Self-adaptive diagnosis method and system for tiny gas leakage fault of pressure container

Abstract

The invention provides a pressure vessel gas tiny leakage fault self-adaptive diagnosis method and system, and the method comprises the steps: carrying out the preprocessing of a leakage signal through a self-adaptive signal decomposition algorithm according to a pressure vessel gas tiny leakage acoustic signal collected by an MEMS ultrasonic microphone array, extracting a plurality of leakage features, and inputting the features into a leakage diagnosis model, and the function of identifying micro leakage working conditions of various apertures of the pressure vessel is realized. According to the invention, by extracting the multi-index characteristics of leakage, the micro leakage working conditions of various different apertures are effectively identified, the accuracy of gas micro leakage fault diagnosis of the pressure vessel and the safety monitoring level of an industrial production link are improved, the detection defects caused by parameter selection by artificial experience are overcome, and the detection efficiency is improved. The stability and generalization ability of the small aperture leakage diagnosis model are effectively improved, a basis is provided for subsequent leakage positioning research, and the method has great practical application value.

Description

Technical Field

[0001] This invention belongs to the field of gas leak fault diagnosis technology, specifically relating to an adaptive diagnosis method and system for minor gas leaks in pressure vessels. Background Technology

[0002] Pressure vessels, as key equipment in the production process, primarily store critical energy media such as compressed air, high-pressure steam, and nitrogen. Their operational safety directly impacts industrial production safety and environmental protection. During long-term operation, factors such as corrosion and fatigue stress can cause leaks of varying sizes to form on the pressure vessel walls. This allows leaked flammable gases to accumulate, potentially igniting fires or even explosions upon contact with an open flame, leading to catastrophic consequences for production safety and the environment. Therefore, a fault diagnosis method for early gas leaks in pressure vessels is urgently needed to achieve timely prevention and control of safety risks.

[0003] Acoustic detection technology has become the mainstream method for gas leak detection in the industrial field due to its significant advantages such as fast response and non-contact measurement. Its basic principle is to determine the existence of a leak by capturing the acoustic signals generated by the leak. However, this technology still faces significant challenges in accurately identifying early-stage micro-orifice leaks in pressure vessels, especially those with orifice diameters smaller than 2 mm. This is because the acoustic signals generated by gas leaks in pressure vessels are inherently nonlinear and non-stationary, with complex and variable signal morphology. Furthermore, the acoustic characteristics of leaks with different micro-orifice diameters differ only slightly, making it difficult for traditional methods to accurately identify early-stage micro-orifice leaks. Existing methods mostly focus on determining the existence of leaks or still suffer from problems such as reliance on empirical parameter settings, poor real-time performance, and wide coverage of orifice diameters. Further research is needed to accurately identify early-stage micro-orifice leaks.

[0004] Therefore, it is necessary to propose an adaptive diagnostic method for minor gas leaks in pressure vessels to solve the problem of early gas leaks in internal pressure vessels in production scenarios, and to improve the identification accuracy and robustness of minor leak apertures. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide an adaptive diagnostic method and system for minor gas leakage faults in pressure vessels, which is used to identify minor leakage conditions of pressure vessels with various orifice diameters.

[0006] The technical solution adopted by this invention to solve the above-mentioned technical problems is: an adaptive diagnostic method for minor gas leakage faults in pressure vessels, comprising the following steps: S1: Acquire the raw acoustic signal of gas leakage from the pressure vessel; S2: The Newton-Raphson optimization algorithm is improved by integrating chaotic mapping population initialization, dynamic decision factors and additional exploration components. The key parameters of the variational mode decomposition algorithm, the penalty factor and the number of mode layers, are adaptively selected using the minimum sample entropy as the fitness function. S3: The original acoustic signal is adaptively decomposed using the optimized variational mode decomposition algorithm to obtain multiple intrinsic mode components; S4: Select effective intrinsic mode components as reconstructed signals by using the Pearson correlation coefficient; S5: Extract various parameter indicators from the reconstructed signal to form a leakage feature vector dataset; S6: Construct a high-precision classification and diagnostic model, and input the leakage feature vector dataset to classify, diagnose, and identify gas leakage conditions in pressure vessels.

[0007] According to the above scheme, in step S2, Indicates the first Random numbers generated by the chaotic iteration. For the control parameters of the Bernoulli chaotic mapping, and The expression for population initialization, improved using Bernoulli chaotic mapping, is as follows: .

[0008] According to the above scheme, in step S2, This represents the current iteration number. To maximize the number of iterations, a dynamic decision factor DF is introduced. Its initial value is set to the default initial value used in the Newton-Raphson optimization algorithm and adjusted during iterations. The expression for the dynamic decision factor DF is: .

[0009] According to the above scheme, in step S2, and This represents candidate positions generated based on different search strategies. To introduce dynamic parameters The next update location, and Random numbers in the range [0,1] are used to maintain population diversity; an additional exploration enhancement component is introduced. By updating the location Adding an extra term based on the parameters of the normal distribution random perturbation, the expression is: .

[0010] According to the above scheme, the specific steps in step S4 are as follows: calculate the Pearson correlation coefficient between each decomposed intrinsic mode component and the original acoustic signal, and select the mode components with Pearson correlation coefficients greater than the selection threshold as effective mode component reconstruction signals.

[0011] Furthermore, in step S4, The representative is the first Each intrinsic mode component and the original acoustic signal The correlation coefficient, For the first The average value of each intrinsic modal component. The average value of the original acoustic signal. This represents all intrinsic modal components and the original acoustic signal. correlation coefficient The maximum value in the correlation coefficient The expression is: ; Screening threshold for intrinsic modal components The expression is: .

[0012] According to the above scheme, the specific steps in step S5 are as follows: The reconstructed signal is analyzed in the time and frequency domains to extract multi-index feature parameter combinations to construct a leakage feature vector dataset. The multi-index feature parameter combinations include time-domain features and frequency-domain features. The time-domain features include dimensional and dimensionless feature indicators. The dimensionless time-domain features are used to represent shape features. The frequency-domain features are used to represent the spectral distribution of leakage energy. The leaked feature vector dataset is divided into a training set and a test set according to a certain ratio.

[0013] According to the above scheme, in step S6, the regularization coefficient of the kernel extreme learning machine is adaptively optimized using the Newton-Raphson optimization algorithm. and kernel function Construct a high-precision classification and diagnostic model; set the population size and maximum number of iterations, and limit the parameters. and The optimization range is defined by setting the minimum recognition error rate as the fitness function.

[0014] An adaptive diagnostic system for minor gas leaks in pressure vessels. The data acquisition submodule is used to acquire the raw acoustic signals of gas leakage from the pressure vessel; The parameter selection submodule is used to improve the Newton-Raphson optimization algorithm by integrating chaotic mapping population initialization, dynamic decision factors and additional exploration components, and adaptively selects the key parameters of the variational mode decomposition algorithm, namely the penalty factor and the number of mode layers, with the minimum sample entropy as the fitness function. The signal decomposition submodule is used to adaptively decompose the original acoustic signal using an optimized variational mode decomposition algorithm to obtain multiple intrinsic mode components. The signal reconstruction submodule is used to select effective intrinsic mode components as the reconstructed signal using the Pearson correlation coefficient. The feature extraction submodule is used to extract various parameter indicators of the reconstructed signal to form a leakage feature vector dataset; The operating condition diagnosis submodule is used to build a high-precision classification diagnosis model and input a leak feature vector dataset to classify and diagnose gas leak conditions in pressure vessels.

[0015] A computer memory storing a computer program executable by a computer processor, the computer program performing an adaptive diagnostic method for minor gas leaks in a pressure vessel.

[0016] The beneficial effects of this invention are as follows: 1. The present invention provides an adaptive diagnostic method and system for minor gas leaks in pressure vessels. Based on the acoustic signals of minor gas leaks in pressure vessels collected by a MEMS ultrasonic microphone array, the leak signals are preprocessed using an adaptive signal decomposition algorithm to extract various leak features and input them into a leak diagnosis model, thereby realizing the function of identifying minor leak conditions of various orifice sizes in pressure vessels.

[0017] 2. This invention effectively identifies various micro-leakage conditions with different orifice sizes by extracting multiple indicator features of leaks, thereby improving the accuracy of fault diagnosis for micro-leakage of gas in pressure vessels and the level of safety monitoring in industrial production. It features complete functions, strong adaptability and high identification accuracy, providing maintenance personnel with a reliable means for early detection of micro-leakage of gas in pressure vessels and providing reliable protection for production safety.

[0018] 3. This invention solves the detection defects caused by manual experience in selecting parameters, effectively improves the stability and generalization ability of the micro-aperture leakage diagnosis model, and provides a foundation for subsequent leakage location research, which has great practical application value.

[0019] Of course, any product implementing this invention does not necessarily need to achieve all of the advantages described above at the same time. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a flowchart of an embodiment of the present invention.

[0022] Figure 2 This is a flowchart of an embodiment of the present invention.

[0023] Figure 3 This is an intrinsic mode component diagram obtained by processing the leakage signal of a small aperture according to an embodiment of the present invention.

[0024] Figure 4 This is a confusion matrix diagram of recognition accuracy in an embodiment of the present invention.

[0025] Figure 5 This is a diagram showing the diagnostic results of the test set in an embodiment of the present invention.

[0026] Figure 6 This is a comparison chart of the recognition accuracy of this invention embodiment with other algorithm models in 10 experiments.

[0027] Figure 7 This is a schematic diagram of the detection device according to an embodiment of the present invention. Detailed Implementation

[0028] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0029] Example 1 See Figure 1 The specific steps of an adaptive diagnostic method for minor gas leaks in pressure vessels are as follows: S1: Acquire the original leakage signal through the MEMS ultrasonic microphone array in the acquisition device; The sampling pressure for collecting the leakage signal was 0.6 MPa. The MEMS ultrasonic microphone array was positioned 1 m away from the leakage source to collect the leakage signal, and the sampling frequency was set to 192 kHz.

[0030] S2: A parameter combination of the variational mode decomposition (VMD) algorithm by introducing chaotic population initialization and incorporating dynamic decision factors and additional exploration components to improve the Newton-Raphson optimization algorithm (INRBO). Perform adaptive selection.

[0031] The specific steps for improving INRBO by integrating chaotic population initialization, dynamic decision factors, and additional exploration components are as follows: Indicates the first Random numbers generated by the chaotic iteration. For the control parameters of the Bernoulli chaotic mapping, and The Bernoulli chaotic mapping is used to improve the initialization strategy, and its expression is:

[0032] In the formula: At this point, the Bernoulli chaotic mapping can generate a uniformly distributed sequence with nearly consistent distribution density for different mapping values.

[0033] A dynamic decision factor DF is introduced, with its initial value still set to the default initial value of 0.6 in the NRBO algorithm, and gradually adjusted in subsequent iterations. Its expression is:

[0034] In the formula: This represents the current iteration number. This represents the maximum number of iterations.

[0035] Introducing an additional exploration enhancement component By updating the location Adding an extra term based on the normal distribution random perturbation parameter, its core expression is:

[0036] In the formula: and This represents candidate positions generated based on different search strategies. To introduce dynamic parameters The next update location, and A random number in the range [0,1], used to maintain population diversity.

[0037] S3: The original leakage signal is decomposed into... using the improved INRBO-VMD algorithm. One IMF component.

[0038] S4: Calculate the correlation coefficient between each decomposed IMF component and the original leakage signal, and screen the effective modal component reconstructed signal.

[0039] The expressions for the correlation coefficient and the screening threshold are as follows:

[0040]

[0041] In the formula The representative is the first Each IMF component and the original leakage signal The correlation coefficient, For the first The average value of each IMF component, This represents the average value of the original leakage signal. This represents all intrinsic modal components and the original acoustic signal. correlation coefficient The maximum value in. The threshold for screening IMF components is used to filter correlation coefficients. The IMF component is used as the effective mode reconstruction signal.

[0042] S5: Based on the reconstructed leakage signal, perform time-domain and frequency-domain analysis, extract multi-index feature parameters to construct a feature vector dataset, and divide it into training set and test set according to the ratio of 70% and 30%.

[0043] S6: Construct a high-precision classification and recognition model, and input the leakage feature vector dataset of different micro-pore diameters from step S5 into the model to perform working condition diagnosis of micro-leaking pore diameters.

[0044] The high-precision classification and recognition model is implemented as follows: A high-precision classification and recognition model is constructed by adaptively optimizing the Kernel Extreme Learning Machine (KELM) using the INRBO algorithm, and the regularization coefficient of the KELM is optimized. and kernel function Set the population size to The maximum number of iterations is set to .parameter and The optimization ranges are limited to [0.1, 1000] and [0.1, 30], respectively.

[0045] This embodiment uses the acoustic signals of minor gas leaks in pressure vessels collected by a MEMS ultrasonic microphone array. The leak signals are preprocessed using an adaptive signal decomposition algorithm, and various leak features are extracted and input into the leak diagnosis model. This enables the identification of minor leak conditions in pressure vessels with various orifice sizes.

[0046] Example 2 The steps in this embodiment are the same as in Embodiment 1, the difference being that each step is applied to a specific instance. See also Figure 2Based on the acoustic signals of minor gas leaks in pressure vessels collected by a MEMS ultrasonic microphone array, an adaptive signal decomposition algorithm is used to preprocess the leak signals, extracting various leak features which are then input into a leak diagnosis model to determine the current operating condition of the minor leak in the pressure vessel, thereby achieving the purpose of minor leak fault diagnosis. Specifically, the following steps are included: S1: Acquire the original leakage signal through the MEMS ultrasonic microphone array in the acquisition device; S2: A parameter combination of the variational mode decomposition (VMD) algorithm by introducing chaotic population initialization and incorporating dynamic decision factors and additional exploration components to improve the Newton-Raphson optimization algorithm (INRBO). Perform adaptive selection.

[0047] In the VMD algorithm, the parameter penalty factor and modal layer number It needs to be set up in advance. The value determines the number of modes obtained during VMD decomposition. If it is too small, sufficient effective information cannot be extracted; if it is too large, pseudo-modes are easily generated. (Penalty factor) Used to control the bandwidth characteristics of each IMF component, and set appropriately. A suitable initialization value can improve the decomposition accuracy of the VMD algorithm, while a poor initialization value can lead to the loss of some important signal information. Therefore, to improve the performance of the VMD algorithm, appropriate parameters must be selected for initialization.

[0048] To avoid under- or over-decomposition caused by manual parameter selection based on experience, this invention introduces the INRBO algorithm as a penalty factor for VMD. and modal layer number Optimization is performed by using the minimum sample entropy as the fitness function for the parameter combination of VMD. Adaptive selection is performed. Sample entropy (SampEn) is a quantitative indicator that measures the complexity and regularity of a time series. A lower sample entropy value indicates stronger self-similarity of the sample sequence, while a higher value indicates greater sequence complexity. For a length of... Time series, given embedding dimension and similarity threshold By statistical embedding dimension and similarity probability of time vector sequences and Then, perform a logarithmic operation to obtain the expression for the sample entropy:

[0049] The core idea of ​​NRBO is to define the solution space using a vector set and search the solution space using operators such as the Newton-Raphson Search Rule (NRSR) and the Trap Avoidance Operator (TAO). The NRSR mechanism significantly improves the convergence rate and global search capability of the NRBO algorithm, while the TAO operation effectively avoids local optima. Specifically: The NRBO algorithm randomly generates an initial population within the solution space boundary, with each population consisting of fuzzy decision variables / vectors. The generated random population is as follows:

[0050] In the formula: The first in the population Individual in the first Position in each dimension For population size, To optimize the problem dimensions, and These represent the lower and upper bounds of the variable, respectively. A random number between (0, 1).

[0051] The population state matrix can be described as:

[0052] The NRSR mechanism calculates the objective function. The first and second derivatives at the current iteration point are used, and this derivative information is then used to determine the search direction and step size. It estimates the function's value near the assumed root region based on the first few components of the Taylor series. The updated function root formula is as follows:

[0053] Use positionality replace This can save a significant amount of iteration computation time, and the updated NRSR can be expressed as:

[0054] In the formula: Let be the parameters of a normally distributed random disturbance with variance of one and mean of zero. For disturbance quantity, and These represent the worst and best performance positions, respectively.

[0055] The TAO trap avoidance operation uses a probabilistic perturbation mechanism to enable the algorithm to avoid local optima, thus enhancing global search performance. This operator integrates the global optimal position. With current position vector To update the iteration vector ,Change The position of a uniformly distributed random variable When the time comes, update the solution according to the following rules:

[0056] In the formula: and They are ( 1,1) and ( A consistent random number between 0.5 and 0.5. and It is a random number, calculated by the following formula:

[0057] To further enhance its global optimization capability and adaptive level, this embodiment integrates Bernoulli chaotic mapping population initialization, dynamic decision factors, and additional exploration components, and proposes an improved Newton-Raphson optimization algorithm (INRBO).

[0058] In the NRBO algorithm, the initial positions of individuals in the population are usually randomly generated, making it difficult to guarantee the uniformity of the initial population distribution in the solution space. To improve the global optimization capability of NRBO, Indicates the first Random numbers generated by the chaotic iteration. For the control parameters of the Bernoulli chaotic mapping, and The Bernoulli chaotic mapping is used to improve the initialization strategy, and its expression is:

[0059] In the formula: At this point, the Bernoulli chaotic mapping can generate a uniformly distributed sequence with a nearly consistent distribution density for different mapping values.

[0060] In the NRBO algorithm, the decision factor DF for TAO is typically set to a fixed value of 0.6. However, when solving complex engineering optimization problems, the algorithm usually requires stronger exploration in the early stages and more refined development in the later stages. To better balance development and exploration, a dynamic decision factor DF is introduced, initially set to 0.6 and gradually adjusted in subsequent iterations. Its expression is:

[0061] In the formula: This represents the current iteration number. This represents the maximum number of iterations.

[0062] Introducing this dynamic decision factor allows the exploration intensity to change from increasing to decreasing over time. Utilizing the periodicity of the sine function enables the algorithm to achieve robust exploration in the early stage, vigorous exploration in the middle stage, and convergent exploration in the later stage, which helps to periodically escape possible local optima.

[0063] In the NRBO algorithm, NRSR and TAO are essentially deterministic mechanisms based on gradient and population information. However, when population diversity declines in the later stages, they are still prone to getting trapped in local optima in complex, multi-modal search spaces. To enhance the algorithm's global search capability, an additional exploration enhancement component is introduced. By updating the location This is achieved by adding an extra term based on the parameters of a normally distributed random perturbation. Its core expression is:

[0064] In the formula: and This represents candidate positions generated based on different search strategies. To introduce dynamic parameters The next update location, and A random number in the range [0,1], used to maintain population diversity.

[0065] This component provides a controlled random exploration in each iteration, enhancing the algorithm's ability to escape local optima.

[0066] Initialize the parameters of INRBO-VMD and set the population size. Maximum number of iterations The upper and lower limits of the optimization parameters are set to... , VMD initial noise margin DC component Initialize center frequency parameters Convergence Criterion Tolerance .

[0067] S3: The original leakage signal is decomposed into... using the improved INRBO-VMD algorithm. One IMF component.

[0068] like Figure 3 The figure shows the nine intrinsic mode components obtained by processing the original leakage signal with a small aperture of 0.5 mm in this embodiment.

[0069] S4: Calculate the correlation coefficient between each decomposed IMF component and the original leakage signal, and screen the effective modal component reconstructed signal.

[0070] Use formula Calculate the Pearson correlation coefficient between each decomposed IMF component and the original leakage signal.

[0071] in The representative is the first Each IMF component and the original leakage signal The correlation coefficient, For the first The average value of each IMF component, This represents the average value of the original leakage signal. This represents all intrinsic modal components and the original acoustic signal. correlation coefficient The maximum value in the range. The screening threshold for IMF components. Set as Filtering correlation coefficients The IMF component is used as the effective mode reconstruction signal.

[0072] S5: Based on the reconstructed leakage signal, perform time-domain and frequency-domain analysis, extract multi-index feature parameters to construct a feature vector dataset, and divide it into training set and test set according to the ratio of 70% and 30%.

[0073] Based on the reconstructed leakage signal, time-domain and frequency-domain analysis is performed. Time-domain features are divided into dimensional and dimensionless features, with dimensionless time-domain features also known as shape features. Frequency-domain features reveal the spectral distribution of leakage energy and are key to identifying the leakage aperture. In this embodiment, the formulas for extracting time-domain features are shown in Table 1, and the frequency-domain features are shown in Table 2, for a total of 15 features.

[0074] Table 1

[0075] Table 2

[0076] in, This is a time-domain sequence of the leakage signal from a small aperture. The total number of sampling points. It is the corresponding frequency of the signal in the frequency spectrum. It is the corresponding amplitude of the signal in the frequency spectrum. ~ Four dimensional time-domain features representing the extracted leakage signal. ~ Five dimensionless time-domain features representing the extracted signal. ~ Six frequency domain features represent the extracted signal. For micro-leakage signals with apertures of 0.1mm, 0.5mm, 1mm, and 2mm, the characteristic indices are... ~ Combined to form a multidimensional feature vector dataset.

[0077] S6: Construct a high-precision classification and recognition model, and input the leakage feature vector dataset of different micro-pore diameters from step S5 into the model to perform working condition diagnosis of micro-leaking pore diameters.

[0078] nuclear parameters and regularization coefficient This will directly affect KELM performance. Control generalization ability; Inappropriate settings can lead to model overfitting or excessive training errors. Therefore, this embodiment uses INRBO adaptive optimization of the above parameters to construct an INRBO-KELM high-precision classification and recognition model for identifying the diameter of minute gas leaks in pressure vessels.

[0079] The 600 feature vector datasets after feature extraction from the reconstructed signals were labeled, and the corresponding label categories are shown in Table 3. Data was randomly selected at a ratio of 70% to 30% to split into training and test sets, which served as input to the INRBO-KELM micro-aperture leakage signal recognition model.

[0080] Table 3

[0081] The INRBO optimization of KELM requires initializing the parameters first, setting the population size to... The maximum number of iterations is set to .parameter and The optimization ranges are limited to [0.1, 1000] and [0.1, 30], respectively. The fitness function is set to minimize the recognition error rate during the optimization iteration process.

[0082] The confusion matrix of the INRBO-KELM model's test set classification accuracy is as follows: Figure 4 As shown, the model achieves 100% accuracy in identifying pore size leaks for label categories 3 and 4. Only a small number of label category 2 samples were incorrectly identified as label category 1; that is, one sample with a 0.5mm pore size leak was misidentified as a 0.1mm pore size. All other pore size samples were correctly identified. The confusion matrix mainly contains true positives. True negative False positives False negative To quantitatively analyze the leakage pore size classification and identification performance of the INRBO-KELM model, four parameters can be used to derive the formula for calculating the identification accuracy:

[0083] Recognition accuracy refers to the percentage of correctly predicted samples out of the total sample size, reflecting the overall discriminative ability. Higher accuracy indicates better overall predictive performance. Calculations show that the INRBO-KELM model has a recognition accuracy of 99.44%, and the diagnostic results are as follows: Figure 5 As shown, it has high diagnostic accuracy for leaks with a diameter of less than 2 mm.

[0084] Furthermore, this embodiment underwent multiple experimental comparisons with other algorithm models, such as... Figure 6 As shown in the figure, based on the results of 10 comparative experiments, it can be seen that the INRBO-VMD-KELM pressure vessel gas micro-orifice leakage diagnostic model of this embodiment has a higher average recognition accuracy for micro-leakage in pressure vessels with orifice diameters of 0.1~2mm. Therefore, the adaptive diagnostic method for micro-leakage faults in pressure vessels adopted in this invention can adaptively detect early-stage gas leaks in pressure vessels with different micro-leakage orifice diameters, and can be applied to practical engineering.

[0085] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0086] Example 3 like Figure 7 As shown, this embodiment is used to build a pressure vessel gas micro-aperture leakage detection device platform to implement the principle of the above method embodiment. It includes a pressure vessel leakage unit and a signal acquisition and processing unit. The pressure vessel leakage unit includes an air compressor, a pressure reducing valve, a pressure storage tank, a ball valve, and a 0.1mm-2mm aperture leakage nozzle. The signal acquisition and processing unit consists of an Infineon MEMS ultrasonic microphone array, a signal acquisition driver board, and a computer.

[0087] The working pressure of the experimental platform was adjusted to 0.6 MPa using a pressure reducing valve. A MEMS microphone array was positioned 1 meter from the leakage source to collect leakage signals, which were then transmitted to a computer. The sampling frequency was set to 192 kHz. The Matlab R2022b platform was used to process the minute leakage signals and identify their apertures. 150 sets of leakage signals each with apertures of 0.1 mm, 0.5 mm, 1 mm, and 2 mm were collected as data samples. Each set of leakage signals had 2048 sampling points, resulting in a total of 600 sets of data. These data were then input into the INRBO-VMD-KELM model for adaptive leakage fault diagnosis.

[0088] It should be noted that, depending on the implementation needs, the various steps / components described in this application can be broken down into more steps / components, or two or more steps / components or parts of the operation of steps / components can be combined into new steps / components to achieve the purpose of this invention.

[0089] This embodiment also includes a processor, a communication interface, a memory, and a communication bus; wherein the processor, the communication interface, and the memory communicate with each other through the communication bus; the memory stores a computer program, which, when executed by the processor, causes the processor to perform the steps of an adaptive diagnostic method for minor gas leaks in a pressure vessel.

[0090] This embodiment also provides a computer-readable storage medium storing executable instructions that, when executed by a processor, enable the processor to implement an adaptive diagnostic method for minor gas leaks in a pressure vessel.

[0091] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects.

[0092] Furthermore, this application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0093] This application is described with reference to the flowchart of the method and computer program product according to Embodiment 1 and the block diagram of the device (system) according to Embodiment 3. It should be understood that each step or block in the flowchart or block diagram, as well as combinations of steps or blocks in the flowchart or block diagram, can be implemented by computer program instructions.

[0094] These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing device to produce a machine, such that the instructions, which are executable by the processor of the computer or other programmable data processing device, produce instructions for implementing the process. Figure 1 One or more processes or boxes Figure 1 An adaptive diagnostic system for minor gas leaks in pressure vessels, specifying functions within one or more boxes.

[0095] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes or boxes Figure 1 The function specified in one or more boxes.

[0096] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes or boxes Figure 1 The steps of an adaptive diagnostic method for minor gas leaks in a pressure vessel are specified in one or more boxes.

[0097] The above embodiments are only used to illustrate the design concept and features of the present invention, and their purpose is to enable those skilled in the art to understand the content of the present invention and implement it accordingly. The protection scope of the present invention is not limited to the above embodiments. Therefore, all equivalent changes or modifications made based on the principles and design ideas disclosed in the present invention are within the protection scope of the present invention.

Claims

1. A method for adaptive diagnosis of a micro-leakage fault of a gas in a pressure vessel, characterized in that: It comprises the following steps: S1: collecting the original acoustic signal of the gas leakage of the pressure vessel; S2: fusing the chaos mapping population initialization, the dynamic decision factor and the additional exploration component to improve the Newton-Raphson optimization algorithm, taking the minimum sample entropy as the fitness function to adaptively select the key parameters of the variational mode decomposition algorithm, including the penalty factor and the mode number; S3: using the optimized variational mode decomposition algorithm to adaptively decompose the original acoustic signal to obtain multiple intrinsic mode components; S4: screening the effective intrinsic mode components as the reconstructed signal through the Pearson correlation coefficient; S5: extracting multiple parameter indexes of the reconstructed signal to form a leakage feature vector dataset; S6: constructing a high-precision classification diagnosis model, and inputting the leakage feature vector dataset to classify and diagnose the working condition of the gas leakage of the pressure vessel.

2. The method according to claim 1, wherein the method is characterized by: The step S2 is performed as follows: representing the first chaotic iteration, is a control parameter of the Bernoulli chaotic mapping, and The expression for improving the population initialization by using the Bernoulli chaotic mapping is as follows: 。 3. The method according to claim 1, wherein the method further comprises: The step S2, is the current iteration number, is the maximum iteration number, a dynamic decision factor DF is introduced, the initial value is set as the default initial value in the Newton-Raphson optimization algorithm, and is adjusted in iteration, and the expression of the dynamic decision factor DF is: ​ 。 4. The method according to claim 1, wherein the method further comprises: The step S2, and represent candidate positions generated based on different search strategies, for introducing dynamic parameters updated positions, and is a random number in the range of [0, 1] for maintaining population diversity; an additional exploration enhancement component is introduced , by adding an additional term based on a normally distributed random disturbance parameter to the updated position , the expression is: ​ 。 5. The method of claim 1, wherein: In the step S4, the specific steps are: calculating the Pearson correlation coefficient between each intrinsic mode component after decomposition and the original acoustic signal, and screening the mode components with the Pearson correlation coefficient greater than the screening threshold as the effective mode components to reconstruct the signal.

6. The method according to claim 5, wherein: The correlation coefficient of the m-th eigenmode component and the original acoustic signal The m-th eigenmode component is the m-th component of the eigenvector The m-th eigenmode component is the m-th component of the eigenvector The average value of the m-th eigenmode component The average value of the m-th eigenmode component The average value of the m-th eigenmode component The average value of the m-th eigenmode component The maximum value of the correlation coefficients of all eigenmode components and the original acoustic signal The maximum value of the correlation coefficients of all eigenmode components and the original acoustic signal The maximum value of the correlation coefficients of all eigenmode components and the original acoustic signal The maximum value of the correlation coefficients of all eigenmode components and the original acoustic signal ； Screening threshold for eigenmode components The expression for the screening threshold is 。 7. The method of claim 1, wherein: In the step S5, the specific steps are: performing time domain and frequency domain analysis on the reconstructed signal, extracting multiple index feature parameter combinations to construct a leakage feature vector dataset; the multiple index feature parameter combinations include time domain features and frequency domain features; the time domain features include dimensional feature indexes and dimensionless feature indexes; the dimensionless time domain features are used to represent shape features; the frequency domain features are used to represent the spectral distribution of leakage energy; the leakage feature vector dataset is divided into a training set and a test set according to a certain proportion.

8. The method of claim 1, wherein: The step S6 is to adaptively optimize the regularization coefficient of the kernel extreme learning machine by Newton-Raphson optimization algorithm and kernel function Construct a high-precision classification diagnosis model; set the population size and the maximum number of iterations, limit the optimization range of parameters and Minimize the recognition error rate and set it as the fitness function.

9. A pressure vessel gas micro-leakage fault adaptive diagnosis system, characterized in that: a data acquisition submodule is configured to collect the original acoustic signal of the gas leakage of the pressure vessel; a parameter selection submodule is configured to fuse the chaos mapping population initialization, the dynamic decision factor and the additional exploration component to improve the Newton-Raphson optimization algorithm, and take the minimum sample entropy as the fitness function to adaptively select the key parameters of the variational mode decomposition algorithm, including the penalty factor and the mode number; a signal decomposition submodule is configured to use the optimized variational mode decomposition algorithm to adaptively decompose the original acoustic signal to obtain multiple intrinsic mode components; a signal reconstruction submodule is configured to screen the effective intrinsic mode components as the reconstructed signal through the Pearson correlation coefficient; a feature extraction submodule is configured to extract multiple parameter indexes of the reconstructed signal to form a leakage feature vector dataset; a working condition diagnosis submodule is configured to construct a high-precision classification diagnosis model, and input the leakage feature vector dataset to classify and diagnose the working condition of the gas leakage of the pressure vessel.

10. A computer memory, characterized by: It has a computer program stored therein, which can be executed by a computer processor, and the computer program executes a pressure vessel gas micro-leakage fault adaptive diagnosis method according to any one of claims 1 to 8.

Images