Industrial process oscillation detection method and system based on adaptive variational mode decomposition

Abstract

The invention discloses an industrial process oscillation detection method and system based on adaptive variational mode decomposition, and relates to the field of industrial process oscillation detection, and the method comprises the steps: collecting a group of process variable signals from an industrial process control system; traversing modal numbers in a preset range, and configuring the modal numbers to a variational modal decomposition algorithm; performing variational mode decomposition on the process variable signal by adopting a variational mode decomposition algorithm to obtain an intrinsic mode function corresponding to each mode number; according to the intrinsic mode function corresponding to each mode number, calculating mode indexes corresponding to different mode numbers; selecting a modal number corresponding to the minimum modal index and a corresponding intrinsic mode function as output; sparsity indexes of the intrinsic mode functions corresponding to all the minimum mode indexes are calculated; calculating robust regularity indexes of the intrinsic mode functions corresponding to all the minimum mode indexes; and judging whether oscillation exists or not according to the sparsity index and the robust regularity index. According to the invention, the industrial process oscillation detection precision is improved.

Description

Technical Field

[0001] The present application relates to the field of industrial process oscillation detection, and in particular to an industrial process oscillation detection method and system based on adaptive variational mode decomposition. Background Art

[0002] Industrial processes include process industries dominated by basic raw material industries such as petroleum, chemicals, steel, nonferrous metals, and building materials, as well as discrete industries dominated by equipment manufacturing industries such as machinery, electronics, automobiles, aerospace, rail transit, and marine engineering. These industries are not only important components of the manufacturing industry, but also pillar industries for my country's national economy and social development, playing an irreplaceable role in ensuring the construction of major national projects and promoting national economic growth. For example, the petroleum and chemical industries provide energy and chemical raw materials to support the production activities of various other industries; the steel and nonferrous metal industries provide key structural materials to support the development of construction and manufacturing; and the machinery and electronics industries provide the necessary equipment and technical support for the entire manufacturing industry. With the continuous advancement of science and technology and the rapid improvement of industrial automation levels, industrial processes are developing towards scale and integration, which puts forward higher requirements for safe, efficient, energy-saving and high-quality production.

[0003] As the "brain" of the industrial production process, the control system's main function is to ensure that the production process runs according to the predetermined program and achieves the expected production goals. It is an indispensable part of ensuring product quality, reducing the material and energy consumption of the industrial process, and ensuring the safe operation of the process. Specifically, the control system monitors various parameters in the production process (such as temperature, pressure, flow, liquid level, controller, etc.) in real time and automatically adjusts according to these parameters to maintain the stability and efficient operation of the process. The proportional-integral-differential (PID) controller has become the most commonly used controller in industrial process control (occupying more than 90% of the control loops) due to its simple structure, stable operation, strong robustness, and fast response speed.

[0004] However, as time goes by, the complex production process and the coupling characteristics of large equipment will change, which will cause the PID controller to have performance degradation problems. Honeywell has conducted performance analysis and statistics on more than 20,000 control loops in the process industry. The results show that about 50% of the closed loops have problems with control performance degradation, and 40% of the loops have abnormal phenomena. Control loops with poor performance will interfere with process operations for a long time, thus having a negative impact on corporate profits.

[0005] Oscillation is one of the main manifestations of PID controller performance deterioration and one of the most common problems in process industries. Oscillation usually manifests itself as the process variable (PV) repeatedly fluctuating around the setpoint (SP) and failing to stabilize at the desired setpoint. This instability not only reduces product quality, but also increases the consumption of raw materials and energy, accelerates equipment aging, and even affects the safety of the production process. Therefore, detecting oscillation and making timely maintenance decisions are crucial to maintaining the stability and economic benefits of the production process.

[0006] With the rapid development of signal processing technology, the focus of existing oscillation detection technology has shifted from time domain and frequency domain to time-frequency domain. Time-frequency domain analysis technology combines the advantages of time domain and frequency domain, and describes the characteristics of the signal more comprehensively by considering the changes of the signal in time and frequency at the same time. In practical applications, by pre-processing PV using advanced signal processing technology, more stable and meaningful data can be obtained. These pre-processing techniques include signal filtering, noise reduction, detrending, etc., which are designed to remove noise and interference in the signal and extract data features in PV. Subsequently, the extracted data features are analyzed based on a series of oscillation detection indicators to complete the detection of oscillation phenomena in a single control loop.

[0007] In signal time-frequency analysis technology, empirical methods represented by empirical mode decomposition (EMD) and optimization methods represented by variational mode decomposition (VMD) are the two mainstream technologies. Among them, EMD is more adaptive and can complete adaptive analysis of data without setting parameters, but it will have modal aliasing problems when processing industrial data under noise and intermittent pollution. In contrast, VMD has deeper theoretical technology, and its decomposition performance is better than EMD under the premise of appropriate parameter configuration (modal number K), but using inappropriate K will lead to more serious modal aliasing problems, which significantly increases the difficulty of oscillation detection. Therefore, determining the appropriate K is crucial to improving VMD performance and accurately completing oscillation detection. Summary of the invention

[0008] The purpose of this application is to provide an industrial process oscillation detection method and system based on adaptive variational modal decomposition. The improvement is to adaptively select the key parameters (modal number K) in the variational modal decomposition method, solve the modal aliasing caused by variational modal decomposition, and effectively improve the accuracy of industrial process oscillation detection. Secondly, the use of "decomposition + detection" means greatly reduces the difficulty of oscillation detection and effectively improves the success rate of oscillation detection.

[0009] To achieve the above objectives, this application provides the following solutions:

[0010] In a first aspect, the present application provides an industrial process oscillation detection method based on adaptive variational mode decomposition, comprising:

[0011] A set of process variable signals are collected from an industrial process control system;

[0012] Traversing the modal numbers of the variational modal decomposition algorithm within a preset range to obtain a set of modal numbers;

[0013] Based on the set of modal numbers, the variational modal decomposition algorithm is used to perform variational modal decomposition on the process variable signal to obtain an intrinsic modal function corresponding to each modal number;

[0014] According to the inherent mode function corresponding to each mode number, the modal index corresponding to different mode numbers is calculated;

[0015] Calculate the sparsity index of all intrinsic mode functions corresponding to the minimum mode index;

[0016] Calculate the robust regularity index of all intrinsic mode functions corresponding to the minimum modal index;

[0017] Whether there is oscillation is determined according to the sparsity index and the robust regularity index.

[0018] Optionally, the industrial process oscillation detection method based on adaptive variational mode decomposition further includes:

[0019] Determine the quadratic penalty factor, fidelity parameter and convergence condition parameters of the variational mode decomposition algorithm.

[0020] Optionally, the modal numbers of the variational modal decomposition algorithm are traversed within a preset range to obtain a set of modal numbers, specifically including:

[0021] Set the traversal range of the modal number of the variational modal decomposition algorithm to [a, b], and obtain the modal number set by traversing the range [a, b] according to the step size λ

[0022]

[0023] Among them, a, b, and λ are set according to the actual signal. The default settings are a=2, b=10, λ=1, and K i represents the i-th mode number.

[0024] Optionally, based on the set of modal numbers, the variational modal decomposition algorithm is used to perform variational modal decomposition on the process variable signal to obtain an intrinsic modal function corresponding to each modal number, specifically including:

[0025] The modal numbers in the modal number set are sequentially assigned to the variational modal decomposition algorithm to perform The sub-variational mode decomposition is performed to obtain the intrinsic mode function corresponding to each mode number; wherein the number of the intrinsic mode functions is consistent with the size of the mode number.

[0026] Optionally, the modal index MI corresponding to the i-th modal number is calculated using the following formula: i :

[0027]

[0028] Among them, d i (t) represents the i-th mode number K of the variational mode decomposition algorithm configuration i The intrinsic mode function obtained by decomposition is represents the i-th mode number K of the variational mode decomposition algorithm configuration i Decomposed K i Intrinsic mode functions, P[·] is the power spectral density operator, ψ{·} is the counting operator, and the power spectral density of the intrinsic mode function is calculated at the threshold Th = p x / 50The sum of the number of process variable signals mentioned above, p x is the average power of the process variable signal, and ||·||2 is the two-norm operator.

[0029] Optionally, the sparsity index of all intrinsic mode functions corresponding to the minimum mode index is calculated, specifically including:

[0030] Calculate the Fourier transformed spectra of all intrinsic modal functions corresponding to the minimum modal index, and calculate the sparsity index of each spectrum; the sparsity index SI of the spectrum of the jth intrinsic modal function is calculated using the following formula: j :

[0031]

[0032] Where N is the length of the vector, represents the spectrum of the jth intrinsic mode function, δ is the balance factor, which is set according to the actual signal and is set to 1 by default.

[0033] Optionally, the robust regularity index r of the jth intrinsic mode function corresponding to the minimum modal index is calculated using the following formula: j :

[0034]

[0035] in, represents the average time period of the autocorrelation function of the jth intrinsic mode function, t j,prepresents the pth zero crossing point in the autocorrelation function of the jth intrinsic mode function, t j,p-1 represents the p-1th zero-crossing point in the autocorrelation function of the jth intrinsic mode function, n represents the total number of zero-crossing points, Represents a set of time intervals The standard deviation of , τ is the oscillation coefficient, is calculated as where d i (t) represents the i-th mode number K of the variational mode decomposition algorithm configuration i The intrinsic mode functions obtained by decomposition, mean[·], max[·] and min[·] are the average operator, maximum operator and minimum operator respectively.

[0036] Optionally, judging whether oscillation exists according to the sparsity index and the robust regularity index specifically includes:

[0037] If there is any intrinsic mode function satisfying SI>0.5 and r>0.5, it is considered that there is an oscillation phenomenon in the process variable signal, otherwise there is no oscillation phenomenon; wherein SI is the sparsity index of the intrinsic mode function, and r is the robust regularity index of the intrinsic mode function.

[0038] In a second aspect, the present application provides an industrial process oscillation detection system based on adaptive variational mode decomposition, comprising:

[0039] An acquisition module, used for acquiring a set of process variable signals from an industrial process control system;

[0040] A configuration module is used to traverse the modal numbers of the variational modal decomposition algorithm within a preset range to obtain a set of modal numbers;

[0041] An output module, configured to perform variational modal decomposition on the process variable signal based on the set of modal numbers by using the variational modal decomposition algorithm to obtain an intrinsic modal function corresponding to each modal number;

[0042] A modal index calculation module is used to calculate the modal index corresponding to different modal numbers according to the inherent modal function corresponding to each modal number;

[0043] A sparsity index calculation module is used to calculate the sparsity index of all intrinsic mode functions corresponding to the minimum mode index;

[0044] A robust regularity index calculation module is used to calculate the robust regularity index of all inherent modal functions corresponding to the minimum modal index;

[0045] A judging module is used to judge whether there is oscillation according to the sparsity index and the robust regularity index.

[0046] Optionally, the data processing process in the configuration module includes:

[0047] Set the traversal range of the modal number of the variational modal decomposition algorithm to [a, b], and obtain the modal number set by traversing the range [a, b] according to the step size λ

[0048]

[0049] Among them, a, b, and λ are set according to the actual signal. The default settings are a=2, b=10, λ=1, and K i represents the i-th mode number.

[0050] According to the specific embodiments provided in this application, this application discloses the following technical effects:

[0051] The present application provides an industrial process oscillation detection method and system based on adaptive variational mode decomposition, which can detect oscillation phenomena in PV signals collected from industrial processes and determine whether the control system is in an abnormal state. Its key technology includes the following steps: First, a mode index (Mode index, MI) is designed to quantify the degree of modal aliasing in the intrinsic mode functions (Intrinsic mode functions, IMFs) decomposed by variational mode decomposition (Variational mode decomposition, VMD). Then, the optimal mode number K is adaptively corrected for different signals to be decomposed (actually collected PV signals) to complete VMD and obtain a series of IMFs. Specifically, the parameter K of VMD is traversed within a preset range to complete the decomposition and calculate the corresponding MI, and the K corresponding to the smallest MI is selected to output IMFs. Finally, the corresponding oscillation detection indicators of IMFs are calculated, including the sparseness index (Sparseness index, SI) and the robust regularity index (Regularity, r), to complete the oscillation detection. This application does not require prior knowledge of industrial processes, is highly robust to noise and intermittency in PV collected from actual process loops, and can automatically decompose PV signals, extract oscillating IMFs with actual physical significance, and achieve accurate oscillation detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0053] Figure 1This is an overall flow chart of an industrial process oscillation detection method based on adaptive variational mode decomposition in one embodiment of the present application;

[0054] Figure 2 A flow chart of a self-tuning variational mode decomposition (SVMD) provided in one embodiment of the present application;

[0055] Figure 3 This is the flow chart of the module for calculating oscillating indicators in this application;

[0056] Figure 4 It is the decomposition result diagram of VMD decomposition simulation signal;

[0057] Figure 5 Calculate the modal index MI corresponding to different K in SVMD;

[0058] Figure 6 The decomposition result diagram of the SVMD decomposition simulation signal;

[0059] Figure 7 Decomposition of real industrial PV results for SVMD;

[0060] Figure 8 This is a functional module diagram of an industrial process oscillation detection system based on adaptive variational mode decomposition in another embodiment of the present application. DETAILED DESCRIPTION

[0061] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0062] The signal time-frequency analysis technology includes an EMD-based oscillation detection algorithm and a VMD-based oscillation detection algorithm.

[0063] 1. Oscillation detection algorithm based on EMD.

[0064] The oscillation detection algorithm based on EMD includes the following steps: first, the PV data is decomposed using EMD to obtain a set of intrinsic mode functions (IMFs), then the relevant detection indicators (absolute error integral, period, regularity, etc.) are calculated, and finally, whether there is oscillation is determined based on the calculated indicators.

[0065] Advantages: The algorithm has a simple structure, is easy to implement, does not require prior knowledge, and has a high degree of adaptability.

[0066] Disadvantages: EMD has poor robustness against noise and intermittency in PV data and is prone to modal aliasing problems, which leads to generally low detection accuracy of oscillations.

[0067] 2. Oscillation detection algorithm based on VMD.

[0068] The VMD-based oscillation detection algorithm includes the following steps: first, the modal number K of VMD is set. For industrial data, K=2 is generally set. PV data is decomposed using VMD to obtain a set of IMFs; then relevant detection indicators (absolute error integral, period, regularity, etc.) are calculated. Finally, whether there is oscillation is determined based on the calculated indicators.

[0069] Advantages: VMD has a solid theoretical basis and is robust to noise and intermittency in the signal.

[0070] Disadvantages: The decomposition performance of VMD is highly dependent on the preset modal number K. An inappropriate K will lead to the failure of VMD construction and optimization problem solving, resulting in modal aliasing problems and affecting the accuracy of oscillation detection.

[0071] In view of this, the present application proposes an industrial process oscillation detection method and system based on adaptive variational modal decomposition, and its key technology includes the following steps: First, a modal index (Mode index, MI) is designed to quantify the degree of modal aliasing in the intrinsic mode functions IMFs decomposed by VMD. Then, the parameter K of VMD is traversed within a preset range to complete the decomposition and calculate the corresponding MI, and the K corresponding to the smallest MI is selected to output IMFs. Finally, the corresponding oscillation detection indicators of IMFs are calculated, including the sparseness index (Sparseness index, SI) and the robust regularity index (Regularity, r), to complete the oscillation detection.

[0072] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0073] The full Chinese names of the symbols involved in this application and their corresponding abbreviations are detailed in Table 1.

[0074] Table 1 Abbreviations

[0075] EMD Empirical mode decomposition IMFs Intrinsic mode functions MI Modeindex PID Proportional-Integral-Derivative PV Process variable r Robust regularity index (Regularity) SI Sparseness index SP Setpoint SVMD Self-tuning variational mode decomposition VMD Variational mode decomposition

[0076] The industrial process oscillation detection method based on adaptive variational mode decomposition provided in the embodiment of the present application is as follows: Figure 1 As shown, including:

[0077] (1) When the controller of an industrial process is in an abnormal condition (oscillation), its process variable PV will oscillate. Therefore, in order to detect the oscillation of the controller, the PV signal is first collected from the industrial process control system.

[0078] Step 1: Collect a set of process variable signals from the industrial process control system.

[0079] Disadvantage 1: VMD will have serious modal aliasing problems when an inappropriate mode number K is configured. Specifically, a too small K will cause components of different frequencies in the original signal to be aliased in the same intrinsic mode function IMF, while a too large K will cause components that should be decomposed in one IMF to be split into different IMFs.

[0080] To this end, the present application proposes a self-tuning variational mode decomposition (SVMD), which effectively suppresses the modal aliasing problem of VMD by automatically correcting the key parameter (K) of the variational mode decomposition VMD, making the extracted oscillation components more accurate. The embodiments are as follows:

[0081] (2) Use the proposed SVMD to decompose the collected PV signal to extract the oscillation components for subsequent oscillation detection. The first step of the proposed SVMD is to determine the optimal mode number k of VMD, which seriously affects the decomposition performance of VMD.

[0082] Step 2: Traverse the modal numbers of the variational modal decomposition algorithm within a preset range to obtain a set of modal numbers. This step specifically includes:

[0083] The traversal range of the modal number K of the VMD algorithm is set to [a, b], and the most suitable K for VMD is determined from the range. Among them, the modal number is traversed within the preset range and configured to the variational mode decomposition algorithm, specifically including:

[0084] Set the range of the modal number of the variational mode decomposition algorithm to [a, b], and obtain a selectable set of modal numbers by traversing the range [a, b] according to the step size λ

[0085]

[0086] Among them, a, b, and λ are set according to the actual signal. The default settings are a=2, b=10, λ=1, and K i represents the i-th mode number.

[0087] (3) To complete VMD decomposition, in addition to K, other VMD parameters need to be fixed.

[0088] Step 3: Determine the quadratic penalty factor α, fidelity parameter τ, and convergence condition parameter ε of the variational mode decomposition algorithm; where α, τ, and ε are set according to the actual signal, and their default settings are α = 1000, τ = 0, ε = 10 -7 .

[0089] Disadvantage 2: Accurately evaluating the degree of modal aliasing in the IMFs decomposed by VMD is the key to determining whether the modal number K of the current VMD configuration is appropriate, but it cannot be achieved with existing technologies.

[0090] To this end, the present application proposes MI, which accurately quantifies the degree of modal aliasing in IMFs by calculating the degree of power spectrum overlap and the bandwidth of the power spectrum of IMFs (by using the two-norm operator). The embodiments are as follows:

[0091] (4) After configuring the VMD parameters, complete the SVMD decomposition for the PV signal and select the optimal modal number K (the flowchart is shown in Figure 2 As shown), and the decomposed IMFs corresponding to the optimal K are output. Specifically, by calculating the MI index of the variational mode decomposition VMD corresponding to different K, and selecting the mode number K and intrinsic mode function IMFs corresponding to the minimum MI as the output, the steps are as follows:

[0092] Step 4: Based on the set of modal numbers, the variational modal decomposition algorithm is used to perform variational modal decomposition on the process variable signal to obtain the intrinsic modal function corresponding to each modal number; the modal index corresponding to different modal numbers is calculated according to the intrinsic modal function corresponding to each modal number; and the modal number corresponding to the smallest modal index and the corresponding intrinsic modal function are selected as output. This step specifically includes:

[0093] Step 4-1: Convert the modal numbers Configured to the variational mode decomposition algorithm, the process variable signal is The sub-variational mode decomposition is performed to obtain the corresponding intrinsic mode functions, where the number of intrinsic mode functions is consistent with the mode number.

[0094] Step 4-2: Use the following formula to calculate the modal index MI of the intrinsic modal function obtained by decomposing the strain corresponding to the i-th modal number using the modal decomposition algorithm: i :

[0095]

[0096] Among them, d i (t) represents the i-th mode number K of the variational mode decomposition algorithm configuration i The intrinsic mode function obtained by decomposition is represents the i-th mode number K of the variational mode decomposition algorithm configuration iDecomposed K i Intrinsic mode functions, P[·] is the power spectral density operator, ψ{·} is the counting operator, and the power spectral density of the intrinsic mode function is calculated at the threshold Th = p x / 50The sum of the number of process variable signals (sample points) mentioned above, where p x is the average power of the process variable signal, and ||·||2 is the two-norm operator.

[0097] Step 4-3, select the modal number K corresponding to the smallest modal index o and the corresponding intrinsic mode functions As output,

[0098] The modal index MI proposed in the embodiment of the present application is a key indicator for evaluating modal aliasing in the decomposition result, and accurately quantifies the modal aliasing degree in IMFs by calculating the power spectrum overlap degree of IMFs and the bandwidth of the power spectrum (by using the two-norm operator). This indicator ensures that the selected modal number K is optimal.

[0099] Disadvantage 3: Due to the modal aliasing problem of existing decomposition tools, the oscillation components cannot be accurately extracted, which will seriously affect the subsequent oscillation detection and may make the oscillation components unable to be accurately detected.

[0100] To this end, the present application completes oscillation detection by calculating the oscillation index of the intrinsic mode function IMFs decomposed using the proposed SVMD. This "decomposition + detection" method greatly reduces the difficulty of oscillation detection and effectively improves the success rate of oscillation detection. The embodiments are as follows:

[0101] (5) For the oscillating components decomposed by SVMD, that is, By calculating the oscillation detection index (SI and r) of each IMFs, it is determined whether there is an oscillation phenomenon in each IMF. If the SI and r indicators of any IMF meet the oscillation conditions, it is determined that there is an oscillation phenomenon in the PV, that is, the controller is in an abnormal condition. The specific flow chart is as follows Figure 3 As shown, the steps are as follows:

[0102] Step 5-1, calculate the sparsity index of all intrinsic mode functions corresponding to the minimum mode index.

[0103] Specifically, the sparsity indexes of all intrinsic mode functions corresponding to the minimum modal index are calculated, including:

[0104] Calculate K o The spectrum of the intrinsic mode function after Fourier transformation is calculated, and the sparsity index of the spectrum is calculated Among them, the j-th SI is calculated using the following formula: j for:

[0105]

[0106] Where N is the length of the vector, represents the spectrum of the jth intrinsic mode function. δ is the balance factor, which can be set according to the actual signal and is set to 1 by default.

[0107] Step 5-2, calculate the robust regularity index of all inherent mode functions corresponding to the minimum modal index.

[0108] Specifically, the robust regularity indexes of all intrinsic mode functions corresponding to the minimum modal index are calculated, including:

[0109] Calculate K o Robust regularity index of intrinsic mode functions Among them, the following formula is used to calculate the jth r j for:

[0110]

[0111] in, represents the average time period of the autocorrelation function of the jth intrinsic mode function and is calculated using the following formula for:

[0112]

[0113] Among them, t j,p represents the pth zero crossing point in the autocorrelation function of the jth intrinsic mode function, t j,p-1 represents the p-1th zero-crossing point in the autocorrelation function of the jth intrinsic mode function, n represents the total number of zero-crossing points, Represents a set of time intervals The standard deviation of , τ is the oscillation coefficient, is calculated as Where mean[·], max[·] and min[·] are the average operator, maximum operator and minimum operator respectively.

[0114] Step 5-3: determine whether oscillation exists according to the sparsity index and the robust regularity index.

[0115] Specifically, judging whether there is oscillation according to the sparsity index and the robust regularity index specifically includes:

[0116] If there is any intrinsic mode function satisfying SI>0.5 and r>0.5, it is considered that there is an oscillation phenomenon in the process variable signal, otherwise there is no oscillation phenomenon; wherein SI is the sparsity index of the intrinsic mode function, and r is the robust regularity index of the intrinsic mode function.

[0117] The embodiment of the present application calculates the oscillation index of the IMFs decomposed using the proposed SVMD to complete the oscillation detection. This "decomposition + detection" approach greatly reduces the difficulty of oscillation detection and effectively improves the success rate of oscillation detection.

[0118] The overall workflow of the industrial process oscillation detection method based on adaptive variational mode decomposition in the above embodiment is as follows:

[0119] (1) A set of PV signals is collected from an industrial process control system.

[0120] (2) Set the traversal range of K.

[0121] (3) Fix other VMD parameters.

[0122] (4) Complete SVMD.

[0123] (5) Output a set of intrinsic mode functions IMFs.

[0124] (6) Calculate the oscillation index (sparseness index and robust regularity index) and output the oscillation detection result. In order to verify the effectiveness of this application, first implement the above step (1) and analyze a simulation signal:

[0125] x(t)=sin(2π×1×t)+cos(2π×4×t)+0.2×w(t).

[0126] Where w(t) is a Gaussian white noise signal w(t)~N(0,1), data length is 712, sampling frequency is 100Hz. Use conventional VMD to decompose the simulation signal, configure the mode number K=2 (two oscillation components in the simulation data), the decomposition result is as follows Figure 4 As shown. Among them, VMD can be performed according to the prior art "Dragomiretskiy K, Zosso D. Variational mode decomposition [J]. IEEE transactions on signal processing, 2013, 62 (3): 531-544."

[0127] Implement the above steps (2) to (4) to complete SVMD. First, set the traversal range of K to [2, 10]. With a step size of 1, traversing K can obtain the set of modal numbers in advance. The above K configurations are respectively assigned to VMD to complete the decomposition and calculate the corresponding MI, such as Figure 5 shown.

[0128] Implement the above step (5) and select the K (K=4) and corresponding IMFs corresponding to the minimum MI as output, where the IMFs corresponding to K=4 are as follows: Figure 6 As shown. Figure 6 It can be seen that SVMD has a better performance in extracting oscillation signals (two sinusoidal signals). There is no frequency aliasing between the modes, and the characteristics of the oscillation signals are successfully extracted. Figure 4 In the decomposition result of the VMD with conventional parameters, there is a serious modal aliasing phenomenon, that is, both sinusoidal signals are in the first IMF. Therefore, the invented method can effectively adjust the modal number of VMD and suppress the modal aliasing problem.

[0129] Implement the above step (6) and calculate the corresponding oscillation indicators (sparseness index SI and robust regularity index r) according to the IMFs obtained by SVMD decomposition, as shown in Table 2. It can be observed from the table that the indicators of the first IMF and the second IMF are consistent with oscillation, that is, SI>0.5 and r>0.5. The proposed oscillation detection system successfully identifies the oscillation phenomenon in the simulation case.

[0130] Table 2 Oscillation detection results of the system of this application for simulation signals

[0131] IMF SI r Whether to oscillate <![CDATA[d1(t)]]> 0.9278 25 yes <![CDATA[d2(t)]]> 0.8063 7.98 yes <![CDATA[d3(t)]]> 0.5549 0 no <![CDATA[d4(t)]]> 0.5878 0.4921 no

[0132] The implementation case of this application uses the process variable PV of Chem.loop14 provided in the industrial data set isdb for analysis and processing. The isdb data set references "M.Jelali and B.Huang, Detection and diagnosis of stiction in control loops: state of the art and advanced methods, 2009.". The PV data of Chem.loop14 is collected from the flow control loop of a chemical plant, with sampling every 20 seconds, and a total of 1500 sample points are collected. In addition, there is stickiness in the valve in the Chem.loop14 loop, which causes oscillation in the process variable PV of the loop, such as Figure 7 As shown in PV.

[0133] Using SVMD to decompose actual industrial signals, the decomposition results (various modal functions) and the corresponding oscillation indicators (SI and r) are as follows: Figure 7 As shown in Table 3. First, from Figure 7It is found that SVMD extracts the oscillation features in PV data in the second IMF, while the trend, interference, and noise in PV are extracted in the first IMF. Effective and accurate extraction of oscillation features in PV data is crucial to achieve accurate oscillation detection. In addition, according to the oscillation indicators of IMFs provided in Table 3, it can be found that the oscillation indicators of the first IMF containing interference do not meet the oscillation conditions, while the SI of the second IMF is greater than 0.5 and r is greater than 0.5, so the second IMF is identified as an oscillation.

[0134] Table 3 Oscillation detection results of the system of this application for actual signals

[0135] IMF SI r Whether to oscillate <![CDATA[d1(t)]]> 0.5378 0.4654 no <![CDATA[d2(t)]]> 0.8063 3.5423 yes

[0136] Through experiments on simulated signals and actual signals, it can be seen that the industrial process oscillation detection method and system based on adaptive variational modal decomposition provided by this application can automatically correct the optimal modal number K of variational modal decomposition, effectively alleviate the modal aliasing problem, improve the decomposition performance of the method, and enable the oscillation components in PV to be effectively extracted. The method proposed in this application has the characteristics of excellent decomposition performance, no prior knowledge of industrial processes, and a solid theoretical foundation, and can better realize oscillation detection in process control systems.

[0137] Based on the same inventive concept, the embodiment of the present application also provides an industrial process oscillation detection system based on adaptive variational modal decomposition for implementing the industrial process oscillation detection method based on adaptive variational modal decomposition involved above. The implementation scheme for solving the problem provided by the system is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more embodiments of the industrial process oscillation detection system based on adaptive variational modal decomposition provided below can refer to the limitations of the industrial process oscillation detection method based on adaptive variational modal decomposition above, and will not be repeated here.

[0138] In another exemplary embodiment of the present application, Figure 8 As shown, an industrial process oscillation detection system based on adaptive variational mode decomposition is provided, comprising:

[0139] The acquisition module is used to acquire a set of process variable signals from the industrial process control system.

[0140] The configuration module is used to traverse the number of modes within a preset range and configure it to the variational mode decomposition algorithm.

[0141] The output module is used to perform variational modal decomposition on the process variable signal using the variational modal decomposition algorithm to obtain the intrinsic modal function corresponding to each modal number.

[0142] The modal index calculation module is used to calculate the modal index corresponding to different modal numbers according to the inherent modal function corresponding to each modal number.

[0143] The sparsity index calculation module is used to calculate the sparsity index of all intrinsic mode functions corresponding to the minimum mode index.

[0144] The robust regularity index calculation module is used to calculate the robust regularity index of all inherent modal functions corresponding to the minimum modal index.

[0145] A judging module is used to judge whether there is oscillation according to the sparsity index and the robust regularity index.

[0146] In an exemplary embodiment, the industrial process oscillation detection system based on adaptive variational mode decomposition further includes:

[0147] The parameter determination module is used to determine the quadratic penalty factor α, the fidelity parameter τ, and the convergence condition parameter ε of the variational mode decomposition algorithm; wherein α, τ, and ε are set according to the actual signal, and the default settings are α=1000, τ=0, ε=10 -7 .

[0148] In an exemplary embodiment, the data processing process in the configuration module includes:

[0149] Set the range of the modal number of the variational mode decomposition algorithm to [a, b], and obtain a selectable set of modal numbers by traversing the range [a, b] according to the step size λ

[0150]

[0151] Among them, a, b, and λ are set according to the actual signal. The default settings are a=2, b=10, λ=1, and K i represents the i-th mode number.

[0152] This application has the following advantages:

[0153] 1. The proposed MI is a key indicator for evaluating modal aliasing in the decomposition results. It accurately quantifies the degree of modal aliasing in IMFs by calculating the degree of power spectrum overlap of IMFs and the bandwidth of the power spectrum (by utilizing the two-norm operator).

[0154] 2. Based on the proposed MI, SVMD is proposed. This method effectively suppresses the modal aliasing problem of VMD by automatically correcting the key parameters of VMD (modal number K), making the extracted oscillation components more accurate.

[0155] 3. The robust regularity index is improved to make it more robust, and the oscillation detection is completed by calculating the oscillation index of the IMFs decomposed by the proposed SVMD. This "decomposition + detection" method greatly reduces the difficulty of oscillation detection and effectively improves the success rate of oscillation detection.

[0156] The above content is a further detailed description of the present application in combination with specific preferred implementations. The specific implementations of the present application cannot be limited to these. The present application can be applied to any field that requires data processing. Without departing from the concept of the present application, several simple deductions or substitutions can be made, which should be regarded as the scope of patent protection determined by the claims submitted by the present application.

[0157] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0158] This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and core ideas of this application. At the same time, for those skilled in the art, according to the ideas of this application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. An industrial process oscillation detection method based on adaptive variational mode decomposition, characterized in that: The industrial process oscillation detection method based on adaptive variational mode decomposition includes: A set of process variable signals are collected from an industrial process control system; Traversing the modal numbers of the variational modal decomposition algorithm within a preset range to obtain a set of modal numbers; Based on the set of modal numbers, the variational modal decomposition algorithm is used to perform variational modal decomposition on the process variable signal to obtain an intrinsic modal function corresponding to each modal number; According to the inherent mode function corresponding to each mode number, the modal index corresponding to different mode numbers is calculated; Calculate the sparsity index of all intrinsic mode functions corresponding to the minimum mode index; Calculate the robust regularity index of all intrinsic mode functions corresponding to the minimum modal index; Whether there is oscillation is determined according to the sparsity index and the robust regularity index.

2. The industrial process oscillation detection method based on adaptive variational mode decomposition according to claim 1 is characterized in that: The industrial process oscillation detection method based on adaptive variational mode decomposition also includes: Determine the quadratic penalty factor, fidelity parameter and convergence condition parameters of the variational mode decomposition algorithm.

3. The industrial process oscillation detection method based on adaptive variational mode decomposition according to claim 1 is characterized in that: Traverse the modal numbers of the variational modal decomposition algorithm within a preset range to obtain a set of modal numbers, including: Set the traversal range of the modal number of the variational modal decomposition algorithm to [a, b], and obtain the modal number set by traversing the range [a, b] according to the step size λ Among them, a, b, and λ are set according to the actual signal. The default settings are a=2, b=10, λ=1, and K i represents the i-th mode number.

4. The industrial process oscillation detection method based on adaptive variational mode decomposition according to claim 3 is characterized in that: Based on the set of modal numbers, the variational modal decomposition algorithm is used to perform variational modal decomposition on the process variable signal to obtain the intrinsic modal function corresponding to each modal number, specifically including: The modal numbers in the modal number set are sequentially assigned to the variational modal decomposition algorithm to perform The sub-variational mode decomposition is performed to obtain the intrinsic mode function corresponding to each mode number; wherein the number of the intrinsic mode functions is consistent with the size of the mode number.

5. The industrial process oscillation detection method based on adaptive variational mode decomposition according to claim 1, characterized in that: The modal index MI corresponding to the i-th modal number is calculated using the following formula: i : Among them, d i (t) represents the i-th mode number K of the variational mode decomposition algorithm configuration i The intrinsic mode function obtained by decomposition is represents the i-th mode number K of the variational mode decomposition algorithm configuration i Decomposed K i Intrinsic mode functions, P[·] is the power spectral density operator, ψ{·} is the counting operator, and the power spectral density of the intrinsic mode function is calculated at the threshold Th = p x / 50The sum of the number of process variable signals mentioned above, p x is the average power of the process variable signal, and ||·||2 is the two-norm operator.

6. The industrial process oscillation detection method based on adaptive variational mode decomposition according to claim 1, characterized in that: Calculate the sparsity index of all intrinsic mode functions corresponding to the minimum modal index, including: Calculate the Fourier transformed spectra of all intrinsic modal functions corresponding to the minimum modal index, and calculate the sparsity index of each spectrum; the sparsity index SI of the spectrum of the jth intrinsic modal function is calculated using the following formula: j : Where N is the length of the vector, represents the spectrum of the jth intrinsic mode function, δ is the balance factor, which is set according to the actual signal and is set to 1 by default.

7. According to the industrial process oscillation detection method based on adaptive variational modal decomposition according to claim 1, the robust regularity index r of the jth intrinsic modal function corresponding to the minimum modal index is calculated using the following formula: j : in, represents the average time period of the autocorrelation function of the jth intrinsic mode function, t j,p represents the pth zero crossing point in the autocorrelation function of the jth intrinsic mode function, t j,p-1 represents the p-1th zero-crossing point in the autocorrelation function of the jth intrinsic mode function, n represents the total number of zero-crossing points, Represents a set of time intervals The standard deviation of , τ is the oscillation coefficient, is calculated as where d i (t) represents the i-th mode number K of the variational mode decomposition algorithm configuration i The intrinsic mode functions obtained by decomposition, mean[·], max[·] and min[·] are the average operator, maximum operator and minimum operator respectively.

8. The industrial process oscillation detection method based on adaptive variational mode decomposition according to claim 1, judging whether oscillation exists according to the sparsity index and the robust regularity index, specifically comprises: If there is any intrinsic mode function satisfying SI>0.5 and r>0.5, it is considered that there is an oscillation phenomenon in the process variable signal, otherwise there is no oscillation phenomenon; wherein SI is the sparsity index of the intrinsic mode function, and r is the robust regularity index of the intrinsic mode function.

9. An industrial process oscillation detection system based on adaptive variational mode decomposition, characterized in that: The industrial process oscillation detection system based on adaptive variational mode decomposition includes: An acquisition module, used for acquiring a set of process variable signals from an industrial process control system; A configuration module is used to traverse the modal numbers of the variational modal decomposition algorithm within a preset range to obtain a set of modal numbers; An output module, configured to perform variational modal decomposition on the process variable signal based on the set of modal numbers by using the variational modal decomposition algorithm to obtain an intrinsic modal function corresponding to each modal number; A modal index calculation module is used to calculate the modal index corresponding to different modal numbers according to the inherent modal function corresponding to each modal number; A sparsity index calculation module is used to calculate the sparsity index of all intrinsic mode functions corresponding to the minimum mode index; A robust regularity index calculation module is used to calculate the robust regularity index of all inherent modal functions corresponding to the minimum modal index; A judging module is used to judge whether there is oscillation according to the sparsity index and the robust regularity index.

10. The industrial process oscillation detection system based on adaptive variational mode decomposition according to claim 9, characterized in that: The data processing process in the configuration module includes: Set the traversal range of the modal number of the variational modal decomposition algorithm to [a, b], and obtain the modal number set by traversing the range [a, b] according to the step size λ Among them, a, b, and λ are set according to the actual signal. The default settings are a=2, b=10, λ=1, and K i represents the i-th mode number.

Images