse proportion (1 / L) and a constant, thus enabling a better fit to the nonlinear behavior of real-world systems. Therefore, L is a key geometric feature.

[0094] S413 combines the remaining parameters in the physical calculation formula into system characteristic constants and calibrates the system characteristic constants.

[0095] In this embodiment, all relatively stable but difficult-to-obtain-precise parameters in the original complex physical calculation formula are packaged and merged into a single system characteristic constant K. Specifically, K ∝ (k, ,R, ,M,Ψ).

[0096] In this embodiment, the calibration of the system characteristic constant can be achieved as follows: During the system design or initial installation phase, when there is no operational data and K cannot be calibrated immediately, the K constant can be set to a theoretical value based on design parameters (such as the effective area of ​​the nozzle and gas properties), or an empirical value based on similar equipment can be used as a temporary initial value. At this time, the prediction accuracy of the physical model may be limited, but it can ensure the operation of basic functions and lay the foundation for subsequent data accumulation and calibration.

[0097] S414 transforms the physical calculation formula into a simplified calculation formula based on system characteristic constants, dominant factor characteristics, and key geometric features. As a physical model, the input of the physical model is the dominant factor characteristics and key geometric features, and the output of the physical model is the first ideal pressure difference of the pulse valve.

[0098] Specifically, in this embodiment, the expression for the physical model is:

[0099]

[0100] Where ΔP_ideal_1 represents the first ideal pressure difference of the pulse valve, K represents the system characteristic constant, P_real_time represents the real-time steady-state pressure of the pipeline before the injection event, T represents the injection time of the pulse valve, and L represents the equivalent distance of the pipeline between the pressure sensor and the pulse valve.

[0101] Furthermore, in this embodiment, after the system has been running for a period of time and the accumulated historical pulse valve pulse event data in a healthy state reaches a certain scale (e.g., thousands to tens of thousands of pulse events), the physical model can be precisely calibrated, that is, the optimal calibration value of K can be obtained:

[0102] S4131, acquire historical pulse valve injection event data in a healthy state, including the real-time steady-state pressure of the pipeline before the injection event, the injection time of the pulse valve, the equivalent distance between the pressure sensor and the pulse valve in the pipeline, and the actual pressure difference of the pulse valve during the injection event.

[0103] Under normal conditions, the actual pressure difference ΔP_actual of the pulse valve during a measured pulse event can be considered the first ideal pressure difference of the pulse valve under that operating condition. Therefore, in this embodiment, ΔP_actual is used as the "true value" or expected value of the first ideal pressure difference ΔP_ideal_1. This step is crucial in connecting the physical world with the model.

[0104] S4132 uses the least squares method to fit the historical jetting event data under healthy conditions to obtain the optimal calibration value of the system characteristic constant.

[0105] For example, the simplified calculation formula of the physical model. This can be viewed as a linear equation: y = K*x. Here, y is the measured ΔP_actual, and x is a composite variable. Using historical jetting event data samples collected across all health states, a unique unknown parameter K is fitted using the least squares method, minimizing the sum of squared errors between the physical model's predicted values ​​and all measured values. This yields a more accurate K constant optimized for the current system, significantly improving the accuracy and reliability of the physical model's predictions.

[0106] It is understandable that the simplification of the physical calculation formula in this embodiment is not arbitrary, but based on a deep understanding of the physical process. The dominant factor characteristics (P_real_time, T) and key geometric features (L) in the complex physical relationships are extracted, and other stable secondary factors, uncertain factors, and complex nonlinear relationships are integrated into a single system characteristic constant K. Finally, K is calibrated using historical pulse-jet event data under healthy conditions, enabling this simplified physical model to highly accurately reflect the true characteristics of the ideal pressure difference of the pulse valve under specific operating conditions.

[0107] S42, based on a regression model architecture, constructs a data-driven model. It uses historical pulse-jet event data under healthy conditions to train the model, learning the nonlinear relationship between the pulse valve's second ideal pressure difference and input characteristics. This allows for data-driven adaptive optimization, capturing complex factors (such as local resistance variations) not covered by the physical model.

[0108] Specifically, in this embodiment, a data-driven model is trained using historical pulse-jet event data under healthy conditions to learn the nonlinear relationship between the second ideal pressure difference of the pulse valve and the input characteristics, including:

[0109] S421, acquire historical pulse valve pulse event data in a healthy state, including real-time steady-state pressure of the pipeline before the pulse event, pulse valve pulse time, equivalent pipeline distance between pressure sensor and pulse valve, and actual pressure difference of pulse valve during the pulse event.

[0110] In the case of a healthy state, the actual pressure difference ΔP_actual of the pulse valve during the actual measured blow event can be considered as the expected value of the second ideal pressure difference ΔP_ideal_2 of the pulse valve under this operating condition.

[0111] S422, the actual pressure difference of the pulse valve during the blow-up event under healthy conditions is used as the output of the data-driven model, and the real-time steady-state pressure of the pipeline before the blow-up event, the blow-up time of the pulse valve, and the equivalent distance between the pressure sensor and the pulse valve are used as the input features of the data-driven model. The data-driven model is trained by a regression algorithm to obtain the trained data-driven model.

[0112] Optionally, the data-driven model can be trained using a regression algorithm via linear regression, where the model is viewed as ΔP_ideal_2 = W1*P_real_time + W2*T + W3*L + b, using historical pulse-jet event data under healthy conditions to determine the optimal weights W and biases b. Alternatively, gradient boosting decision trees (such as XGBoost, LightGBM, etc.) can be used, iteratively constructing multiple decision trees, each learning the residuals of all previous trees, ultimately combining them into a powerful data-driven model. This embodiment does not limit the implementation as long as the data-driven model can learn the nonlinear relationship between the second ideal pressure difference of the pulse valve and the input features.

[0113] S43 collects real-time data on blowdown events with a health status diagnosis result, using it as a new sample set of health data.

[0114] Understandably, the system will continuously generate new diagnostic data during subsequent operation. For those spray event data diagnosed as "healthy," this embodiment automatically adds them to the new health data sample set.

[0115] S44, based on the amount of data in the new health data sample set, performs a weighted fusion of the physical model and the trained data-driven model as a prediction model. The prediction model is used to predict the final ideal pressure difference of the pulse valve in the blow-off event based on the input features.

[0116] Specifically, in this embodiment, based on the amount of data in the new health data sample set, a weighted fusion of the physical model and the trained data-driven model is performed as a prediction model, including:

[0117] S441, if the cumulative amount of data in the new health data sample set is less than or equal to the preset quantity threshold, then the weight of the physical model is assigned a value of 1, and the weight of the data-driven model is assigned a value of 0.

[0118] Understandably, in the early stages of diagnosis, the predictive model in this embodiment prioritizes the use of the physical model, because the physical model does not require historical diagnostic data, and the data-driven model is prone to overfitting or instability when data is insufficient.

[0119] S442, if the cumulative amount of data in the new health data sample set is greater than the preset quantity threshold, then the weight assignment of the physical model is negatively correlated with the current cumulative amount of data, and the weight assignment of the data-driven model is positively correlated with the current cumulative amount of data.

[0120] Understandably, after accumulating diagnostic data, the predictive model in this embodiment employs a weighted fusion strategy because the data-driven model has learned more complex real-world patterns from the continuously collected new samples of health data. This weighted fusion strategy smoothly transitions from "theory-guided" to "data-driven," ensuring the predictive reliability of the ideal pressure difference throughout the system's lifecycle.

[0121] Furthermore, this embodiment can also design a continuous update mechanism for the predictive model. Regularly (e.g., monthly), it uses a new, larger set of health data to continuously update the K-constant of the calibrated physical model and update the training data to drive the model. This allows the predictive model to adapt to slow changes in system equipment (such as slight filter clogging or slight sensor drift), thereby achieving "up-to-date" predictive accuracy. Simultaneously, the continuous update mechanism can also record the long-term deviation trend between the actual pressure difference and the ideal pressure difference, providing a data foundation for remaining life prediction and supporting predictive maintenance plans (e.g., indicating a replacement cycle when the deviation rate continues to increase).

[0122] Understandably, compared to existing technologies that can only determine whether a pulse valve "operates," the predictive model in this embodiment provides gas dynamics theoretical support through a designed physical model. This ensures that the prediction of the ideal pressure difference of the pulse valve conforms to physical laws, avoiding the risk of a "black box." Simultaneously, a data-driven model is designed to learn the nonlinear relationships of the actual system through historical pulse-jet event data under healthy conditions, correcting simplification errors in the physical model. This achieves accurate quantification of pulse-jet intensity and makes the diagnostic results objective and repeatable, reducing maintenance costs (e.g., eliminating the need for frequent manual inspections of multiple pulse valves). Furthermore, by inputting the real-time steady-state pressure of the pipeline before a pulse-jet event, real-time gas source pressure fluctuations are captured, physically binding the predictive model's output to the pulse-jet intensity. The ideal pressure difference of the pulse valve is dynamically adjusted according to operating conditions, enabling early detection of performance degradation (such as diaphragm fatigue causing a slow deviation of the actual pressure difference from the ideal pressure difference) during subsequent diagnostic processes, truly achieving predictive maintenance.

[0123] S5, compare the actual pressure difference of each pulse valve with the corresponding ideal pressure difference, and determine the status diagnosis result of each pulse valve based on the comparison result.

[0124] Specifically, in this embodiment, the actual pressure difference of each pulse valve is compared with the corresponding ideal pressure difference, and the state diagnosis result of each pulse valve is determined based on the comparison result, including:

[0125] If the actual pressure difference is less than the ideal pressure difference × (1-ε), the condition diagnosis result is insufficient pulse valve injection intensity. Possible causes include incomplete diaphragm opening, partial blockage of the air passage, insufficient air supply pressure (locally), or valve body wear.

[0126] If the actual pressure difference > ideal pressure difference × (1 + ε), the condition diagnosis result is pulse valve over-spraying or severe leakage. Possible causes include continuous leakage due to diaphragm incomplete closure, abnormally prolonged spraying time, or leaks in related pipelines.

[0127] If the actual pressure difference is less than the preset lower threshold, and the pulse drive signal is confirmed to be issued, the status diagnosis result is that the pulse valve does not operate. Possible causes include a damaged solenoid coil, a severely jammed valve core, or a complete blockage of the air supply.

[0128] If ideal pressure difference × (1-ε) ≤ actual pressure difference ≤ ideal pressure difference × (1+ε), then the state diagnosis result is a healthy state, where ε represents the preset error coefficient.

[0129] For example, please refer to the appendix. Figure 3 and attached Figure 4 , Figure 3 This is a timing diagram of the pressure data and the pulse drive signal when the pulse valve is in a healthy state according to an embodiment of the present invention. Figure 4 This is a timing diagram of the pressure data and the pulse valve's blowing signal under abnormal conditions, according to an embodiment of the present invention. The vertical axis represents the pressure difference, the curve "—" represents the actual pressure difference of each pulse valve, and "…" represents the ideal pressure difference corresponding to each pulse valve output by the prediction model. The shaded area represents the error range of the ideal pressure difference corresponding to each pulse valve. Based on the dust collector pulse valve state diagnosis method based on pressure data provided in this embodiment, it can be obtained that… Figure 4 The status diagnosis results for pulse valve 2 are: insufficient pulse valve blowing intensity; the status diagnosis results for pulse valve 4 are: pulse valve not operating; and the status diagnosis results for pulse valve 6 are: pulse valve over-spraying or serious leakage.

[0130] For example, maintenance personnel, based on the historical trends of pulse valve diagnostic results, discovered that the actual pressure differential of a certain pulse valve had been slowly decreasing over the past week. During shutdown maintenance, disassembly of this pulse valve revealed slight deformation and elasticity loss in its diaphragm. After replacement, the actual pressure differential of the pulse valve returned to the error range corresponding to the ideal pressure differential, and the condition diagnostic result returned to "healthy state." This successfully achieved early warning of the fault, preventing the filter bag from operating under prolonged high pressure differential due to inadequate dust removal.

[0131] Understandably, this embodiment effectively eliminates common interferences such as pipeline pressure fluctuations and sensor location differences by comparing the actual pressure difference of the pulse valves with the ideal pressure difference predicted based on the real-time operating conditions of the system. This achieves highly targeted diagnostic conclusions and significantly reduces the misdiagnosis rate. Simultaneously, in the process of determining the status diagnosis results of each pulse valve based on the comparison results, relative error analysis enables refined classification of fault modes, which helps guide maintenance personnel to accurately locate the cause of the fault (such as distinguishing between "diaphragm wear" and "insufficient air supply"), thereby improving maintenance efficiency.

[0132] In summary, this embodiment upgrades pulse valve condition diagnosis from the existing "qualitative judgment based on control signals" to "quantitative analysis based on gas dynamics." By directly monitoring the pipeline pressure dynamics during pulse valve injection and combining event alignment to separate single-valve injection events, precise component-level positioning is achieved. A predictive model based on gas dynamics theory and a regression model architecture dynamically predicts the ideal pressure difference of the pulse valves, allowing the diagnostic criteria (ideal pressure difference) for each pulse valve to adaptively adjust with operating conditions, while meeting the interpretability requirements of industrial scenarios. By comparing the actual pressure difference with the ideal pressure difference of each pulse valve over a long period, progressive faults such as diaphragm elastic decay can be detected, thus enabling early warning. The entire condition diagnosis system requires only a small number of pressure sensors to achieve real-time condition monitoring of all pulse valves in the system, at a cost far lower than installing a diagnostic device on each pulse valve.

[0133] Example 2:

[0134] Please see the appendix Figure 5 , Figure 5 This is a schematic diagram of a status diagnosis system for a dust collector pulse valve based on pressure data, provided as an embodiment of this specification.

[0135] like Figure 5 As shown, the condition diagnosis system for the dust collector pulse valve based on pressure data may include at least the following:

[0136] Data acquisition module 1 is used to collect pressure data on the compressed air pipeline where the pulse valve is located in real time, and to acquire the injection drive signal of each pulse valve;

[0137] Pressure waveform capture module 2 is used to capture pressure waveform segments of each pulse valve during the blowing event from pressure data, based on the blowing drive signal.

[0138] The actual pressure difference calculation module 3 is used to calculate the actual pressure difference of each pulse valve in the injection event based on the pressure waveform segment of each pulse valve in the injection event;

[0139] The theoretical pressure difference prediction module 4 is used to establish a prediction model based on gas dynamics theory and regression model architecture. The prediction model is used to predict the ideal pressure difference of each pulse valve in the jetting event.

[0140] The status diagnosis module 5 is used to compare the actual pressure difference of each pulse valve with the corresponding ideal pressure difference, and determine the status diagnosis result of each pulse valve based on the comparison result.

[0141] It is understood that the technical concept of the dust collector pulse valve status diagnosis system based on pressure data provided in this embodiment is similar to the technical concept of the aforementioned dust collector pulse valve status diagnosis method based on pressure data, and will not be repeated here.

[0142] The above description is merely a preferred embodiment disclosed in this application and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of protection involved in this disclosure is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the above-disclosed concept. For example, technical solutions formed by substituting the above features with (but not limited to) technical features with similar functions disclosed in this disclosure.

[0143] Furthermore, while the operations are described in a specific order, this should not be construed as requiring these operations to be performed in the specific order shown or in a sequential order. In certain environments, multitasking and parallel processing may be advantageous. Similarly, while several specific implementation details are included in the above discussion, these should not be construed as limiting the scope of this disclosure. Certain features described in the context of individual embodiments may also be implemented in combination in a single embodiment. Conversely, various features described in the context of a single embodiment may also be implemented individually or in any suitable sub-combination in multiple embodiments.

Claims

1. A method for diagnosing the state of a pulse valve of a dust collector based on pressure data, characterized by, Includes the following steps: Real-time acquisition of pressure data on the compressed air pipeline where the pulse valve is located, and acquisition of the jet drive signal of each pulse valve; Based on the jet drive signal, pressure waveform segments of each pulse valve during the jet event are extracted from the pressure data; Based on the pressure waveform segments of each pulse valve during the jetting event, the actual pressure difference of each pulse valve during the jetting event is calculated. A prediction model is established based on gas dynamics theory and regression model architecture. The prediction model is used to predict the ideal pressure difference of each pulse valve in the jetting event. The actual pressure difference of each pulse valve is compared with the corresponding ideal pressure difference, and the condition diagnosis result of each pulse valve is determined based on the comparison result.

2. The method for diagnosing the condition of a dust collector pulse valve based on pressure data as described in claim 1, characterized in that, Based on the pressure waveform segments of each pulse valve during the jetting event, the actual pressure difference of each pulse valve during the jetting event is calculated, including: Based on the pressure waveform segment of the pulse valve during the injection event, the steady-state pressure of the pipeline before the injection event and the valley value of the pipeline pressure during the injection event are extracted. The difference between the steady-state pressure of the pipeline before the purging event and the trough of the pipeline pressure during the purging event is calculated to obtain the actual pressure difference of the pulse valve during the purging event.

3. The method for diagnosing the condition of a dust collector pulse valve based on pressure data as described in claim 1, characterized in that, A prediction model is established based on gas dynamics theory and a regression model framework. This prediction model is used to predict the ideal pressure difference of each pulse valve during a jetting event, including: A physical model is constructed based on gas dynamics theory. The physical model is used to calculate the first ideal pressure difference of the pulse valve according to the input characteristics, so as to simplify the blowing process into the adiabatic expansion process of the constant volume gas tank. A data-driven model is built based on a regression model architecture. The model is trained using historical pulse-jet event data under healthy conditions to learn the nonlinear relationship between the second ideal pressure difference of the pulse valve and the input characteristics. Real-time collection of blowdown event data with a status diagnosis result of "healthy" is used as a new sample set of health data. Based on the amount of data in the new health data sample set, the physical model and the trained data-driven model are weighted and fused to form a prediction model. The prediction model is used to predict the final ideal pressure difference of the pulse valve in the blow-off event based on the input features.

4. The method for diagnosing the condition of a dust collector pulse valve based on pressure data as described in claim 3, characterized in that, A physical model is constructed based on gas dynamics theory. This model is used to calculate the first ideal pressure difference of the pulse valve according to the input characteristics, thereby simplifying the injection process into an adiabatic expansion process of a constant-volume gas tank, including: A physical calculation formula for the first ideal pressure difference of a pulse valve is constructed based on the law of mass conservation and the gas law. Extract the dominant factor features and key geometric features from the physical calculation formulas; The remaining parameters in the physical calculation formula are combined into system characteristic constants, and the system characteristic constants are calibrated. The physical calculation formula is transformed into a simplified calculation formula based on system characteristic constants, dominant factor characteristics, and key geometric features, which serves as a physical model. The input of the physical model is the dominant factor characteristics and key geometric features, and the output of the physical model is the first ideal pressure difference of the pulse valve.

5. The method for diagnosing the condition of a dust collector pulse valve based on pressure data as described in claim 4, characterized in that, The system characteristic constants are calibrated, including: Acquire historical pulse valve pulse event data under healthy conditions, including real-time steady-state pressure of the pipeline before the pulse event, pulse valve pulse time, equivalent pipeline distance between pressure sensor and pulse valve, and actual pressure difference of pulse valve during the pulse event. The optimal calibration values ​​of the system characteristic constants were obtained by fitting the historical purging event data under healthy conditions using the least squares method.

6. The method for diagnosing the condition of a dust collector pulse valve based on pressure data as described in claim 4, characterized in that, The dominant factors include the real-time steady-state pressure of the pipeline before the injection event and the injection time of the pulse valve. The key geometric features include the equivalent distance in the pipeline between the pressure sensor and the pulse valve. Therefore, the expression for the physical model is: ， Where ΔP_ideal_1 represents the first ideal pressure difference of the pulse valve, K represents the system characteristic constant, P_real_time represents the real-time steady-state pressure of the pipeline before the injection event, T represents the injection time of the pulse valve, and L represents the equivalent distance of the pipeline between the pressure sensor and the pulse valve.

7. The method for diagnosing the condition of a dust collector pulse valve based on pressure data as described in claim 3, characterized in that, A data-driven model is trained using historical pulse-jet event data under healthy conditions to learn the nonlinear relationship between the second ideal pressure difference of the pulse valve and input characteristics, including: Acquire historical pulse valve pulse event data under healthy conditions, including real-time steady-state pressure of the pipeline before the pulse event, pulse valve pulse time, equivalent pipeline distance between pressure sensor and pulse valve, and actual pressure difference of pulse valve during the pulse event. The actual pressure difference of the pulse valve during the pulse-driven event under healthy conditions is used as the output of the data-driven model. The real-time steady-state pressure of the pipeline before the pulse-driven event, the pulse-driven time of the pulse valve, and the equivalent distance between the pressure sensor and the pulse valve are used as the input features of the data-driven model. A regression algorithm is used to train the data-driven model to learn the nonlinear relationship between the second ideal pressure difference of the pulse valve and the input features, thus obtaining the trained data-driven model.

8. The method for diagnosing the condition of a dust collector pulse valve based on pressure data as described in claim 3, characterized in that, Based on the data volume in the new health data sample set, a weighted fusion of the physical model and the trained data-driven model is performed as a prediction model, including: If the cumulative amount of data in the new health data sample set is less than or equal to the preset quantity threshold, the weight of the physical model is assigned a value of 1, and the weight of the data-driven model is assigned a value of 0. If the cumulative amount of data in the new health data sample set exceeds a preset threshold, the weight assignment of the physical model is negatively correlated with the current cumulative amount of data, while the weight assignment of the data-driven model is positively correlated with the current cumulative amount of data.

9. The method for diagnosing the condition of a dust collector pulse valve based on pressure data as described in claim 1, characterized in that, The actual pressure difference of each pulse valve is compared with the corresponding ideal pressure difference. Based on the comparison results, the condition diagnosis results of each pulse valve are determined, including: If the actual pressure difference is less than the ideal pressure difference × (1-ε), then the condition diagnosis result is that the pulse valve injection intensity is insufficient. If the actual pressure difference is greater than the ideal pressure difference × (1 + ε), the condition diagnosis result is that the pulse valve is over-sprayed or has serious leakage. If the actual pressure difference is less than the preset lower threshold, and the jet drive signal is confirmed to be issued, the status diagnosis result is that the pulse valve does not operate. If ideal pressure difference × (1-ε) ≤ actual pressure difference ≤ ideal pressure difference × (1+ε), then the state diagnosis result is a healthy state, where ε represents the preset error coefficient.

10. A condition diagnosis system for dust collector pulse valves based on pressure data, characterized in that, include: The data acquisition module is used to collect pressure data on the compressed air pipeline where the pulse valve is located in real time, and to acquire the injection drive signal of each pulse valve. The pressure waveform capture module is used to capture pressure waveform segments of each pulse valve during the blowing event from the pressure data, based on the blowing drive signal. The actual pressure difference calculation module is used to calculate the actual pressure difference of each pulse valve in the injection event based on the pressure waveform segment of each pulse valve in the injection event; The theoretical pressure difference prediction module is used to establish a prediction model based on gas dynamics theory and regression model architecture. The prediction model is used to predict the ideal pressure difference of each pulse valve in the jetting event. The status diagnosis module is used to compare the actual pressure difference of each pulse valve with the corresponding ideal pressure difference, and determine the status diagnosis result of each pulse valve based on the comparison result.

Images