Oscillation detection method for feedback control loop of process industrial process control system

Abstract

The invention relates to the technical field of flow industry, in particular to a flow industry process control system feedback control loop oscillation detection method, which comprises the following steps: acquiring a data time sequence of a set value, a measured value and a valve position value of a control loop, performing data preprocessing on the data time sequence of the set value, the measured value and the valve position value of the control loop to obtain first data; according to the first data, determining a time domain fluctuation characteristic and a frequency characteristic characteristic of the preprocessed measurement value data time sequence; inputting the time domain fluctuation characteristics and the frequency characteristic characteristics into a preset oscillation degree obtaining algorithm to obtain an oscillation degree result of the control loop; the oscillation degree acquisition algorithm is set based on a fuzzy adaptive rule. The problems that a traditional control loop is low in oscillation positioning and analysis efficiency, high in use threshold and multiple in limitation are solved, and technical method support is provided for efficient operation and maintenance of the production device control loop and reduction of labor intensity of operators.

Description

Technical Field

[0001] The present invention relates to the technical field of process industry, and in particular to an oscillation detection method for a feedback control loop of a process industry process control system. Background Art

[0002] Process industry control is widely present in all areas of human life and production. Rapid and accurate control of process industry processes has extremely important practical value and economic benefits. In process industry production equipment, proportional, integral, and differential control (PID control for short) strategy is the most widely used, and more than 90% of automatic control loops use PID control strategy.

[0003] Due to the influence of internal and external factors such as process, equipment, production load and instrumentation, up to 60% of the control loops in the device have performance defects to varying degrees, especially control loop oscillation caused by various reasons. The workload of improving the performance of these loops is huge, but the number of system maintenance engineers in the enterprise is limited, and quantitative positioning and qualitative analysis of oscillation loops are difficult.

[0004] Existing oscillation detection and analysis methods are often based on professional technicians observing the control loop's measured values, set values, valve position values ​​and other curves with their naked eyes, and then making subjective judgments based on industrial knowledge and device characteristics. However, there are many on-site control loops, and the number of professional technicians in enterprises is very limited. Such methods are inefficient and cannot meet the loop operation and maintenance needs of large and medium-sized devices in the process industry.

[0005] In addition, the existing ones are mainly used to detect discontinuous oscillations of signals. Most of these methods are based on the analysis of data from stationary processes, mainly using time domain analysis. These methods are often only applicable to stationary processes. For non-stationary or multi-oscillation industrial processes, a lot of human intervention is required for separation, and they are easily interfered by medium-frequency noise. In addition, these evaluation methods often require a lot of prior knowledge of the process for comparison, lack a unified qualitative analysis method, and the given parameters of the calculation process require a lot of prior experience input, which has a high threshold and is difficult to promote widely.

[0006] Therefore, there is an urgent need for a feedback control loop oscillation detection method for process industry process control systems. Summary of the invention

[0007] 1. Technical issues to be resolved In view of the above-mentioned shortcomings and deficiencies of the prior art, the present invention provides a method for detecting oscillation of a feedback control loop in a process industrial control system, which gets rid of the technical problems of low accuracy, multiple usage bottlenecks, heavy reliance on analysis by professional technicians, high technical threshold, low automation level, and high operating intensity for operators in the existing methods for locating and analyzing oscillation defects in feedback control loops.

[0008] (II) Technical solution In order to achieve the above object, the main technical solutions adopted by the present invention include: An embodiment of the present invention provides a method for detecting oscillation of a feedback control loop of a process control system in a process industry, comprising: S100, obtaining a data time series of a set value, a measured value, and a valve position value of a control loop, and performing data preprocessing on the data time series of the set value, the measured value, and the valve position value of the control loop to obtain first data; The first data includes a preprocessed set value data time series, a preprocessed measured value data time series, and a preprocessed valve position value data time series; S200, determining the time domain fluctuation characteristics and frequency characteristic characteristics of the preprocessed measurement value data time series according to the first data; S300, inputting the time domain fluctuation characteristics and the frequency characteristic characteristics into a pre-set oscillation degree acquisition algorithm to obtain an oscillation degree result of the control loop; The oscillation degree acquisition algorithm is based on fuzzy adaptive rule setting.

[0009] Optionally, the S100 includes: S110, obtaining a fluctuating set point and a corresponding fluctuating position according to the set value data time series, deleting the fluctuating set point, and obtaining a preprocessed set value data time series; S120, sorting the data time series of the measurement values ​​and the data time series of the valve position values, removing the largest 0.5% and the smallest 0.5% data segments, and obtaining the preprocessed data time series of the measurement values ​​and the data time series of the valve position values.

[0010] Optionally, the S200 includes: S210, input the preprocessed measurement data time series into the following formula to obtain the time domain fluctuation characteristics: ; in, Std - pv is the standard deviation of the preprocessed measurement data time series, corresponding to the time domain fluctuation characteristics of the preprocessed measurement data time series,n is the data length, pv i is the time series of preprocessed measurement data, is the average value of the preprocessed measurement data time series.

[0011] Optionally, the S200 further includes: S220, performing a fast Fourier transform on the preprocessed measurement value data time series based on a time extraction DIT radix 2 FFT algorithm to obtain a frequency characteristic sequence of the preprocessed measurement value data time series; S230 , quantitatively analyzing the frequency characteristic sequence of the preprocessed measurement value data time series to obtain frequency characteristic features of the preprocessed measurement value data time series.

[0012] Optionally, the S220 specifically includes: The preprocessed measured value data time series is converted into a complex number sequence, and the complex number sequence is input into the following formula to obtain a fast Fourier transform result, that is, a frequency characteristic sequence of the preprocessed measured value data time series: ; in, Z 0 is the value of the complex sequence Z(n) when n=0, that is, the first complex value of the measured value data time series after preprocessing, j is an imaginary unit, ( Z 0 ) * for Z 0 The complex conjugate of F N / 2 It is the frequency component of point N in the fast Fourier transform result that is located in the middle position of the FFT result, corresponding to the characteristic sequence of the preprocessed measurement value data time series at the N / 2 frequency.

[0013] Optionally, converting the preprocessed measurement value data time series into a complex number sequence includes: The preprocessed measurement value data time series is separated into an odd part and an even part according to the parity, the odd part is used as the real part of the complex number, and the even part is used as the imaginary part of the complex number to obtain a complex number sequence.

[0014] Optionally, the S230 includes: S231, performing amplitude normalization processing on the frequency characteristic sequence of the preprocessed measurement value data time series, and extracting the DC component and the signal amplitude other than the DC component to obtain the low- and medium-frequency band signals and the high-frequency band signals; S232, obtaining the peak value of the mid-low frequency band and the standard deviation of the amplitude of the high frequency band according to the mid-low frequency band signal and the high frequency band signal; S233: Obtain frequency characteristic features of the preprocessed measurement value data time series according to the peak values ​​of the medium and low frequency bands and the standard deviation of the amplitudes of the high frequency bands.

[0015] Optionally, the S233 specifically includes: The standard deviation of the peak value of the medium and low frequency bands and the amplitude of the high frequency band is input into the following formula to obtain the frequency characteristic characteristics of the preprocessed measurement value data time series: ; in, Index - Osc is the oscillation coefficient of the control loop frequency characteristic, corresponding to the frequency characteristic characteristics of the preprocessed measurement value data time series, is the oscillation conversion coefficient, PV range is the range of the control loop measurement value, e is the interpolation between the set value and the measured value, Amp - max is the peak amplitude in the low and medium frequency bands, is the standard deviation of the amplitude in the high frequency band.

[0016] Optionally, the S300 includes: The time domain fluctuation characteristics and the frequency characteristic characteristics are input into the following formula to obtain the oscillation degree result of the control loop: ; in, Y is the result of the oscillation degree of the control loop, PV range is the range of the control loop measurement value, Index - Osc is the frequency characteristic of the time series of the preprocessed measurement data, Std - pv is the time domain fluctuation characteristic of the preprocessed measurement data time series, k 1 and k 2 is the corresponding weight, and k 1 and k 2 The sum of is 1.

[0017] Optionally, also include: Obtain the relative excitation size and response time of the control loop for oscillation judgment, input them into the fuzzy knowledge base, obtain the oscillation conversion coefficient and the weights corresponding to the frequency characteristic characteristics and time domain fluctuation characteristics of the preprocessed measurement value data time series; The fuzzy knowledge base is generated through the judgment of experienced engineers on the fusion of indicators.

[0018] Optional, The fuzzy knowledge base uses 7 linguistic variables for fuzzy description, and the fuzzy domain is divided into 7 levels, each level represents a fuzzy subset, and each fuzzy subset is represented by a corresponding linguistic variable; The language variables include: negative large, negative medium, negative small, zero, positive small, positive medium and positive large.

[0019] (III) Beneficial effects The beneficial effects of the present invention are as follows: a feedback control loop oscillation detection method for a process industry process control system of the present invention can make full use of the data characteristics of the control loop in the time domain and frequency domain, adopt the fast Fourier transform method and data standard deviation operation for qualitative calculation and analysis, and use fuzzy mathematics technology combined with the prior knowledge and experience of different loops in the process industry to quantify and fuse the analysis results in the time domain and frequency domain. Compared with the existing technology, it can make full use of the understanding and judgment of a large number of professionals on process knowledge, effectively make up for the problems of difficult parameter design and inconsistent evaluation standards of existing schemes, and can effectively improve the accuracy and efficiency of feedback control loop oscillation detection, and quickly improve the automation level of production equipment in the process industry. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 A schematic flow chart of an oscillation detection method for a feedback control loop of a process control system in a process industry according to an embodiment of the present invention; Figure 2 It is a signal flow diagram of the 8-point time extraction DIT radix 2 FFT algorithm time extraction algorithm in Embodiment 2 of the present invention; Figure 3 is a flow chart of the fuzzy adaptive process in embodiment 1 of the present invention; Figure 4 is the membership function of the relative excitation size of the data to be identified in embodiment 1 of the present invention; Figure 5 is the membership function of the response time of the object to be identified in Embodiment 1 of the present invention; Figure 6 is the weight value membership function of the relative excitation size of the data to be identified in embodiment 1 of the present invention; Figure 7 is the fuzzy rule table of weight values ​​of the accuracy loss function in Embodiment 1 of the present invention; Figure 8 This is the expected accuracy threshold fuzzy rule table in Example 1 of the present invention. DETAILED DESCRIPTION

[0021] In order to better explain the present invention and facilitate understanding, the present invention is described in detail below through specific implementation modes in conjunction with the accompanying drawings.

[0022] Loop oscillation detection: Loop oscillation refers to regular or irregular fluctuations in loop measurement values ​​caused by poor PID parameters, equipment aging, upstream and downstream disturbances, etc. It is a common problem in control loops. This oscillation can cause production fluctuations and affect product quality at the least, or endanger production safety at the worst, requiring continuous attention from enterprise operation and maintenance personnel.

[0023] The embodiment of the present invention proposes a method for detecting oscillation of a feedback control loop in a process control system of a process industry. It adopts a method combining time domain and frequency domain, and uses fuzzy adaptive rules combined with expert knowledge to design index fusion rules and realize parameter adaptation, thereby solving the problem of oscillation detection of feedback control loops in process control of process industries. Its main advantages are: The new oscillation detection method adopts frequency domain indicators based on traditional time domain detection. Compared with the single time domain evaluation method, it can be applied to industrial processes with non-stationary or multiple oscillation cycles. The new detection method covers the oscillation characteristics in the frequency domain and time domain, combines the advantages of the two indicators, and designs a detection formula to achieve quantitative calculation, thereby improving the accuracy and consistency of detection, making it easy to implement in product form, and greatly improving the efficiency of oscillation detection. The fast Fourier transform method is introduced in the qualitative calculation of the frequency domain, and a quantitative evaluation strategy is designed to realize the oscillation detection calculation in the frequency domain, which effectively eliminates the interference of high-frequency noise and the adverse effects of low-frequency disturbance on the results that are prone to occur in the frequency domain calculation process; The frequency domain and time domain indicators are weighted and fused, and fuzzy adaptive rules are used to achieve weight adaptive calculation for circuits with different characteristics. This effectively solves the problem that in the traditional method, parameters need to be given by manual experience during the consistency calculation process, which has a high technical threshold and easily leads to inaccurate and inconsistent evaluation results.

[0024] In order to better understand the above technical solution, exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present invention are shown in the accompanying drawings, it should be understood that the present invention can be implemented in various forms and should not be limited by the embodiments described herein. On the contrary, these embodiments are provided to enable a clearer and more thorough understanding of the present invention and to fully convey the scope of the present invention to those skilled in the art.

[0025] Example 1

[0026] See also Figure 1 , a method for detecting oscillation of a feedback control loop of a process control system in a process industry of this embodiment includes: Step S100, obtaining a data time series of a set value, a measured value, and a valve position value of a control loop, and performing data preprocessing on the data time series of the set value, the measured value, and the valve position value of the control loop to obtain first data; The first data includes a preprocessed set value data time series, a preprocessed measured value data time series, and a preprocessed valve position value data time series; Step S200: determining the time domain fluctuation characteristics and frequency characteristic characteristics of the preprocessed measurement value data time series according to the first data; Step S300, inputting the time domain fluctuation characteristics and the frequency characteristic characteristics into the oscillation degree acquisition algorithm set in advance to obtain the oscillation degree result of the control loop; The oscillation degree acquisition algorithm is based on fuzzy adaptive rule setting.

[0027] In step S300, in order to use unified indicators to analyze the oscillation of the feedback control loop of the process industry, an indicator fusion method based on fuzzy adaptive rules is proposed to achieve the fusion and unified calculation of parameters of different dimensions, so as to facilitate operation and maintenance personnel to conduct qualitative evaluation and quantitative analysis of the oscillation of loops of different types and characteristics based on a unified evaluation system and parameters under the same dimensional dimension.

[0028] Assume that in the final indicator, the weight of the time domain fluctuation feature is k 1 , the weight of the frequency characteristic feature is k 2 , then the oscillation degree result Y after fusion can be expressed as: ; in, Y is the result of the oscillation degree of the control loop, PV range is the range of the control loop measurement value, Index - Osc is the frequency characteristic of the time series of the preprocessed measurement data, Std - pv is the time domain fluctuation characteristic of the preprocessed measurement data time series, k 1 and k 2 is the corresponding weight, and k 1 andk 2 The sum of is 1.

[0029] Through generative fuzzy rules, the judgment method of experienced engineers on indicator fusion is generated into a fuzzy knowledge base, and fuzzy reasoning is performed using a fuzzy controller to obtain the final weight value, thereby lowering the technical threshold for indicator evaluation.

[0030] Through analysis, the factors affecting data quality are extracted and the input of the fuzzy knowledge base is summarized as the relative excitation size of the control loop to be oscillated. X e and the response time of the object to be identified ts , the fuzzy output is the weight value k 1 , oscillation conversion coefficient .

[0031] in: 0< X e <100%, 2s< ts <1000s; 0< k 1 <100%,0%< <100%; Among them, the relative incentive size X e Indicates the degree of change of the control loop set value, expressed as the ratio of the set value fluctuation to the set value range. The specific calculation is as follows: ; in, SP is the set value, SP range is the range of the set value, Std - sp is the standard deviation of the set value, which is used to characterize the fluctuation of the set value itself: ; in, SP is the set value, n is the data length, is the average value of the set value.

[0032] The response time calculation process of the object to be judged is as follows: According to the measured value and valve position value sequence of the circuit to be judged, the valve position value is considered to be the input of the circuit, and the measured value is the output of the circuit. The input-output time function of the circuit can be obtained by using the system identification method. The input-output time function of the object to be judged is expressed in the form of transfer function G(s) as follows: ; in, K is the gain, which is a model parameter that describes the magnitude of the change in the measured variable caused by the control output. It can be obtained by dividing the change in the measured variable by the change in the controller output that causes the change. T is the time constant, which describes the response speed of the measured variable when the controller output changes; τs is the lag time, which refers to the time from the step change of the system output to the time when the measured variable shows a clear initial response to the change, and s is the complex frequency variable.

[0033] The fuzzy adaptive process of parameters is as follows Figure 3 shown.

[0034] The basic form of fuzzy reasoning is shown as follows: if E=A i and if EC=B j then U 1 =C ji ; if E=A i and if EC=B j then U 2 =C ji ; in, E is the error, EC is the error change rate, A i , B j , U 1 , U 2 Relative incentive size X e , Response time ts , output weights and fuzzy subsets of expected accuracy thresholds. The present invention uses the Mamdani type MAX-MIN synthesis algorithm for fuzzy reasoning.

[0035] The algorithm uses the MAX-MIN synthesis algorithm to define the relationship between variables expressed by fuzzy implications, and uses the centroid method to accurately calculate the output.

[0036] For the input and output of the fuzzy system, this embodiment uses 7 linguistic variables for fuzzy description: negative large (NL), negative medium (NM), negative small (NS), zero (ZE), positive small (PS), positive medium (PM), positive large (PL). The fuzzy domain is divided into 7 levels, each of which represents a fuzzy subset, represented by the corresponding linguistic variable. The input and output are both represented by triangular membership functions. The fuzzy subset membership function of the linguistic variables of the input and output of the fuzzy controller is established as follows: Figure 4 , Figure 5 as well as Figure 6 shown.

[0037] Combined with the expert experience in the actual model and parameter application, the fuzzy rule table of output weight and expected accuracy is given according to the object characteristics and data characteristics of the actual process. Figure 7 as well as Figure 8 shown.

[0038] Therefore, after fuzzy rule calculation and defuzzification, accurate weight values ​​and oscillation conversion coefficients can be obtained, realizing indicator fusion based on fuzzy adaptive rules.

[0039] A method for detecting oscillations in a feedback control loop of a process control system in a process industry according to an embodiment of the present invention takes into account the oscillation characteristics of the loop data in the time domain and frequency domain, and performs effective fusion of indicators to effectively detect oscillations. In view of the diverse characteristics of field control loops in process industries and the coexistence of noise and external disturbances, the frequency domain detection method is applied to the oscillation detection process, and a method for quantitative calculation of frequency domain characteristic oscillations based on fast Fourier transform analysis is proposed. At the same time, this embodiment regularizes the experience of professional engineers and designs corresponding fuzzy adaptive strategies based on the characteristics of different control loops, effectively solving the problem that the weights of time domain indicators and frequency domain indicators in practical applications are difficult to judge due to the influence of objects and data conditions during the fusion process, solves the consistency problem of qualitative and quantitative evaluation of loop oscillations of different types and characteristics, and improves the accuracy of evaluation.

[0040] Example 2

[0041] A method for detecting oscillation of a feedback control loop of a process control system in a process industry according to this embodiment includes: Step S100, obtaining a data time series of a set value, a measured value, and a valve position value of a control loop, and performing data preprocessing on the data time series of the set value, the measured value, and the valve position value of the control loop to obtain first data; The first data includes a preprocessed set value data time series, a preprocessed measured value data time series, and a preprocessed valve position value data time series; Step S200: determining the time domain fluctuation characteristics and frequency characteristic characteristics of the preprocessed measurement value data time series according to the first data; Step S300, inputting the time domain fluctuation characteristics and the frequency characteristic characteristics into the oscillation degree acquisition algorithm set in advance to obtain the oscillation degree result of the control loop; The oscillation degree acquisition algorithm is based on fuzzy adaptive rule setting.

[0042] In the specific implementation process, the data time series of the set value (SP), measured value (PV) and valve position value (MV) of the control loop are collected using a real-time database system or other industrial data collection tools. The length of the data time series is set according to the specific situation of the control loop.

[0043] In this embodiment, step S100 includes: Step S110, according to the set value data time series, obtain the fluctuating set point and the corresponding fluctuation position, delete the data around the fluctuating set point, and obtain the pre-processed set value data time series; In the specific implementation process, we first need to define what is a fluctuating set point, which usually refers to those points that vary greatly relative to the surrounding points, usually due to short-term operational adjustments, system failures, or the system actively suppressing external interference. The fluctuating set point is determined by calculating the difference between adjacent points. When the difference between adjacent points is greater than the pre-set threshold, one of them is considered to be a fluctuating set point.

[0044] By removing fluctuating set points, it helps reduce the impact of active control loop actions on the data, making subsequent processing smoother and more reliable.

[0045] Step S120, sort the data time series of the measured values ​​and the data time series of the valve position values, remove the largest 0.5% and the smallest 0.5% data segments, and obtain the preprocessed data time series of the measured values ​​and the data time series of the valve position values.

[0046] For the measured values ​​and valve position values, they are sorted according to the numerical value, and the largest 0.5% and the smallest 0.5% data segments are removed after sorting. For example, assuming that the measured value data time series includes 100 data, the largest five data and the smallest five data are removed after sorting, thereby obtaining the preprocessed measured value data time series.

[0047] In this embodiment, step S200 includes: Step S210: Input the preprocessed measurement data time series into the following formula to obtain the time domain fluctuation characteristics: ; in, Std - pv is the standard deviation of the preprocessed measurement data time series, corresponding to the time domain fluctuation characteristics of the preprocessed measurement data time series, n is the data length, pv i is the time series of preprocessed measurement data, is the average value of the preprocessed measurement data time series.

[0048] In the specific implementation process, the time domain fluctuation characteristics of the measurement value data time series can also be mapped by the standard deviation of the control deviation. The formula for solving the standard deviation of the control deviation is as follows: ; e i =SP i -PV i ; in, Std - e is the standard deviation of the control deviation, n is the data length, e i To control the deviation, is the mean value of the control deviation, SP i is the preprocessed set value data time series, PV i is the time series of the preprocessed measurement data.

[0049] In this embodiment, step S200 further includes: Step S220, performing fast Fourier transform on the preprocessed measurement value data time series based on the time extraction DIT radix 2 FFT algorithm to obtain the frequency characteristic sequence of the preprocessed measurement value data time series; Step S230: quantitatively analyze the frequency characteristic sequence of the preprocessed measurement value data time series to obtain the frequency characteristic features of the preprocessed measurement value data time series.

[0050] In the specific implementation process, step S220 specifically includes: The preprocessed measurement value data time series PV(n) is separated into the odd part x(n) and the even part y(n) according to the parity, and the odd part is used as the real part of the complex number and the even part is used as the imaginary part of the complex number to obtain the complex sequence Z(n).

[0051] Input the complex sequence into the following formula to obtain the fast Fourier transform result, that is, the frequency characteristic sequence of the preprocessed measurement data time series: ; in, Z 0 is the value of the complex sequence Z(n) when n=0, that is, the first complex value of the measured value data time series after preprocessing, j is an imaginary unit, ( Z 0 ) * for Z 0 The complex conjugate of F N / 2 It is the frequency component of point N in the fast Fourier transform result that is located in the middle position of the FFT result, corresponding to the characteristic sequence of the preprocessed measurement value data time series at the N / 2 frequency.

[0052] In the specific implementation process, the complex sequence Z(n) is as follows: ; ; ; Then, the complex sequence Z(n) is subjected to a fast Fourier transform to obtain the transformed sequence Z(n). The fast Fourier transform result of the original sequence is: ; k = 0, 1, ... N / 2-1; Fe k It is equivalent to the fast Fourier transform result of x(n). Fo k It is equivalent to the fast Fourier transform result of y(n), where N is the total number of samples and k is the frequency index; Therefore, for k=N / 2+1 ,..., N-1, according to the conjugate property of real number sequence, we can get: F k = ( F N-k ) * ; For k=N / 2, since Z has periodicity Z 0 =Z N / 2 ,get: ; Among them, the fast Fourier transform process adopts the time extraction (DIT) based 2 FFT algorithm, the specific process is as follows: The Decimation in Time (DIT) Radix 2 FFT algorithm is an algorithm for fast calculation of Discrete Fourier Transform (DFT). It expresses the DFT of a long sequence as the DFT of multiple short sequences, and recursively decomposes the sequence into smaller parts, thus significantly reducing the amount of calculation. This algorithm is particularly suitable for processing DFT calculations of power-of-two sequences.

[0053] The basic principle of the time-decimated radix-2 FFT algorithm is to split a long sequence according to the odd and even subscripts and express it as the DFT of two or more short sequences. The specific process is as follows: For x(n) in an N-point sequence, its DFT transform pair is defined as: ; For the above formula, let N=2 M , M is a positive integer. According to the original DFT calculation principle, it is split into odd and even parts for calculation, that is, let n=2r and n=2r+1, r=0,1,...,N / 2-1, then: ; In the formula ,again ,make: ; Then we can get the following formula: N / 2 points A(k), B(k) can fully represent N points X(k), ; When N=8, N / 2 is still a DFT with a high complex number, so we need to continue to decompose it according to the above method: ; ; Thus, we get the formula that N / 4 points C(k), D(k) can fully represent N / 2 points A(k): ; Similarly, B(k) can be obtained from E(k) and F(k). At this time, C(k), D(k), E(k), and F(k) are all two-point DFTs, so there is no need to divide them further.

[0054] If N is a higher power of 2, continue dividing according to the above method until the DFT of the two points is obtained (such as Figure 2 as shown).

[0055] In this embodiment, step S230 includes: Step S231, performing amplitude normalization processing on the frequency characteristic sequence of the preprocessed measurement value data time series, and extracting the DC component and the signal amplitude other than the DC component to obtain the low-frequency band signal and the high-frequency band signal; Step S232: Obtain the peak value of the mid-low frequency band and the standard deviation of the amplitude of the high frequency band according to the mid-low frequency band signal and the high frequency band signal; Step S233: Obtain the frequency characteristic features of the preprocessed measurement value data time series according to the peak values ​​of the medium and low frequency bands and the standard deviation of the amplitudes of the high frequency bands.

[0056] Taking the low- and medium-frequency bands as an example, we first normalize the spectrum amplitude of the obtained process frequency domain characteristic sequence, and extract the signal amplitude corresponding to 0Hz from the original signal, also known as the DC component. The corresponding signal amplitude is: 0Hz corresponding amplitude = current value / number of sampling points len; The signal amplitude corresponding to the unexpected DC component is: ; Then determine the mid- and low-frequency limits freq - low : ; Among them, Ts is the sampling period of data. Generally, the sampling period of industrial process data varies from 1 to 10 seconds. Tss is the length of the selected data segment. Different loop lengths are selected according to the characteristics of the control loop, generally ranging from 20 minutes to 2 hours.

[0057] At this time, 0- freq - low The frequency band is the medium and low frequency band, and the peak amplitude value in the medium and low frequency band is obtained by traversing. Amp - max .

[0058] Step S233 specifically includes: The peak values ​​of the low and medium frequency bands and the standard deviation of the amplitude of the high frequency band are input into the following formula to obtain the frequency characteristic characteristics of the preprocessed measurement value data time series: ; in, Index - Osc is the oscillation coefficient of the control loop frequency characteristic, corresponding to the frequency characteristic characteristics of the preprocessed measurement value data time series, is the oscillation conversion coefficient, PV range is the range of the control loop measurement value,e is the interpolation between the set value and the measured value, Amp - max is the peak amplitude in the planting frequency band, is the standard deviation of the amplitude in the high frequency band.

[0059] in, It is the oscillation conversion coefficient. The specific parameter value is adjusted according to the characteristics of the loop. It is used to convert the frequency increase and the fluctuation size in the time domain into the same dimension for consideration, so as to achieve the unification of calculation and evaluation of different loop indicators.

[0060] In the specific implementation process, when the frequency fluctuates due to the oscillation of the loop itself, Index - Osc Obviously greater than 0, when the data oscillates due to noise, Index - Osc Obviously less than 0.

[0061] In this embodiment, step S300 includes: Input the time domain fluctuation characteristics and frequency characteristic characteristics into the following formula to obtain the quantitative results of the oscillation degree of the control loop: ; in, Y A quantitative calculation value for the degree of oscillation of the control loop, PV range is the range of the control loop measurement value, Index - Osc is the frequency characteristic of the time series of the preprocessed measurement data, Std - pv is the time domain fluctuation characteristic of the preprocessed measurement data time series, k 1 and k 2 is the corresponding weight, and k 1 and k 2 The sum of is 1.

[0062] In the specific implementation process, an indicator fusion method based on fuzzy adaptive rules is used to achieve the fusion and unified calculation of parameters of different dimensions, which facilitates operation and maintenance personnel to conduct qualitative evaluation and quantitative analysis of the oscillation conditions of different types and characteristic circuits based on a unified evaluation system and parameters under the same dimensional dimension.

[0063] The method of the embodiment of the present invention further includes: Obtain the relative excitation size and response time of the control loop for oscillation judgment, input them into the fuzzy knowledge base, obtain the oscillation conversion coefficient and the weights corresponding to the frequency characteristic characteristics and time domain fluctuation characteristics of the preprocessed measurement value data time series; The fuzzy knowledge base is generated through the judgment of experienced engineers on the fusion of indicators.

[0064] In the fuzzy knowledge base, seven linguistic variables are used for fuzzy description, and the fuzzy domain is divided into seven levels, each level represents a fuzzy subset, and each fuzzy subset is represented by a corresponding linguistic variable; The linguistic variables include: negative large, negative medium, negative small, zero, positive small, positive medium, and positive large.

[0065] The present embodiment is a method for detecting oscillation of a feedback control loop of a process control system in a process industry. By taking into account the oscillation characteristics of the combination of time domain and frequency domain, the fast Fourier transform method and data standard deviation operation are used for qualitative calculation and analysis, and the fuzzy mathematics technology is used in combination with the prior knowledge and experience of different loops in the process industry to quantify and integrate the analysis results of the time domain and frequency domain. The advantage of this method is that it can make full use of the understanding and judgment of a large number of experts on process knowledge, and effectively make up for the problem of difficult parameter design and inconsistent evaluation standards in existing schemes. This method can effectively improve the accuracy and efficiency of feedback control loop oscillation detection, and quickly improve the automation level of production equipment in the process industry.

[0066] In the description of the present invention, it should be understood that the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of the present invention, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0067] In the present invention, unless otherwise clearly specified and limited, the terms "installed", "connected", "connected", "fixed" and the like should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium; it can be the internal connection of two elements or the interaction relationship between two elements. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0068] In the present invention, unless otherwise clearly specified and limited, when a first feature is “on” or “below” a second feature, it may be that the first and second features are in direct contact, or that the first and second features are in indirect contact through an intermediate medium. Moreover, when a first feature is “above”, “above” or “above” a second feature, it may be that the first feature is directly above or obliquely above the second feature, or it may simply mean that the first feature is higher in level than the second feature. When a first feature is “below”, “below” or “below” a second feature, it may be that the first feature is directly below or obliquely below the second feature, or it may simply mean that the first feature is lower in level than the second feature.

[0069] In the description of this specification, the description of the terms "one embodiment", "some embodiments", "embodiment", "example", "specific example" or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, unless they are contradictory.

[0070] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations of the present invention. A person skilled in the art may alter, modify, replace and modify the above embodiments within the scope of the present invention.

Claims

1. A method for detecting oscillation of a feedback control loop of a process control system in a process industry, characterized in that: include: S100, obtaining a data time series of a set value, a measured value, and a valve position value of a control loop, and performing data preprocessing on the data time series of the set value, the measured value, and the valve position value of the control loop to obtain first data; The first data includes a preprocessed set value data time series, a preprocessed measured value data time series, and a preprocessed valve position value data time series; S200, determining the time domain fluctuation characteristics and frequency characteristic characteristics of the preprocessed measurement value data time series according to the first data; S300, inputting the time domain fluctuation characteristics and the frequency characteristic characteristics into a pre-set oscillation degree acquisition algorithm to obtain an oscillation degree result of the control loop; The oscillation degree acquisition algorithm is based on fuzzy adaptive rule setting.

2. The process control system feedback control loop oscillation detection method of process industry according to claim 1, characterized in that: The S100 includes: S110, obtaining a fluctuating set point and a corresponding fluctuating position according to the set value data time series, deleting the fluctuating set point, and obtaining a preprocessed set value data time series; S120, sorting the data time series of the measurement values ​​and the data time series of the valve position values, removing the largest 0.5% and the smallest 0.5% data segments, and obtaining the preprocessed data time series of the measurement values ​​and the data time series of the valve position values.

3. The method for detecting oscillation of a feedback control loop of a process control system in a process industry according to claim 1, characterized in that: The S200 includes: S210, input the preprocessed measurement data time series into the following formula to obtain the time domain fluctuation characteristics: ； in, Std - pv is the standard deviation of the preprocessed measurement data time series, corresponding to the time domain fluctuation characteristics of the preprocessed measurement data time series, n is the data length, pv i is the time series of preprocessed measurement data, is the average value of the preprocessed measurement data time series.

4. The method for detecting oscillation of a feedback control loop of a process control system in a process industry according to claim 3, characterized in that: The S200 further includes: S220, performing a fast Fourier transform on the preprocessed measurement value data time series based on a time extraction DIT radix 2 FFT algorithm to obtain a frequency characteristic sequence of the preprocessed measurement value data time series; S230 , quantitatively analyzing the frequency characteristic sequence of the preprocessed measurement value data time series to obtain frequency characteristic features of the preprocessed measurement value data time series.

5. The method for detecting oscillation of a feedback control loop of a process control system in a process industry according to claim 4, characterized in that: The S220 specifically includes: The preprocessed measured value data time series is converted into a complex number sequence, and the complex number sequence is input into the following formula to obtain a fast Fourier transform result, that is, a frequency characteristic sequence of the preprocessed measured value data time series: ； in, Z 0 is the value of the complex sequence Z(n) when n=0, that is, the first complex value of the measured value data time series after preprocessing, j is an imaginary unit, ( Z 0 ) * for Z 0 The complex conjugate of F N / 2 It is the frequency component of point N in the fast Fourier transform result that is located in the middle position of the FFT result, corresponding to the characteristic sequence of the preprocessed measurement value data time series at the N / 2 frequency.

6. The method for detecting oscillation of a feedback control loop in a process control system of a process industry according to claim 5, characterized in that: Converting the preprocessed measurement value data time series into a complex number sequence includes: The preprocessed measurement value data time series is separated into an odd part and an even part according to the parity, the odd part is used as the real part of the complex number, and the even part is used as the imaginary part of the complex number to obtain a complex number sequence.

7. The method for detecting oscillation of a feedback control loop of a process control system in a process industry according to claim 4, characterized in that: The S230 includes: S231, performing amplitude normalization processing on the frequency characteristic sequence of the preprocessed measurement value data time series, and extracting the DC component and the signal amplitude other than the DC component to obtain the low- and medium-frequency band signals and the high-frequency band signals; S232, obtaining the peak value of the mid-low frequency band and the standard deviation of the amplitude of the high frequency band according to the mid-low frequency band signal and the high frequency band signal; S233: Obtain frequency characteristic features of the preprocessed measurement value data time series according to the peak values ​​of the medium and low frequency bands and the standard deviation of the amplitudes of the high frequency bands.

8. The method for detecting oscillation of a feedback control loop of a process control system in a process industry according to claim 7, characterized in that: The S233 specifically includes: The standard deviation of the peak value of the medium and low frequency bands and the amplitude of the high frequency band is input into the following formula to obtain the frequency characteristic characteristics of the preprocessed measurement value data time series: ； in, Index - Osc is the oscillation coefficient of the control loop frequency characteristic, corresponding to the frequency characteristic characteristics of the preprocessed measurement value data time series, is the oscillation conversion coefficient, PV range is the range of the control loop measurement value, e is the interpolation between the set value and the measured value, Amp - max is the peak amplitude in the low and medium frequency bands, is the standard deviation of the amplitude in the high frequency band.

9. The method for detecting oscillation of a feedback control loop of a process control system in a process industry according to claim 1, characterized in that: The S300 includes: The time domain fluctuation characteristics and the frequency characteristic characteristics are input into the following formula to obtain the oscillation degree result of the control loop: ； in, Y is the result of the oscillation degree of the control loop, PV range is the range of the control loop measurement value, Index - Osc is the frequency characteristic of the time series of the preprocessed measurement data, Std - pv is the time domain fluctuation characteristic of the preprocessed measurement data time series, k 1 and k 2 is the corresponding weight, and k 1 and k 2 The sum of is 1.

10. The method for detecting oscillation of a feedback control loop of a process control system in a process industry according to any one of claims 8 or 9, characterized in that: Also includes: Obtain the relative excitation size and response time of the control loop for oscillation judgment, input them into the fuzzy knowledge base, obtain the oscillation conversion coefficient and the weights corresponding to the frequency characteristic characteristics and time domain fluctuation characteristics of the preprocessed measurement value data time series; The fuzzy knowledge base is generated through the judgment of experienced engineers on the fusion of indicators.

11. The method for detecting oscillation of a feedback control loop of a process control system in a process industry according to claim 10, characterized in that: The fuzzy knowledge base uses 7 linguistic variables for fuzzy description, and the fuzzy domain is divided into 7 levels, each level represents a fuzzy subset, and each fuzzy subset is represented by a corresponding linguistic variable; The language variables include: negative large, negative medium, negative small, zero, positive small, positive medium and positive large.

Images