Fault feature extraction method and device of reciprocating compressor, equipment and medium

Abstract

The invention relates to the technical field of reciprocating compressors, and discloses a fault feature extraction method, device and equipment of a reciprocating compressor and a medium, and the method comprises the steps: obtaining thermal parameters and vibration parameters of the reciprocating compressor; respectively constructing a high-dimensional thermal characteristic space and a high-dimensional vibration characteristic space according to the thermal parameters and the vibration parameters; performing nonlinear feature dimension reduction on the high-dimensional thermal feature space and the high-dimensional vibration feature space based on kernel principal component analysis to obtain principal components of thermal parameters and principal components of vibration parameters; constructing a high-dimensional reciprocating compressor feature space according to the principal components of the thermal parameters and the principal components of the vibration parameters; and feature extraction is carried out on the high-dimensional reciprocating compressor feature space based on t distribution-random adjacent embedding, and fault features of the reciprocating compressor are obtained. Potential fault features can be mined in the thermal parameters and the vibration parameters, and the recognition rate of the reciprocating compressor on common faults is increased.

Description

Technical Field

[0001] The present application relates to the technical field of reciprocating compressors, and in particular to a method, device, equipment and medium for extracting fault features of a reciprocating compressor. Background Art

[0002] At present, reciprocating compressors have become irreplaceable key equipment in the process flow of the petrochemical industry with their stable performance, excellent driving ability, wide displacement range and efficient equipment operation efficiency. Therefore, ensuring the efficient and safe operation of reciprocating compressors is of great significance to the smooth progress of production and life.

[0003] Due to the complexity of the structure of reciprocating compressors and the diversity of manifestations of reciprocating compressor faults, it is difficult to effectively diagnose faults based on a single vibration feature. Therefore, the accuracy of reciprocating compressor fault diagnosis in the prior art is not high. Summary of the invention

[0004] The purpose of the embodiments of the present application is to provide a method, device, equipment and medium for extracting fault features of a reciprocating compressor, so as to solve the defect of low accuracy of fault diagnosis of reciprocating compressors in the prior art.

[0005] In order to achieve the above-mentioned object, the first aspect of the present application provides a method for extracting fault features of a reciprocating compressor, the method comprising:

[0006] Obtain the thermal parameters and vibration parameters of the reciprocating compressor, wherein the thermal parameters include cylinder pressure, fuel pressure, valve temperature, speed, torque and power, and the vibration parameters include root mean square, true peak, kurtosis, skewness, peak factor, margin factor, waveform factor, pulse factor in the time domain and center of gravity frequency, mean square frequency, frequency variance in the frequency domain;

[0007] According to the thermal parameters and vibration parameters, a high-dimensional thermal feature space and a high-dimensional vibration feature space are constructed respectively;

[0008] Based on kernel principal component analysis, nonlinear feature dimensionality reduction is performed on high-dimensional thermal feature space and high-dimensional vibration feature space respectively to obtain the principal components of thermal parameters and vibration parameters.

[0009] A high-dimensional reciprocating compressor feature space is constructed based on the principal components of thermal parameters and vibration parameters;

[0010] Based on t-distribution-random neighbor embedding, feature extraction is performed on the high-dimensional reciprocating compressor feature space to obtain the fault characteristics of the reciprocating compressor.

[0011] In the embodiment of the present application, nonlinear feature dimensionality reduction is performed on the high-dimensional thermal feature space and the high-dimensional vibration feature space based on kernel principal component analysis to obtain the principal components of the thermal parameters and the principal components of the vibration parameters, including:

[0012] According to the kernel function, determine the kernel matrix;

[0013] Centralize the kernel matrix to obtain a centralized kernel matrix;

[0014] Determine the eigenvalues ​​and corresponding eigenvectors of the centralized kernel matrix, and arrange each eigenvector according to the eigenvalues;

[0015] Based on the contribution rate of each eigenvalue, determine the cumulative contribution rate;

[0016] The eigenvectors whose cumulative contribution rate is greater than or equal to the preset cumulative contribution rate are determined as the principal components of the thermal parameters and the vibration parameters.

[0017] In the embodiment of the present application, the kernel function is a Gaussian radial basis kernel function.

[0018] In the embodiment of the present application, feature extraction is performed on the feature space of the high-dimensional reciprocating compressor based on t-distribution-random neighbor embedding to obtain the fault features of the reciprocating compressor, including:

[0019] Mapping the principal components of the thermal parameters and the vibration parameters in the high-dimensional reciprocating compressor feature space to the low-dimensional space to obtain the principal components of the thermal parameters and the vibration parameters in the low-dimensional space;

[0020] Determine the joint probability distribution in the feature space of the high-dimensional reciprocating compressor according to the Gaussian distribution;

[0021] Determine the joint probability distribution in low-dimensional space based on t-distribution;

[0022] The KL divergence is minimized by the gradient descent method, and the principal components of the thermal parameters and the vibration parameters in the low-dimensional space are updated to obtain the fault characteristics of the reciprocating compressor.

[0023] In the embodiment of the present application, the formula for the joint probability distribution in the high-dimensional reciprocating compressor feature space includes:

[0024]

[0025] In the formula, p ij represents the joint probability distribution in the feature space of high-dimensional reciprocating compressors, p j|i represents the probability of j occurring under given conditions i, p i|j represents the probability of i occurring under the given j condition, n represents the number in the high-dimensional reciprocating compressor feature space, x i 、xj and x k represents a data point in the high-dimensional reciprocating compressor feature space, σ i 2 Indicates x i The Gaussian variance centered at

[0026] The formula for the joint probability distribution in low-dimensional space includes:

[0027]

[0028] In the formula, q ij represents the joint probability distribution in low-dimensional space, y i ,y i and k Represents a data point in the high-dimensional reciprocating compressor feature space.

[0029] In the embodiment of the present application, the KL divergence is minimized by the gradient descent method, including:

[0030] Minimize the KL divergence based on the gradient formula, where the gradient formula includes:

[0031]

[0032] In the formula, C represents KL divergence, y i and i represents the data points in the low-dimensional space, p ij represents the joint probability distribution of data points i and j in the high-dimensional reciprocating compressor feature space, q ij Represents the joint probability distribution of data points i and j in the low-dimensional space.

[0033] In an embodiment of the present application, the principal components of the thermodynamic parameters and the principal components of the vibration parameters in the low-dimensional space are updated by an updating formula, wherein the updating formula includes:

[0034]

[0035] Where Y represents the principal components of the thermal parameters and the vibration parameters in the low-dimensional space, t represents the number of iterations, η represents the learning rate, and α(t) is the momentum factor.

[0036] A second aspect of the present application provides a fault feature extraction device for a reciprocating compressor, the device comprising:

[0037] An acquisition module is used to acquire thermal parameters and vibration parameters of the reciprocating compressor, wherein the thermal parameters include cylinder pressure, fuel pressure, valve temperature, speed, torque and power, and the vibration parameters include root mean square, true peak, kurtosis, skewness, peak factor, margin factor, waveform factor, pulse factor in the time domain and center of gravity frequency, mean square frequency, frequency variance in the frequency domain;

[0038] The first construction module is used to construct a high-dimensional thermal feature space and a high-dimensional vibration feature space according to the thermal parameters and the vibration parameters respectively;

[0039] The first data processing module is used to perform nonlinear feature dimensionality reduction on the high-dimensional thermal feature space and the high-dimensional vibration feature space based on kernel principal component analysis to obtain the principal components of the thermal parameters and the principal components of the vibration parameters;

[0040] The second construction module is used to construct a high-dimensional reciprocating compressor feature space according to the principal components of the thermal parameters and the principal components of the vibration parameters;

[0041] The second data processing module is used to extract features from the high-dimensional reciprocating compressor feature space based on t-distribution-random neighbor embedding to obtain fault features of the reciprocating compressor.

[0042] A third aspect of the present application provides a fault feature extraction device for a reciprocating compressor, comprising:

[0043] a memory configured to store instructions;

[0044] The processor is configured to call instructions from the memory and implement the fault feature extraction method for the reciprocating compressor according to the first aspect when executing the instructions.

[0045] A fourth aspect of the present application provides a machine-readable storage medium having instructions stored thereon, the instructions being used to enable a machine to execute the method for extracting fault features of a reciprocating compressor according to the first aspect above.

[0046] Through the above technical solution, the thermal parameters and vibration parameters generated by the reciprocating compressor during operation are deeply mined to reveal and extract the fault characteristics hidden in the thermal parameters and vibration parameters, thereby improving the accuracy of identifying a series of common faults of reciprocating compressors.

[0047] Other features and advantages of the embodiments of the present application will be described in detail in the subsequent specific implementation section. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the following specific implementations, they are used to explain the embodiments of the present application, but do not constitute a limitation on the embodiments of the present application. In the accompanying drawings:

[0049] Figure 1 A schematic diagram of a flow chart of a method for extracting fault features of a reciprocating compressor according to an embodiment of the present application is schematically shown;

[0050] Figure 2Schematically showing a schematic diagram of pressure curves under four gas valve states according to an embodiment of the present application;

[0051] Figure 3 A schematic diagram of temperature curves under four gas valve states according to an embodiment of the present application is shown;

[0052] Figure 4 Schematically shows waveform diagrams under four gas valve states according to an embodiment of the present application;

[0053] Figure 5 Schematically shows a spectrum diagram under four gas valve states according to an embodiment of the present application;

[0054] Figure 6 A schematic diagram of the clustering effect of the nonlinear characteristics of the thermal parameters before dimensionality reduction according to an embodiment of the present application is schematically shown;

[0055] Figure 7 A schematic diagram of the clustering effect of the nonlinear characteristics of vibration parameters before dimensionality reduction according to an embodiment of the present application is shown;

[0056] Figure 8 A schematic diagram of the clustering effect after the nonlinear characteristics of the thermal parameters are reduced in dimension according to an embodiment of the present application is shown;

[0057] Fig. 9 A schematic diagram of the clustering effect after the dimension reduction of the nonlinear characteristics of the vibration parameters according to the embodiment of the present application is schematically shown;

[0058] Fig.10 The schematic diagram shows the clustering effect after feature extraction of the feature space of a high-dimensional fuel-driven reciprocating compressor based on t-distribution-random neighbor embedding according to an embodiment of the present application. DETAILED DESCRIPTION

[0059] In order to make the purpose, technical scheme and advantages of the embodiments of the present application clearer, the technical scheme in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. It should be understood that the specific implementation methods described herein are only used to illustrate and explain the embodiments of the present application, and are not used to limit the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0060] It should be noted that the acquisition, transmission, storage, use, and processing of data in the technical solution of this application are in compliance with the relevant provisions of national laws and regulations. In the embodiments of this application, some existing solutions in the industry such as certain software, components, and models may be mentioned, which should be considered as exemplary. Their purpose is only to illustrate the feasibility of implementing the technical solution of this application, but it does not mean that the applicant has or will necessarily use the solution.

[0061] It should be noted that if the embodiments of the present application involve directional indications (such as up, down, left, right, front, back...), the directional indications are only used to explain the relative position relationship, movement status, etc. between the components under a certain specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indication will also change accordingly.

[0062] In addition, if there are descriptions involving "first", "second", etc. in the embodiments of the present application, the descriptions of "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or suggesting their relative importance or implicitly indicating the number of technical features indicated. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In addition, the technical solutions between the various embodiments can be combined with each other, but they must be based on the ability of ordinary technicians in the field to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be deemed that such combination of technical solutions does not exist and is not within the scope of protection required by this application.

[0063] This application aims to reveal and extract hidden fault characteristics in reciprocating compressors by deeply analyzing the thermal parameters and vibration parameters generated by the working reciprocating compressors. This process can not only enhance the comprehensive understanding of the internal operating status of the reciprocating compressor, but also significantly improve the accuracy of identifying a series of common faults in reciprocating compressors. By accurately identifying the early signs of these faults, this application can provide a scientific basis for the preventive maintenance of reciprocating compressors, effectively extend the service life of the equipment, reduce unplanned downtime, and ultimately improve production efficiency and economic benefits.

[0064] Figure 1 The flowchart of a method for extracting fault features of a reciprocating compressor according to an embodiment of the present application is schematically shown. Figure 1 As shown, an embodiment of the present application provides a method for extracting fault features of a reciprocating compressor, which may include the following steps.

[0065] S110: Acquire thermal parameters and vibration parameters of the reciprocating compressor, wherein the thermal parameters include cylinder pressure, fuel pressure, valve temperature, speed, torque and power; and the vibration parameters include root mean square, true peak, kurtosis, skewness, peak factor, margin factor, waveform factor, pulse factor in the time domain and center of gravity frequency, mean square frequency, and frequency variance in the frequency domain.

[0066] Among them, the thermal parameters mainly reflect the thermodynamic state of the reciprocating compressor during operation. By installing relevant sensors at key locations such as the air inlet, air outlet and compression chamber of the reciprocating compressor, the pressure, temperature and speed of the gas are monitored in real time to obtain the thermal parameters. The vibration parameters mainly reflect the mechanical vibration state of the compressor during operation. By installing vibration sensors on key components of the reciprocating compressor (such as cylinders, crankshafts, connecting rods, etc.), the vibration conditions are monitored in real time, and the signals measured by the vibration sensors are subjected to spectrum analysis and time domain analysis to extract the required vibration parameters. The calculation formulas for the root mean square, true peak, kurtosis, skewness, peak factor, margin factor, waveform factor, and pulse factor in the time domain are as follows:

[0067]

[0068] In the formula, x rms represents the root mean square, x(n) represents the nth signal in the signal sequence, and N represents the total number of signal values.

[0069] x p =max|x(n)|

[0070] In the formula, x p Represents the true peak value, that is, the maximum signal value in the signal sequence.

[0071]

[0072] In the formula, x kur Represents kurtosis, which is a statistic used to describe the steepness of the signal distribution. represents the arithmetic mean of the signal sequence, x std represents the standard deviation of the signal sequence.

[0073]

[0074] In the formula, x ske Represents skewness, which is a statistic used to describe the symmetry of the signal distribution shape.

[0075] x CF =x p / x rms

[0076] In the formula, x CFIndicates the crest factor, which is used to describe the size of the peak value in the signal relative to the root mean square value.

[0077]

[0078] In the formula, x CIF Represents the margin factor, which is used to describe the stability and anti-interference ability of the signal.

[0079]

[0080] In the formula, x SF Indicates the form factor.

[0081]

[0082] In the formula, x IF Represents the impulse factor.

[0083] The sampling frequency of this embodiment is 25600 Hz.

[0084] The calculation formulas for the centroid frequency, mean square frequency, and frequency variance in the frequency domain are as follows:

[0085]

[0086] In the formula, FC represents the center of gravity frequency, K represents the number of spectral lines, and f k represents the frequency of the kth spectral line, and s(k) represents the spectrum of x(n).

[0087]

[0088] Where MSF represents mean square frequency.

[0089]

[0090] Where VF represents the frequency variance.

[0091] This embodiment takes valve faults (normal state, valve notch, valve plate breakage, valve spring shortage) as an example to specifically explain a method for extracting fault features of a reciprocating compressor in this embodiment.

[0092] The thermal parameters include inlet pressure, outlet pressure, valve temperature, inlet temperature and outlet temperature under four conditions: normal valve state, valve gap, valve plate breakage and valve spring shortage. 75 samples were selected for each of the four conditions, for a total of 300 samples. Figure 2 The schematic diagram of the pressure curve under four gas valve states according to the embodiment of the present application is shown as follows: Figure 2 As shown, curve 10 represents the outlet pressure and curve 20 represents the inlet pressure. Figure 3The schematic diagram of temperature curves under four gas valve states according to the embodiment of the present application is shown as follows: Figure 3 As shown, curve 30 represents the outlet temperature, curve 40 represents the valve temperature, and curve 50 represents the inlet temperature.

[0093] The vibration parameters include the time domain root mean square, true peak, kurtosis, skewness, peak factor, margin factor, waveform factor, pulse factor and frequency domain center of gravity frequency, mean square frequency and frequency variance under four states: normal valve state, valve gap, valve plate fracture and valve spring deficiency. 75 samples are selected for each of the four states, totaling 300 samples. Figure 4 Schematic diagrams of waveforms under four valve states according to an embodiment of the present application are shown, such as Figure 4 As shown, the waveform diagrams include the waveform diagram under normal state, the waveform diagram under valve notch, the waveform diagram under valve rupture and the waveform diagram under valve spring shortage. The time domain root mean square, true peak value, kurtosis, skewness, peak factor, margin factor, waveform factor and pulse factor can be obtained from the waveform diagrams. Figure 5 The schematic diagram of the spectrum under four valve states according to the embodiment of the present application is shown as follows: Figure 5 As shown, the spectrum diagrams include the spectrum diagram under normal state, the spectrum diagram under valve gap, the spectrum diagram under valve rupture and the spectrum diagram under valve spring shortage. The frequency domain center frequency, mean square frequency and frequency variance can be obtained from the spectrum diagram. Among them, the waveform length is 5678 points and the sampling frequency is 12.8k.

[0094] Among them, the thermal parameters extracted from the pressure curves and temperature curves under the four valve states are as follows: Figure 6 As shown, Figure 6 The schematic diagram shows the clustering effect of the nonlinear characteristics of the thermal parameters before dimensionality reduction according to the embodiment of the present application. The data points of the thermal parameters are scattered in the original thermal characteristic space. The vibration parameters extracted from the waveforms and spectrum diagrams under the four valve states are as follows: Figure 7 As shown, Figure 7 The schematic diagram shows the clustering effect of the nonlinear characteristics of the vibration parameters before dimensionality reduction according to the embodiment of the present application, and the various data points of the vibration parameters are scattered in the original vibration feature space.

[0095] S120: constructing a high-dimensional thermal feature space and a high-dimensional vibration feature space according to the thermal parameters and the vibration parameters respectively.

[0096] Among them, when constructing a high-dimensional thermal feature space based on thermal parameters, preprocessing operations such as denoising and de-meaning are performed on the thermal parameters to improve data quality, and standardization or normalization is performed to ensure that different parameters are compared on the same scale. When constructing a high-dimensional vibration feature space based on vibration parameters, preprocessing operations such as denoising and de-meaning are performed on the vibration parameters to improve data quality, and standardization or normalization is performed to ensure that different parameters are compared on the same scale.

[0097] Based on the Gaussian radial basis kernel function, the parameter data points x1, x2, ..., x N Mapped to the high-dimensional feature space, we get the data points Φ(x1),Φ(x2),…,Φ(x N ), the mapping process satisfies the centralization condition

[0098] Gaussian radial basis kernel function:

[0099]

[0100] In the formula, K(x i ,x j ) represents the data point x in the high-dimensional thermal feature space or high-dimensional vibration feature space i and data point x j The distance between them, σ represents the standard deviation. In the high-dimensional thermal feature space, σ = 0.1, and in the high-dimensional vibration feature space, σ = 0.5.

[0101] In this embodiment, mapping is performed based on thermal parameters and vibration parameters, respectively. The thermal parameters and vibration parameters correspond to a number of data points in the high-dimensional thermal feature space and the high-dimensional vibration feature space, respectively, and each data point in the high-dimensional thermal feature space and the high-dimensional vibration feature space corresponds to the position information of each data point.

[0102] S130: Based on kernel principal component analysis, nonlinear feature dimensionality reduction is performed on the high-dimensional thermal feature space and the high-dimensional vibration feature space respectively to obtain the principal components of the thermal parameters and the principal components of the vibration parameters.

[0103] Among them, the basic principle of Kernel Principal Component Analysis (KPCA) is to make the originally linearly inseparable thermodynamic parameters become linearly separable in the high-dimensional thermodynamic feature space, and to make the originally linearly inseparable vibration parameters become linearly separable in the high-dimensional vibration feature space. Then, the principal component analysis (PCA) method is applied in the high-dimensional thermodynamic feature space and the high-dimensional vibration feature space to extract linear features. This method utilizes the properties of the kernel function and implicitly calculates the inner product of the data points corresponding to the thermodynamic parameters and the vibration parameters in the high-dimensional thermodynamic feature space and the high-dimensional vibration feature space, respectively, thereby avoiding complex calculations directly in the high-dimensional thermodynamic feature space and the high-dimensional vibration feature space.

[0104] Further, S130 includes the following steps:

[0105] Step 131: Determine a kernel matrix according to the kernel function.

[0106] Among them, common kernel functions include Sigmoid kernel function, polynomial kernel function and Gaussian radial basis kernel function. Sigmoid kernel function is similar to the activation function in neural network and is suitable for certain specific nonlinear problems. Polynomial kernel function can handle nonlinear relationships in data, but may require higher polynomial degrees to capture complex patterns. Gaussian radial basis kernel function is a commonly used nonlinear kernel function that can handle a wide range of data distribution and complexity. The thermal parameters and vibration parameters in this embodiment are nonlinear data. In order to improve the calculation efficiency, the kernel function of this embodiment is Gaussian radial basis kernel function, and the expression can refer to the following formula:

[0107] K(x i ,x j )=[b·s(x i ,x j )+c] d

[0108] In the above formula, K(x i ,x j ) represents the data point x in the high-dimensional thermal feature space or high-dimensional vibration feature space i and data point x j The distance between them, s(x i ,x j ) represents the data point x in the original space i and data point x j The distance between them; b and c are constant terms, which are used to adjust the shape and range of the kernel function; d represents the exponential term, which is used to control the strength of the nonlinear transformation. The kernel matrix consists of all K(x i ,xj )composition.

[0109] Step 132: Centralize the kernel matrix to obtain a centralized kernel matrix.

[0110] Among them, the kernel matrix K is centralized to obtain the centralized kernel matrix K c :

[0111] K c =Kl N K-Kl N +l N K N

[0112] In the above formula, K c represents the centralized kernel matrix; K represents the kernel matrix; l N It is an N×N matrix, and each element is 1 / N.

[0113] Step 133: Determine the eigenvalues ​​and corresponding eigenvectors of the centralized kernel matrix, and arrange each eigenvector according to the eigenvalues.

[0114] Among them, the calculation of the centralized kernel matrix K c The eigenvalues ​​λ1,λ2,…,λ N and the corresponding eigenvectors v1,v,…,v N , and each eigenvector v1,v,…,v N Sort in descending order.

[0115] Step 134: Determine the cumulative contribution rate based on the contribution rate of each eigenvalue.

[0116] Among them, each eigenvector v1,v,…,v is calculated N The contribution rate of the , and determine the cumulative contribution rate r1, r2, ..., r N , the cumulative contribution rate expression can refer to the following formula:

[0117]

[0118] In the above formula, r k represents the cumulative contribution rate, represents the sum of the contribution rates of the first k eigenvectors, Represents the sum of the contribution rates of all eigenvectors.

[0119] Step 135: Determine the eigenvectors whose cumulative contribution rates are greater than or equal to the preset cumulative contribution rates as the principal components of the thermal parameters and the vibration parameters.

[0120] The preset cumulative contribution rate is set to 95%, that is, when the cumulative contribution rate reaches 95%, the first t eigenvectors with cumulative contribution rates equal to or exceeding 95% are selected as the principal components of the thermal parameters. The principal component determination method of the vibration parameters is the same as above.

[0121] Figure 8 The schematic diagram shows the clustering effect of the nonlinear characteristics of the thermal parameters after dimensionality reduction according to the embodiment of the present application. Figure 8 As shown in the figure, the main components of the thermal parameters are Figure 6 In comparison, the clustering effect of nonlinear feature dimensionality reduction of high-dimensional thermal feature space based on kernel principal component analysis is better than the clustering effect of the original thermal feature space.

[0122] Fig. 9 The schematic diagram shows the clustering effect of the vibration parameter nonlinear characteristics after dimensionality reduction according to the embodiment of the present application. Fig. 9 As shown in the figure, the main components of the vibration parameters are Figure 7 In comparison, the clustering effect of nonlinear feature dimensionality reduction of high-dimensional vibration feature space based on kernel principal component analysis is better than the clustering effect of the original vibration feature space.

[0123] Through this method, the dimension of parameters can be effectively reduced, while most of the information in thermal parameters and vibration parameters can be retained to the maximum extent.

[0124] S140: Constructing a high-dimensional reciprocating compressor feature space according to the principal components of the thermal parameters and the principal components of the vibration parameters.

[0125] Among them, the principal components of thermal parameters and vibration parameters are combined to construct a high-dimensional reciprocating compressor feature space. The establishment of the high-dimensional reciprocating compressor feature space provides support for the subsequent acquisition of the fault characteristics of the reciprocating compressor.

[0126] S150: extracting features from the high-dimensional reciprocating compressor feature space based on t-distribution-random neighbor embedding to obtain fault features of the reciprocating compressor.

[0127] Among them, the principle of t-Distributed Stochastic Neighbor Embedding (t-SNE) is to maintain the local structure between data points, that is, points that are close in the high-dimensional reciprocating compressor feature space should also be kept close in the low-dimensional space. It represents the similarity of points in high-dimensional and low-dimensional spaces through probability distribution, and finds the best low-dimensional representation by minimizing the difference in the joint probability distribution of the two spaces.

[0128] Further, S150 includes the following steps:

[0129] Step 151: Map the principal components of the thermal parameters and the principal components of the vibration parameters in the high-dimensional reciprocating compressor feature space to the low-dimensional space to obtain the principal components of the thermal parameters and the principal components of the vibration parameters in the low-dimensional space.

[0130] Step 152: Determine the joint probability distribution in the high-dimensional reciprocating compressor feature space according to the Gaussian distribution.

[0131] Among them, the formula for the joint probability distribution in the high-dimensional reciprocating compressor feature space includes:

[0132]

[0133] In the formula, p ij represents the joint probability distribution in the feature space of high-dimensional reciprocating compressors, p j|i represents the probability of j occurring under given conditions i, p i|j represents the probability of i occurring under the given j condition, n represents the number in the high-dimensional reciprocating compressor feature space, x i 、x j and x k represents a data point in the high-dimensional reciprocating compressor feature space, σ i 2 Indicates x i The Gaussian variance centered at i Determined by binary search based on the perplexity factor.

[0134] Step 153: Determine the joint probability distribution in the low-dimensional space according to the t distribution.

[0135] Among them, the formula for the joint probability distribution in the low-dimensional space includes:

[0136]

[0137] In the formula, q ij represents the joint probability distribution in low-dimensional space, y i ,y j and k Represents a data point in the high-dimensional reciprocating compressor feature space.

[0138] Step 154: Minimize the KL divergence (Kullback-Leibler divergence) by the gradient descent method, and update the principal components of the thermal parameters and the vibration parameters in the low-dimensional space to obtain the fault characteristics of the reciprocating compressor.

[0139] Among them, the KL divergence is minimized by the gradient descent method, including:

[0140] Minimize the KL divergence based on the gradient formula, where the gradient formula includes:

[0141]

[0142] In the formula, C represents KL divergence, y i and j represents the data points in the low-dimensional space, p ij represents the joint probability distribution of data points i and j in the high-dimensional reciprocating compressor feature space, q ij Represents the joint probability distribution of data points i and j in the low-dimensional space.

[0143] The principal components of the thermodynamic parameters and the principal components of the vibration parameters in the low-dimensional space are updated by an updating formula, wherein the updating formula includes:

[0144]

[0145] Where Y represents the principal components of the thermal parameters and the vibration parameters in the low-dimensional space, t represents the number of iterations, η represents the learning rate, and α(t) is the momentum factor.

[0146] Pre-calculate the parameters in the update formula and use the normal distribution N(0,10 -4 I) Randomly initialize Y to obtain the initial solution Y (0) . Start iterative calculation of Y (t) , until the number of iterations reaches the maximum number of iterations, the principal components of the thermal parameters and the principal components of the vibration parameters in the updated low-dimensional space are obtained. The maximum number of iterations is set according to actual needs and is not specifically limited in this embodiment.

[0147] Fig.10 The schematic diagram shows the clustering effect of extracting features from the feature space of a high-dimensional fuel-driven reciprocating compressor based on t-distribution-random neighbor embedding according to an embodiment of the present application. Fig.10 As shown in the figure, based on t-distribution-random neighbor embedding, feature extraction is performed on the feature space of high-dimensional reciprocating compressor. The thermal parameters and vibration parameters under four states, namely, normal valve state, valve gap, valve plate fracture, and valve spring deficiency, are clearly distinguished and identified. The clustering effect is significantly better than the clustering effect of nonlinear dimensionality reduction based only on thermal parameters or only on vibration.

[0148] Optionally, an embodiment of the present application further provides a fault feature extraction device for a reciprocating compressor, the device comprising:

[0149] An acquisition module is used to acquire thermal parameters and vibration parameters of the reciprocating compressor, wherein the thermal parameters include cylinder pressure, fuel pressure, valve temperature, speed, torque and power, and the vibration parameters include root mean square, true peak, kurtosis, skewness, peak factor, margin factor, waveform factor, pulse factor in the time domain and center of gravity frequency, mean square frequency, frequency variance in the frequency domain;

[0150] The first construction module is used to construct a high-dimensional thermal feature space and a high-dimensional vibration feature space according to the thermal parameters and the vibration parameters respectively;

[0151] The first data processing module is used to perform nonlinear feature dimensionality reduction on the high-dimensional thermal feature space and the high-dimensional vibration feature space based on kernel principal component analysis to obtain the principal components of the thermal parameters and the principal components of the vibration parameters;

[0152] A second construction module is used to construct a high-dimensional reciprocating compressor feature space according to the principal components of the thermal parameters and the principal components of the vibration parameters;

[0153] The second data processing module is used to extract features from the high-dimensional reciprocating compressor feature space based on t-distribution-random neighbor embedding to obtain fault features of the reciprocating compressor.

[0154] Optionally, an embodiment of the present application further provides a fault diagnosis device for a reciprocating compressor, comprising:

[0155] a memory configured to store instructions;

[0156] The processor is configured to call instructions from the memory and implement the fault feature extraction method for the reciprocating compressor as described above when executing the instructions.

[0157] Optionally, an embodiment of the present application further provides a machine-readable storage medium, on which instructions are stored, and the instructions are used to enable a machine to execute the fault feature extraction method for the reciprocating compressor as described above.

[0158] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment or an embodiment in combination with software and hardware. Moreover, the present application may adopt 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.) that contain computer-usable program code.

[0159] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0160] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0161] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0162] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0163] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0164] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0165] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0166] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included within the scope of the claims of the present application.

Claims

1. A method for extracting fault features of a reciprocating compressor, characterized in that: The method comprises: Acquire the thermal parameters and vibration parameters of the reciprocating compressor, wherein the thermal parameters include cylinder pressure, fuel pressure, valve temperature, speed, torque and power, and the vibration parameters include root mean square, true peak, kurtosis, skewness, peak factor, margin factor, waveform factor, pulse factor in the time domain and center of gravity frequency, mean square frequency, frequency variance in the frequency domain; Constructing a high-dimensional thermal characteristic space and a high-dimensional vibration characteristic space according to the thermal parameters and the vibration parameters respectively; Based on kernel principal component analysis, nonlinear feature dimensionality reduction is performed on the high-dimensional thermal feature space and the high-dimensional vibration feature space respectively to obtain the principal components of the thermal parameters and the principal components of the vibration parameters; Constructing a high-dimensional reciprocating compressor feature space according to the principal components of the thermal parameters and the principal components of the vibration parameters; Based on t-distribution-random neighbor embedding, feature extraction is performed on the high-dimensional reciprocating compressor feature space to obtain the fault features of the reciprocating compressor.

2. The method according to claim 1, characterized in that The nonlinear feature dimensionality reduction is performed on the high-dimensional thermal feature space and the high-dimensional vibration feature space based on kernel principal component analysis to obtain the principal components of the thermal parameters and the principal components of the vibration parameters, respectively, including: According to the kernel function, determine the kernel matrix; Centralizing the kernel matrix to obtain a centralized kernel matrix; Determine the eigenvalues ​​and corresponding eigenvectors of the centralized kernel matrix, and arrange each of the eigenvectors according to the eigenvalues; Determine a cumulative contribution rate based on the contribution rate of each of the characteristic values; The characteristic vector whose cumulative contribution rate is greater than or equal to a preset cumulative contribution rate is determined as the principal component of the thermal parameter and the principal component of the vibration parameter.

3. The method according to claim 2, characterized in that The kernel function is a Gaussian radial basis kernel function.

4. The method according to claim 1, characterized in that: The method of extracting features from the high-dimensional reciprocating compressor feature space based on t-distribution-random neighbor embedding to obtain the fault features of the reciprocating compressor includes: Mapping the principal components of the thermal parameters and the vibration parameters in the high-dimensional reciprocating compressor feature space to a low-dimensional space to obtain the principal components of the thermal parameters and the vibration parameters in the low-dimensional space; Determining a joint probability distribution in the feature space of the high-dimensional reciprocating compressor according to a Gaussian distribution; Determining a joint probability distribution in the low-dimensional space according to a t distribution; The KL divergence is minimized by the gradient descent method, and the principal components of the thermal parameters and the vibration parameters in the low-dimensional space are updated to obtain the fault characteristics of the reciprocating compressor.

5. The method according to claim 4, characterized in that The formula for the joint probability distribution in the high-dimensional reciprocating compressor feature space includes: In the formula, p ij represents the joint probability distribution in the high-dimensional reciprocating compressor feature space, p j|i represents the probability of j occurring under given conditions i, p i|j represents the probability of i occurring under the given j condition, n represents the number in the high-dimensional reciprocating compressor feature space, x i 、x j and x k represents the data points in the high-dimensional reciprocating compressor feature space, σ i 2 Indicates x i The Gaussian variance centered at The formula for the joint probability distribution in the low-dimensional space includes: In the formula, q ij represents the joint probability distribution in the low-dimensional space, y i ,y j and k represents a data point in the high-dimensional reciprocating compressor feature space.

6. The method according to claim 4, characterized in that The method of minimizing the KL divergence by the gradient descent method includes: Minimize the KL divergence based on the gradient formula, where the gradient formula includes: Where C represents the KL divergence, y i and j represents the data points in the low-dimensional space, p ij represents the joint probability distribution of data points i and j in the high-dimensional reciprocating compressor feature space, q ij Represents the joint probability distribution of data points i and j in the low-dimensional space.

7. The method according to claim 4, characterized in that The principal components of the thermodynamic parameters and the principal components of the vibration parameters in the low-dimensional space are updated by an updating formula, wherein the updating formula includes: Wherein, Y represents the principal component of the thermal parameters and the principal component of the vibration parameters in the low-dimensional space, t represents the number of iterations, η represents the learning rate, and α(t) is the momentum factor.

8. A fault feature extraction device for a reciprocating compressor, characterized in that: The device comprises: an acquisition module, configured to acquire thermal parameters and vibration parameters of the reciprocating compressor, wherein the thermal parameters include cylinder pressure, fuel pressure, valve temperature, rotation speed, torque and power, and the vibration parameters include root mean square, true peak, kurtosis, skewness, peak factor, margin factor, waveform factor, pulse factor in the time domain and center of gravity frequency, mean square frequency and frequency variance in the frequency domain; A first construction module is used to construct a high-dimensional thermal feature space and a high-dimensional vibration feature space according to the thermal parameters and the vibration parameters respectively; A first data processing module is used to perform nonlinear feature dimensionality reduction on the high-dimensional thermal feature space and the high-dimensional vibration feature space based on kernel principal component analysis to obtain the principal components of the thermal parameters and the principal components of the vibration parameters; A second construction module is used to construct a high-dimensional reciprocating compressor feature space according to the principal components of the thermal parameters and the principal components of the vibration parameters; The second data processing module is used to extract features from the high-dimensional reciprocating compressor feature space based on t-distribution-random neighbor embedding to obtain fault features of the reciprocating compressor.

9. A fault diagnosis device for a reciprocating compressor, characterized in that: include: a memory configured to store instructions; A processor is configured to call the instructions from the memory and implement the fault feature extraction method for a reciprocating compressor according to any one of claims 1 to 7 when executing the instructions.

10. A machine-readable storage medium, characterized in that: The machine-readable storage medium stores instructions for causing a machine to execute the fault feature extraction method for a reciprocating compressor according to any one of claims 1 to 7.

Images