Coriolis flowmeter wall failure detection method based on steady-state response

Abstract

The invention discloses a Coriolis flowmeter wall hanging fault detection method based on steady-state response, which specifically comprises the following steps: applying sinusoidal excitation to a Coriolis flowmeter in a fault-free state to obtain a steady-state response signal; filtering the steady-state response signal to obtain a standard characteristic signal; calculating the sample entropy of the standard characteristic signal; applying sinusoidal excitation to a Coriolis flowmeter to be tested to obtain a real-time response signal; filtering the real-time response signal to obtain a detection characteristic signal; calculating the sample entropy of the detection characteristic signal; normalizing the sample entropy of the standard characteristic signal and the sample entropy of the detection characteristic signal to obtain a detection result; the problem in the prior art that the detection of the Coriolis flowmeter wall hanging fault requires shutdown detection and the detection process is complicated is solved.

Description

Technical Field

[0001] The invention belongs to the technical field of Coriolis flowmeter fault detection, and in particular relates to a Coriolis flowmeter wall-hanging fault detection method based on steady-state response. Background Art

[0002] As the leader in the mass flow measurement industry, Coriolis mass flowmeter (hereinafter referred to as Coriolis flowmeter) plays an increasingly important role in process measurement such as fine chemical production, pharmaceutical synthesis, food processing, papermaking, and fluid trade transfer. With the development of the national economy and the continuous improvement of the refined control requirements of high-precision industries, higher requirements are placed on the performance indicators of Coriolis flowmeters such as measurement accuracy, reliability, and safety. The Coriolis flowmeter uses the influence of the Coriolis effect generated when the fluid flows through a vibrating pipe on the vibration phase at both ends of the pipe to measure the mass of the fluid flowing through the pipe. During the service of the Coriolis flowmeter, since the measured fluid is viscous and impure, it is affected by the measured fluid itself and the surrounding environment temperature, pressure and other factors. Impurities in the fluid are very easy to deposit in the pipe and cause wall failure. The wall failure of the Coriolis flowmeter measuring tube changes the physical rigidity value of the measuring tube, resulting in an offset of the measurement calibration factor, which directly affects the measurement accuracy and reliability of mass flow and other fluid information. Therefore, it is an urgent need to detect the service status of the Coriolis flowmeter and identify the wall failure of the measuring tube to improve the measurement accuracy and reliability of the Coriolis flowmeter.

[0003] At present, the detection of tube wall failure measured by Coriolis flowmeter mainly adopts the method of regular disassembly and offline detection to detect the wall failure of mass flowmeter; in the actual detection process, this regular offline detection method has the following problems: first, the offline detection method has a long downtime, which affects the working efficiency of the flowmeter; second, it is necessary to disassemble and reinstall the vibration tube, which consumes a lot of manpower and material resources, and the detection cost is high and the efficiency is low; third, the detection cycle is long, and during the detection cycle, the abnormal working state of the flowmeter cannot be monitored in real time. As the degree of wall failure of the measuring tube continues to increase, the measurement accuracy of the Coriolis flowmeter continues to decrease, which will produce a large measurement error. Therefore, a Coriolis flowmeter wall failure online detection method based on steady-state response is needed to solve the problems existing in the prior art. Summary of the invention

[0004] The purpose of the present invention is to provide a Coriolis flowmeter wall failure detection method based on steady-state response, which solves the problem in the prior art that the Coriolis flowmeter wall failure detection requires shutdown detection and the detection steps are complicated.

[0005] The technical means adopted by the present invention is a Coriolis flowmeter wall failure detection method based on steady-state response, which is specifically implemented according to the following steps:

[0006] Step 1, applying sinusoidal excitation to the Coriolis flowmeter in a fault-free state to obtain a steady-state response signal;

[0007] Step 2: Filter the steady-state response signal to obtain a standard characteristic signal;

[0008] Step 3, calculating the sample entropy of the standard characteristic signal;

[0009] Step 4, applying sinusoidal excitation to the Coriolis flowmeter to be tested to obtain a real-time response signal;

[0010] Step 5: Filter the real-time response signal to obtain a detection characteristic signal;

[0011] Step 6: Calculate the sample entropy of the detection feature signal;

[0012] Step 7: Normalize the sample entropy of the standard characteristic signal and the sample entropy of the detection characteristic signal to obtain a detection result.

[0013] The present invention is also characterized in that:

[0014] The frequencies of the sinusoidal excitation applied in step 1 and step 4 are both the natural frequencies of the Coriolis flowmeter, the excitation coil is used in both the sinusoidal excitation applied in step 1 and step 4, the acquisition coil is used to obtain the steady-state response signal in step 1, and the acquisition coil is used to obtain the real-time response signal in step 4.

[0015] The sampling frequency for obtaining the steady-state response signal in step 1 is 20 to 40 times the sinusoidal excitation frequency, and the sampling frequency for obtaining the real-time response signal in step 4 is 20 to 40 times the sinusoidal excitation frequency.

[0016] The specific process of filtering in step 2 is:

[0017] Step 2.1, construct the second-order low-pass filter transfer function H(s):

[0018]

[0019] In formula (1), ω0 represents the cutoff frequency, Q represents the quality factor, and s represents the frequency in Laplace transform;

[0020] Step 2.2: Discretize the filter transfer function H(s) using the bilinear transformation method, where the formula S of the bilinear transformation method is:

[0021]

[0022] In formula (2), z represents a complex variable in the digital domain, and T represents the sampling period;

[0023] Step 2.3: Multiply the steady-state response signal X1(t) by the low-pass filter transfer function H(s) discretized in step 2.2 to obtain the standard characteristic signal h(t).

[0024] The calculation process of the sample entropy of the standard characteristic signal in step 3 is as follows:

[0025] Step 3.1: Group the standard characteristic signal h(t) into a vector sequence of dimension m according to the sequence number:

[0026] [H m (1),…,H m (i), …, H m (N-m+1)](3)

[0027] In formula (3), H m (i) Specifically:

[0028] H m (i)={h(i), h(i+1), ..., h(N-m+1)} (4)

[0029] In formula (4), the value range of i is (1, N-m+1), i represents any dimension in m dimensions; N represents the total number of data, m represents the total number of dimensions, H m (i) represents the i-th dimension vector of the standard characteristic signal;

[0030] Step 3.2: Calculate H m (i) and H m (j) The distance between m (i),H m (j)]:

[0031] d[H m (i),H m (j)]=max k＝0,...,m-1 {|h(i+k)-h(j+k)|} (5)

[0032] In formula (5), j represents any dimension in m, and the value range of j is (1, N-m+1);

[0033] Step 3.3: Statistical d[H m (i),H m (j)] The number of j's that is not greater than r i And calculate the number B i Probability

[0034]

[0035] In formula (6), r is 0.1 times the standard deviation of the standard characteristic signal;

[0036] Step 3.4: Calculate H m The probability average of (i) is:

[0037]

[0038] Step 3.5: Increase the dimension of the standard characteristic signal to m+1 and calculate H m+1 (i) and H m+1 (j) The number of j whose distance between them is not greater than r A i Probability

[0039]

[0040] Step 3.6: Calculate H m+1 (i) The average value of probability A (m) (r):

[0041]

[0042] Step 3.7: Sample entropy SE of standard characteristic signal 标 for:

[0043]

[0044] The specific process of filtering in step 5 is:

[0045] Multiply the real-time response signal X2(t) by the low-pass filter transfer function H(s) discretized in step 2.2 to obtain the detection standard signal X3(t).

[0046] The calculation process of the sample entropy of the detection characteristic signal in step 6 is:

[0047] Step 6.1, group the detected characteristic signal X3(t) into a vector sequence with a number of m bits according to the sequence number:

[0048] [X m (1), …, X m (i), …X m (N-m+1)](11)

[0049] In formula (11), X m (i) Specifically:

[0050] X m (i)={X(i),X(i+1),…,X(N-m+1)} (12)

[0051] In formula (12), X m (i) represents the i-th dimension vector of the detection feature signal;

[0052] Step 6.2, calculate X m (i) and X m (j) The distance d[X m (i), X m (j)]:

[0053] d[X m (i), X m (j)]=max k＝0，…，m-1 {|X(i+k)-X(j+k)|} (13)

[0054] Step 6.3: Statistic d[X m (i), X m (j)] The number of j’s not greater than g C i And calculate the number C i Probability

[0055]

[0056] In formula (14), g is 0.1 times the standard deviation of the detection characteristic signal;

[0057] Step 6.4, calculate X m The probability average of (i) is:

[0058]

[0059] Step 6.5: Increase the dimension of the standard characteristic signal to m+1 and calculate X m+1 (i) and X m+1 (j) The number of j whose distance between them is not greater than g E i Probability

[0060]

[0061] Step 6.6, calculate X m+1 (i) The average value of probability E (m) (g)

[0062]

[0063] Step 6.7: Detect sample entropy SE of characteristic signal index for:

[0064]

[0065] The formula for normalization in step 7 is:

[0066]

[0067] When the value is not greater than 0.9, the detection result is that there is a wall hanging fault, otherwise the detection result is that there is no wall hanging fault.

[0068] The beneficial effects of the present invention are as follows: when detecting and processing the fault of the Coriolis flowmeter, it is only necessary to collect the steady-state response signal of the flowmeter in a fault-free state after the flowmeter is put into use, and calculate the sample entropy as a detection index to determine whether the existing flowmeter in use has a fault. The information to be processed is small, but the result is accurate. The coil collects its steady-state response signal with the natural frequency as the frequency of the sinusoidal excitation; the interference signal is removed by low-pass filtering, and the standard characteristic signal is extracted; then the sample entropy of the standard characteristic signal is calculated as the detection index, and then the sinusoidal excitation is applied to the measuring tube of the Coriolis flowmeter in operation, the real-time response signal is collected, and the sample entropy of the detection characteristic signal is calculated after the real-time response signal is processed by low-pass filtering, and the two sample entropy data are compared and processed to obtain the required detection results; it has the characteristics of being simple, reliable, easy to operate, low cost, and strong real-time performance, and is suitable for on-site real-time detection of the size of various wall-hanging faults of the Coriolis flowmeter, and has important engineering practical value. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] Figure 1 The Coriolis flowmeter wall failure detection test bench in Example 1;

[0070] Figure 2 is the Fourier spectrum of the steady-state response signal in Example 1;

[0071] Figure 3 is a spectrum diagram of the standard characteristic signal in Example 1;

[0072] Figure 4 These are test result diagrams for different conditions tested in Examples 1 to 3.

[0073] In the figure, 1. Coriolis flowmeter, 2. excitation coil, 3. collection coil, 4. processing unit. DETAILED DESCRIPTION

[0074] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments.

[0075] The Coriolis flowmeter wall failure detection method based on steady-state response of the present invention is specifically implemented according to the following steps:

[0076] Step 1, applying sinusoidal excitation to the Coriolis flowmeter in a fault-free state to obtain a steady-state response signal;

[0077] Step 2: Filter the steady-state response signal to obtain a standard characteristic signal;

[0078] Step 3, calculating the sample entropy of the standard characteristic signal;

[0079] Step 4, applying sinusoidal excitation to the Coriolis flowmeter to be tested to obtain a real-time response signal;

[0080] Step 5: Filter the real-time response signal to obtain a detection characteristic signal;

[0081] Step 6: Calculate the sample entropy of the detection feature signal;

[0082] Step 7: Normalize the sample entropy of the standard characteristic signal and the sample entropy of the detection characteristic signal to obtain a detection result.

[0083] The frequencies of the sinusoidal excitation applied in step 1 and step 4 are both the natural frequencies of the Coriolis flowmeter, the excitation coil is used in both the sinusoidal excitation applied in step 1 and step 4, the acquisition coil is used to obtain the steady-state response signal in step 1, and the acquisition coil is used to obtain the real-time response signal in step 4.

[0084] The sampling frequency for obtaining the steady-state response signal in step 1 is 20 to 40 times the sinusoidal excitation frequency, and the sampling frequency for obtaining the real-time response signal in step 4 is 20 to 40 times the sinusoidal excitation frequency.

[0085] The specific process of filtering in step 2 is:

[0086] Step 2.1, construct the second-order low-pass filter transfer function H(s):

[0087]

[0088] In formula (1), ω0 represents the cutoff frequency, Q represents the quality factor, and s represents the frequency in Laplace transform;

[0089] Step 2.2: Discretize the filter transfer function H(s) using the bilinear transformation method, where the formula S of the bilinear transformation method is:

[0090]

[0091] In formula (2), z represents a complex variable in the digital domain, and T represents the sampling period;

[0092] Step 2.3: Multiply the steady-state response signal X1(t) by the low-pass filter transfer function H(s) discretized in step 2.2 to obtain the standard characteristic signal h(t).

[0093] The calculation process of the sample entropy of the standard characteristic signal in step 3 is as follows:

[0094] Step 3.1: Group the standard characteristic signal h(t) into a vector sequence of dimension m according to the sequence number:

[0095] [H m (1),…,Hm (i), …, H m (N-m+1)](3)

[0096] In formula (3), H m (i) Specifically:

[0097] H m (i)={h(i),h(i+1),…,h(N-m+1)} (4)

[0098] In formula (4), the value range of i is (1, N-m+1), i represents any dimension in m dimensions; N represents the total number of data, m represents the total number of dimensions, H m (i) represents the i-th dimension vector of the standard characteristic signal;

[0099] Step 3.2: Calculate H m (i) and H m (j) The distance between m (i),H m (j)]:

[0100] d[H m (i),H m (j)]=max k＝0,...,m-1 {|h(i+k)-h(j+k)|} (5)

[0101] In formula (5), j represents any dimension in m, and the value range of j is (1, N-m+1);

[0102] Step 3.3: Statistical d[H m (i),H m (j)] The number of j's that is not greater than r i And calculate the number B i Probability

[0103]

[0104] In formula (6), r is 0.1 times the standard deviation of the standard characteristic signal;

[0105] Step 3.4: Calculate H m The probability average of (i) is:

[0106]

[0107] Step 3.5: Increase the dimension of the standard characteristic signal to m+1 and calculate H m+1 (i) and H m+1 (j) The number of j whose distance between them is not greater than r i Probability

[0108]

[0109] Step 3.6: Calculate H m+1 (i) The average value of probability A (m) (r):

[0110]

[0111] Step 3.7: Sample entropy SE of standard characteristic signal 标 for:

[0112]

[0113] The specific process of filtering in step 5 is:

[0114] Multiply the real-time response signal X2(t) by the low-pass filter transfer function H(s) discretized in step 2.2 to obtain the detection standard signal X3(t).

[0115] The calculation process of the sample entropy of the detection characteristic signal in step 6 is:

[0116] Step 6.1, group the detected characteristic signal X3(t) into a vector sequence with a number of m bits according to the sequence number:

[0117] [X m (1), …, X m (i), …X m (N-m+1)](11)

[0118] In formula (11), X m (i) Specifically:

[0119] X m (i)={X(i),X(i+1),…,X(N-m+1)} (12)

[0120] In formula (12), X m (i) represents the i-th dimension vector of the detection feature signal;

[0121] Step 6.2, calculate X m (i) and X m (j) The distance d[X m (i), X m (j)]:

[0122] d[X m (i), X m (j)]=max k＝0，…，m-1 {|X(i+k)-X(j+k)|} (13)

[0123] Step 6.3: Statistic d[X m (i), X m (j)] The number of j’s not greater than g C i And calculate the number C i Probability

[0124]

[0125] In formula (14), g is 0.1 times the standard deviation of the detection characteristic signal;

[0126] Step 6.4, calculate X m The probability average of (i) is:

[0127]

[0128] Step 6.5: Increase the dimension of the standard characteristic signal to m+1 and calculate X m+1 (i) and X m+1 (j) The number of j whose distance between them is not greater than g E i Probability

[0129]

[0130] Step 6.6, calculate X m+1 (i) The average value of probability E (m) (g)

[0131]

[0132] Step 6.7: Detect sample entropy SE of characteristic signal index for:

[0133]

[0134] The formula for normalization in step 7 is:

[0135]

[0136] When the value is not greater than 0.9, the detection result is that there is a wall hanging fault, otherwise the detection result is that there is no wall hanging fault.

[0137] Among them, the steady-state response signal X1(t) is low-pass filtered using the discretized low-pass filter transfer function. The filtering process can be assisted by the filter function h(t)=filter(b,a,x(t)) in Matlab, where b and a are the numerator and denominator coefficients of the filter, respectively, and h(t) is the standard characteristic signal after filtering.

[0138] According to the detection index The size of the flow meter wall failure detection, if the detection index value If it is not greater than 0.9, it means that the flow meter has a wall hanging fault. middle, The smaller it is, the more serious the flow meter wall failure is.

[0139] Example 1

[0140] The wall failure of the Coriolis flowmeter seriously affects the detection accuracy of the flowmeter and even causes serious accidents. Figure 1 The Coriolis flowmeter wall hanging fault detection test bench shown in the figure includes a collection coil 3 and an excitation coil 2 on the Coriolis flowmeter 1. The wall hanging fault is located between the collection coil 3 and the excitation coil 2. The information collected by the collection coil 3 is input into the processing unit 4 for calculation to obtain the detection index to determine whether the Coriolis flowmeter 1 has a fault. Counterweights of different weights are pasted at the bend position where the flowmeter is prone to wall hanging faults to simulate different degrees of wall hanging faults in the flowmeter. The simulation of different degrees of wall hanging faults in the flowmeter is shown in Table 1. The natural frequency of the flowmeter is used as the excitation frequency, a sinusoidal excitation signal is applied to the flowmeter through the excitation coil, and the steady-state response signal of the flowmeter is collected through the collection coil. The steady-state response signal is subjected to fast Fourier transform to obtain the following: Figure 2 The Fourier spectrum is shown in Figure 1. Based on the Fourier spectrum, the filter cutoff frequency is selected as ω = 4*ω0≈400Hz, and the quality factor is Q = 0.625. Then, the second-order low-pass filter transfer function is constructed and discretized. The steady-state response signal is low-pass filtered using the discretized low-pass filter function to obtain the spectrum of the standard characteristic signal, as shown in Figure 1. Figure 3 As shown, the sample entropy of the standard feature signal is calculated as the detection index.

[0141] The specific steps are as follows:

[0142] Step 1, applying sinusoidal excitation to the Coriolis flowmeter in a fault-free state to obtain a steady-state response signal;

[0143] Step 2: Filter the steady-state response signal to obtain a standard characteristic signal;

[0144] Step 3, calculating the sample entropy of the standard characteristic signal;

[0145] Step 4, applying sinusoidal excitation to the Coriolis flowmeter to be tested to obtain a real-time response signal;

[0146] Step 5: Filter the real-time response signal to obtain a detection characteristic signal;

[0147] Step 6: Calculate the sample entropy of the detection feature signal;

[0148] Step 7: Normalize the sample entropy of the standard characteristic signal and the sample entropy of the detection characteristic signal to obtain a detection result.

[0149] After calculation, when the Coriolis flowmeter is in a fault-free state, five sets of data are tested. The values ​​are 0.99, 1, 0.97, 0.99, 0.95. When the Coriolis flowmeter to be tested is weighed 17g, five different data are measured. The values ​​of are 0.8, 0.79, 0.8, 0.81, and 0.82. It can be seen that when If the value is lower than 0.85, the Coriolis flowmeter to be tested is faulty.

[0150] Example 2

[0151] The Coriolis flowmeter wall failure detection method based on steady-state response of the present invention is specifically implemented according to the following steps:

[0152] Step 1, applying sinusoidal excitation to the Coriolis flowmeter in a fault-free state to obtain a steady-state response signal;

[0153] Step 2: Filter the steady-state response signal to obtain a standard characteristic signal;

[0154] Step 3, calculating the sample entropy of the standard characteristic signal;

[0155] Step 4, applying sinusoidal excitation to the Coriolis flowmeter to be tested to obtain a real-time response signal;

[0156] Step 5: Filter the real-time response signal to obtain a detection characteristic signal;

[0157] Step 6: Calculate the sample entropy of the detection feature signal;

[0158] Step 7: Normalize the sample entropy of the standard characteristic signal and the sample entropy of the detection characteristic signal to obtain a detection result.

[0159] The frequencies of the sinusoidal excitation applied in step 1 and step 4 are both the natural frequencies of the Coriolis flowmeter, the excitation coil is used in both the sinusoidal excitation applied in step 1 and step 4, the acquisition coil is used to obtain the steady-state response signal in step 1, and the acquisition coil is used to obtain the real-time response signal in step 4.

[0160] The sampling frequency for obtaining the steady-state response signal in step 1 is 20 to 40 times the sinusoidal excitation frequency, and the sampling frequency for obtaining the real-time response signal in step 4 is 20 to 40 times the sinusoidal excitation frequency.

[0161] The specific process of filtering in step 2 is:

[0162] Step 2.1, construct the second-order low-pass filter transfer function H(s):

[0163]

[0164] In formula (1), ω0 represents the cutoff frequency, Q represents the quality factor, and s represents the frequency in Laplace transform;

[0165] Step 2.2: Discretize the filter transfer function H(s) using the bilinear transformation method, where the formula S of the bilinear transformation method is:

[0166]

[0167] In formula (2), z represents a complex variable in the digital domain, and T represents the sampling period;

[0168] Step 2.3: Multiply the steady-state response signal X1(t) by the low-pass filter transfer function H(s) discretized in step 2.2 to obtain the standard characteristic signal h(t).

[0169] After calculation, when the Coriolis flowmeter is in a fault-free state, five sets of data are tested. The values ​​are 0.99, 1, 0.97, 0.99, 0.95. When the Coriolis flowmeter to be tested is weighed 34g, five different data are measured. The values ​​are 0.65, 0.62, 0.615, 0.61, and 0.62. Compared with Example 1, the larger the hanging weight, The smaller the value of, the more likely the Coriolis flowmeter to be tested is The smaller the value, the more serious the fault.

[0170] Example 3

[0171] The Coriolis flowmeter wall failure detection method based on steady-state response of the present invention is specifically implemented according to the following steps:

[0172] Step 1, applying sinusoidal excitation to the Coriolis flowmeter in a fault-free state to obtain a steady-state response signal;

[0173] Step 2: Filter the steady-state response signal to obtain a standard characteristic signal;

[0174] Step 3, calculating the sample entropy of the standard characteristic signal;

[0175] Step 4, applying sinusoidal excitation to the Coriolis flowmeter to be tested to obtain a real-time response signal;

[0176] Step 5: Filter the real-time response signal to obtain a detection characteristic signal;

[0177] Step 6: Calculate the sample entropy of the detection feature signal;

[0178] Step 7: Normalize the sample entropy of the standard characteristic signal and the sample entropy of the detection characteristic signal to obtain a detection result.

[0179] The frequencies of the sinusoidal excitation applied in step 1 and step 4 are both the natural frequencies of the Coriolis flowmeter, the excitation coil is used in both the sinusoidal excitation applied in step 1 and step 4, the acquisition coil is used to obtain the steady-state response signal in step 1, and the acquisition coil is used to obtain the real-time response signal in step 4.

[0180] The sampling frequency for obtaining the steady-state response signal in step 1 is 20 to 40 times the sinusoidal excitation frequency, and the sampling frequency for obtaining the real-time response signal in step 4 is 20 to 40 times the sinusoidal excitation frequency.

[0181] The specific process of filtering in step 2 is:

[0182] Step 2.1, construct the second-order low-pass filter transfer function H(s):

[0183]

[0184] In formula (1), ω0 represents the cutoff frequency, Q represents the quality factor, and s represents the frequency in Laplace transform;

[0185] Step 2.2: Discretize the filter transfer function H(s) using the bilinear transformation method, where the formula S of the bilinear transformation method is:

[0186]

[0187] In formula (2), z represents a complex variable in the digital domain, and T represents the sampling period;

[0188] Step 2.3: Multiply the steady-state response signal X1(t) by the low-pass filter transfer function H(s) discretized in step 2.2 to obtain the standard characteristic signal h(t).

[0189] The calculation process of the sample entropy of the standard characteristic signal in step 3 is as follows:

[0190] Step 3.1: Group the standard characteristic signal h(t) into a vector sequence with dimension m according to the sequence number:

[0191] [H m (1),…,H m (i), …, H m (N-m+1)](3)

[0192] In formula (3), H m (i) Specifically:

[0193] Hm (i)={h(i),h(i+1),…,h(N-m+1)} (4)

[0194] In formula (4), the value range of i is (1, N-m+1), i represents any dimension in m dimensions; N represents the total number of data, m represents the total number of dimensions, H m (i) represents the i-th dimension vector of the standard characteristic signal;

[0195] Step 3.2: Calculate H m (i) and H m (j) The distance between m (i),H m (j)]:

[0196] d[H m (i),H m (j)]=max k＝0,...,m-1 {|h(i+k)-h(j+k)|} (5)

[0197] In formula (5), j represents any dimension in m, and the value range of j is (1, N-m+1);

[0198] Step 3.3: Statistical d[H m (i),H m (j)] The number of j's that is not greater than r i And calculate the number B i Probability

[0199]

[0200] In formula (6), r is 0.1 times the standard deviation of the standard characteristic signal;

[0201] Step 3.4: Calculate H m The probability average of (i) is:

[0202]

[0203] Step 3.5: Increase the dimension of the standard characteristic signal to m+1 and calculate H m+1 (i) and H m+1 (j) The number of j whose distance between them is not greater than r A i Probability

[0204]

[0205] Step 3.6: Calculate H m+1 (i) The average value of probability A (m) (r):

[0206]

[0207] Step 3.7: Sample entropy SE of standard characteristic signal 标 for:

[0208]

[0209] The specific process of filtering in step 5 is:

[0210] Multiply the real-time response signal X2(t) by the low-pass filter transfer function H(s) discretized in step 2.2 to obtain the detection standard signal X3(t).

[0211] The calculation process of the sample entropy of the detection characteristic signal in step 6 is:

[0212] Step 6.1, group the detected characteristic signal X3(t) into a vector sequence with a number of m bits according to the sequence number:

[0213] [X m (1), …, X m (i), …X m (N-m+1)](11)

[0214] In formula (11), X m (i) Specifically:

[0215] X m (i)={X(i),X(i+1),…,X(N-m+1)} (12)

[0216] In formula (12), X m (i) represents the i-th dimension vector of the detection feature signal;

[0217] Step 6.2, calculate X m (i) and X m (j) The distance d[X m (i), X m (j)]:

[0218] d[X m (i), X m (j)]=max k＝0，…，m-1 {|X(i+k)-X(j+k)|} (13)

[0219] Step 6.3: Statistic d[X m (i), X m (j)] The number of j’s not greater than g C i And calculate the number C i Probability

[0220]

[0221] In formula (14), g is 0.1 times the standard deviation of the detection characteristic signal;

[0222] Step 6.4, calculate X m The probability average of (i) is:

[0223]

[0224] Step 6.5: Increase the dimension of the standard characteristic signal to m+1 and calculate X m+1 (i) and X m+1 (j) The number of j whose distance between them is not greater than g E i Probability

[0225]

[0226] Step 6.6, calculate X m+1 (i) The average value of probability E (m) (g)

[0227]

[0228] Step 6.7: Detect sample entropy SE of characteristic signal index for:

[0229]

[0230] The formula for normalization in step 7 is:

[0231]

[0232] When the value is not greater than 0.9, the detection result is that there is a wall hanging fault, otherwise the detection result is that there is no wall hanging fault, such as Figure 4 The figure shows the detection results.

[0233] After calculation, when the Coriolis flowmeter is in a fault-free state, five sets of data are tested. The values ​​are 0.99, 1, 0.97, 0.99, and 0.95. When the Coriolis flowmeter to be tested is weighed 51g, five different data are measured. The values ​​are 0.55, 0.545, 0.54, 0.55, and 0.56. When the Coriolis flowmeter to be tested is weighed 68g, four different data are measured. The values ​​are 0.54, 0.53, 0.52, and 0.53. Compared with Example 1 and Example 2, the larger the hanging weight, The smaller the value of, the more likely the Coriolis flowmeter to be tested is The smaller the value, the more serious the fault. The results of the hanging weight simulation calculation are shown in Table 1:

[0234] Table 1 Simulation of Coriolis flowmeter wall failure

[0235] Flow meter wall failure state Fault simulation method Normal state Without weight Wall hanging state 1 Add 17g of counterweight Wall hanging state 2 Add 34g of counterweight Wall hanging state 2 Add 51g of counterweight Wall hanging state 4 Add 68g of counterweight

Claims

1. A Coriolis flowmeter wall failure detection method based on steady-state response, characterized in that: Follow the steps below to implement it: Step 1, applying sinusoidal excitation to the Coriolis flowmeter in a fault-free state to obtain a steady-state response signal; Step 2: Filter the steady-state response signal to obtain a standard characteristic signal; Step 3, calculating the sample entropy of the standard characteristic signal; Step 4, applying sinusoidal excitation to the Coriolis flowmeter to be tested to obtain a real-time response signal; Step 5: Filter the real-time response signal to obtain a detection characteristic signal; Step 6: Calculate the sample entropy of the detection feature signal; Step 7: normalize the sample entropy of the standard characteristic signal and the sample entropy of the detection characteristic signal to obtain a detection result; The formula for the normalization process is: （19） In formula (19), Represents the sample entropy of the detection feature signal, Represents the sample entropy of the standard characteristic signal; When the value is not greater than 0.9, the detection result is that there is a wall hanging fault, otherwise the detection result is that there is no wall hanging fault.

2. The method for detecting wall-hanging fault of Coriolis flowmeter based on steady-state response according to claim 1 is characterized in that: The frequencies of applying sinusoidal excitation in step 1 and step 4 are both the natural frequencies of the Coriolis flowmeter, the sinusoidal excitations applied in step 1 and step 4 use an excitation coil, the steady-state response signal obtained in step 1 uses a collection coil, and the real-time response signal obtained in step 4 uses a collection coil.

3. The method for detecting wall failure of a Coriolis flowmeter based on steady-state response according to claim 2 is characterized in that: The sampling frequency for obtaining the steady-state response signal in step 1 is 20 to 40 times the sinusoidal excitation frequency, and the sampling frequency for obtaining the real-time response signal in step 4 is 20 to 40 times the sinusoidal excitation frequency.

4. The method for detecting wall-hanging fault of Coriolis flowmeter based on steady-state response according to claim 1 is characterized in that: The specific process of the filtering process described in step 2 is: Step 2.1: Construct the second-order low-pass filter transfer function H(s) : （1） In formula (1), represents the cutoff frequency, Q represents the quality factor, s represents the frequency in the Laplace transform; Step 2.2: Use the bilinear transformation method to transform the filter transfer function H(s) Discretization is performed, where the formula S of the bilinear transformation method is: （2） In formula (2), z represents a complex variable in the digital domain, and T represents the sampling period; Step 2.3: Compare the steady-state response signal X1(t) with the low-pass filter transfer function discretized in step 2.2 H(s) Multiply to get the standard characteristic signal h(t) .

5. The method for detecting wall-hanging fault of Coriolis flowmeter based on steady-state response according to claim 4 is characterized in that: The calculation process of the sample entropy of the standard characteristic signal in step 3 is as follows: Step 3.1: Group the standard characteristic signals h(t) into a set of dimensions according to the sequence number. m The vector sequence: （3） In formula (3), H m ( i ) Specifically: （4） In formula (4), the value range of i is (1, N-m+1), i represents any dimension in m dimensions; N represents the total number of data, m represents the total dimension, H m (i) The standard characteristic signal i dimensional vector; Step 3.2, calculation and Distance : （5） In formula (5), j represents any dimension in m, and the value range of j is (1, N-m+1); Step 3.3: Statistics The number of j not greater than r And calculate the number Probability : （6） In formula (6), r is 0.1 times the standard deviation of the standard characteristic signal; Step 3.4, calculation The average probability of: （7） Step 3.5: Increase the dimension of the standard characteristic signal to m+1 and calculate and The number of j whose distance is not greater than r Probability : （8） Step 3.6, calculation The average probability : （9） Step 3.7: Sample entropy of standard characteristic signal for: （10）。 6. The method for detecting wall-hanging fault of Coriolis flowmeter based on steady-state response according to claim 5 is characterized in that: The specific process of the filtering process described in step 5 is: The real-time response signal X2(t) is compared with the low-pass filter transfer function after discretization in step 2.2 H(s) Multiply them together to obtain the detection feature signal X3(t).

7. The method for detecting wall-hanging fault of Coriolis flowmeter based on steady-state response according to claim 6 is characterized in that: The calculation process of the sample entropy of the detection characteristic signal in step 6 is: Step 6.1, group the detected characteristic signal X3(t) into a vector sequence with a number of m bits according to the sequence number: （11） In formula (11), Specifically: （12） In formula (12), Represents the i-th dimension vector of the detection feature signal; Step 6.2: Calculation and Distance : （13） Step 6.3: Statistics The number Ci of j not greater than g and calculate the probability of number Ci : （14） In formula (14), g is 0.1 times the standard deviation of the detection characteristic signal; Step 6.4: Calculation The average probability of: （15） Step 6.5: Increase the dimension of the detection feature signal to m+1 and calculate and The distance between The number of j E i Probability : （16） Step 6.6, Calculation The average probability : （17） Step 6.7: Detect sample entropy of characteristic signal for: （18）。

Images