A method for online detection of surge of turbocharger centrifugal compressor

Abstract

The present invention discloses an online detection method for surge of a turbocharger centrifugal compressor, which belongs to the technical field of state detection and diagnosis of turbocharger centrifugal compressors. The present invention can identify surges at other speeds based on the judgment threshold determined at the lowest speed, thereby expanding the scope of application. After surge occurs, the judgment threshold is determined based on the maximum #imgabs0# under steady-state conditions and the maximum #imgabs1# in the first surge cycle. There is no need to collect periodic pressure signals, and the surge occurring in the compressor can be judged only within half a surge cycle, thereby significantly improving the real-time performance of surge detection. The present invention increases the signal-to-noise ratio of the signal to be detected and improves the accuracy of the detection result by removing singular points and performing wavelet filtering. The present invention replaces manual judgment of surge with online surge detection, thereby ensuring the consistency of surge detection results. The present invention can improve the real-time, universality and accuracy of surge detection during the operation of a supercharged engine and during compressor performance testing.

Description

Technical Field

[0001] The invention belongs to the technical field of turbocharger centrifugal compressor state detection and diagnosis, and relates to an online detection method for identifying turbocharger centrifugal compressor surge. Background Art

[0002] Turbocharging is an effective measure to improve engine power and fuel economy, greatly improving the performance of internal combustion engines and is an important direction for the development of internal combustion engines. However, during the operation of the supercharger, the surge of the compressor will reduce the performance of the supercharger and affect the power output of the engine. Surge is caused by the reduction of the inlet flow of the compressor to a certain extent at a certain speed. At this time, the stability of the forward movement of the air in the compressor is destroyed, the airflow vibrates strongly, and the flow, pressure, and power in the compressor fluctuate greatly. Surge can cause damage to the components in the supercharger and is the most harmful and destructive type of fault in the supercharger.

[0003] Real-time surge detection is required during the operation of the supercharged engine and when testing the performance of the compressor, but there are some difficulties in surge detection. For example, the characteristics of the surge signal are complex. Even the same model of compressor has different surge signal characteristics under different operating conditions and installation methods, resulting in poor versatility of the surge detection method. In addition, the signal may be affected by external electromagnetic interference during the signal acquisition process, resulting in singular points in the signal, affecting the accuracy of the surge detection method. Therefore, how to improve the real-time, versatility and accuracy of surge detection has always been the goal pursued by researchers. Summary of the invention

[0004] The main purpose of the present invention is to provide an online detection method for turbocharger centrifugal compressor surge, which can realize online detection of turbocharger centrifugal compressor surge and improve the real-time, universality and accuracy of surge detection during the operation of the supercharged engine and the compressor performance test.

[0005] The technical solution of the present invention is implemented as follows:

[0006] The present invention discloses a method for online detection of turbocharger centrifugal compressor surge, comprising the following steps:

[0007] Step 1: Install a dynamic pressure sensor on the turbocharger compressor outlet pipe, collect the compressor outlet dynamic pressure signal through a surge test, and realize surge signal collection.

[0008] Step 2: Perform M-layer binary wavelet decomposition on the pressure signal, determine the location of the singular point using the wavelet transform modulus maximum and the Lipschitz index α calculated at the point, and remove the singular points in the signal.

[0009] Step 2.1: Perform M-layer dyadic wavelet decomposition on the pressure signal;

[0010] Select the wavelet basis, decompose the pressure signal into different scales, and obtain the wavelet coefficients of each layer: Wf(2 1 ,b)、Wf(2 2 , b),…,Wf(2 M ,b),2 m is the wavelet transform scale, m=1,2,…,M, b is the displacement parameter;

[0011] Step 2.2: In 2 M-1 , 2 M Scale to find the maximum value of signal modulus;

[0012] The method for extracting the maximum value of the wavelet sequence modulus is as follows:

[0013] If |Wf(2 m ,b)|≤|Wf(2 m ,b 0 )|, m = M-1, M, then the wavelet transform is 2 m b under scale 0 The modulus maximum is obtained at point

[0014] Step 2.3: Set the test threshold T 1 , when the maximum value of the wavelet coefficient modulus is greater than the test threshold T 1 When , the Lipschitz index α of the signal at the modulus maximum is calculated to remove the singular points of the signal;

[0015] When 2 M The wavelet coefficient at the scale modulus maximum exceeds the test threshold T 1 And 2 M-1 When the wavelet coefficient at the corresponding time under the scale is not 0, calculate the Lipschitz index of the signal.

[0016] For 2 M-1 , 2 M Scale, in b 0 The calculation formula of Lipschitz index α at point is as follows

[0017]

[0018] If the Lipschitz index of the signal at the modulus maximum is less than 0, then the value of this point is equal to the sampling value of the previous point, and the value of the signal after 2 points is equal to the sampling value of the previous point. M -No more singularity detection is performed within 1 sampling period.

[0019] Preferably, the test threshold T in step 2.3 is 1 The setting method is as follows:

[0020] When testing the performance of a turbocharger compressor, select the signal at any speed among the measured speeds. When a singular point appears, M The maximum value of the wavelet coefficient under steady-state conditions is taken from the wavelet coefficients under the scale 1 And the wavelet coefficient W when the singular point appears 2 , the wavelet threshold takes the average of the two, that is, the test threshold T is determined 1 , the calculation formula is as follows

[0021]

[0022] in,

[0023]

[0024] Where P is the steady-state condition2 M The number of wavelet coefficients at the scale,

[0025] W 2 =Wf(2 M ,b s ) (4)

[0026] Where b s is the moment when the singular point of the sampled signal appears.

[0027] Step 3: Perform wavelet transform on the signal after removing the singular points in step 2, retain all wavelet coefficients at large scales, and set the wavelet threshold T for the wavelet coefficients at other scales. 2 , below the wavelet threshold T 2 The coefficients are set to 0, and the processed wavelet coefficients are reconstructed using inverse wavelet transform to restore the valid signal.

[0028] Step 4: After each sampling, calculate the mean of the sum of squares of the differences between the latest sampling point and other sampling points in the time window for the signal filtered in step 3.

[0029] The calculation formula is as follows

[0030]

[0031] Where N is the number of sampling points in the time window, p i is the sampling value at other moments in the time window; p N is the latest sample value in the time window.

[0032] When the pressure signal p′(t) is disturbed by noise n(t) during the sampling process, the discrete expression of the signal obtained after passing through the sampling system in the time domain is:

[0033] p i =p′i +n i (6)

[0034] but Another way to express it is

[0035]

[0036] in, is the mean value of the pressure sampling data without noise interference, is the mean value of the pressure signal data collected by the sampling system, is the noise mean.

[0037] The first term on the right side of the equal sign in formula (7) represents the fluctuation of the compressor outlet pressure signal; the second term is the covariance between the compressor outlet pressure signal and the noise signal, reflecting the degree of linearity between the two random variables, and this value can be considered to be equal to 0; the third term is caused by noise and is the variance of the noise.

[0038] When the compressor outlet pressure signal is detected under steady-state or surge conditions, the noise generated by the sampling system obeys the same distribution, and the influence of the speed on the third term on the right side of the equation (7) can be ignored. Under steady-state conditions, the change in compressor outlet pressure can be ignored, and the influence of the speed on the first term on the right side of the equation (7) can be ignored. However, under surge conditions, the sampling value deviates from its mean by a large margin, and as the speed increases, the degree of deviation also gradually increases. The first term on the right side of the equation (7) increases with the increase in speed. That is, under steady-state conditions, the speed has a significant influence on the mean. The impact can be ignored, and under surge conditions Increases with the increase of speed.

[0039] Step 5: Determine the judgment threshold ΔS according to the calculation result of step 4.

[0040] Select the lowest speed among the measured speeds, and take the maximum value S of the sum of squares of the difference between the latest sampling point and other sampling points in the steady-state working time window. 1 The maximum value S of the sum of squares of the difference between the latest sampling point and other sampling points in the first surge cycle time window 2 , the judgment threshold is the average of the two, that is, the judgment threshold ΔS is determined, and the calculation formula is as follows:

[0041]

[0042] in,

[0043]

[0044] In the formula, Q k,k=1,2, is the number of times the mean of the sum of squares of the differences between the latest sampling point and other sampling points in the time window is calculated, It is the mean of the sum of squares of the differences between the latest sampling point and other sampling points in the time window obtained by the j-th calculation under steady-state conditions.

[0045] In step 5, the judgment threshold obtained based on the lowest speed is obtained by the maximum value S of the sum of squares of the differences between the latest sampling point and other sampling points in the steady-state operating time window. 1 The maximum value S of the sum of squares of the difference between the latest sampling point and other sampling points in the first surge cycle time window 2 When the measured supercharger speed increases, the speed has no effect on the steady-state condition. The impact can be ignored, but under surge conditions It increases with the increase of the rotation speed, that is, surge identification can be performed based on the determination threshold determined at the lowest rotation speed. The determination threshold determined at the lowest rotation speed is applicable to surge identification at other rotation speeds, thereby expanding the scope of application of the present invention.

[0046] In step 5, based on the two working conditions The judgment threshold obtained by the average value is that when the pressure signal sampling value gradually decreases to the minimum value in the first surge cycle, the difference between the latest sampling point value and the other sampling point values ​​continues to increase. When the minimum value point of the pressure signal in the first surge cycle is collected, the difference between the latest sampling point value and the other sampling point values ​​reaches the maximum, that is, After half a surge cycle, it reaches the maximum value. and the maximum value in the first surge cycle To determine the judgment threshold and judge whether surge occurs, there is no need to collect periodic pressure signals. It can judge whether the compressor surges within only half a surge cycle, which significantly improves the surge detection efficiency and realizes online surge detection of turbocharger centrifugal compressor. According to the surge online detection result, measures can be taken when the compressor surges to ensure the reliability of engine operation.

[0047] Step 6: Collect the dynamic pressure sensor signal installed in step 1 in real time, use step 2 to remove the signal singularity points and step 3 to filter the signal to improve the signal-to-noise ratio of the signal, and calculate the mean of the sum of square differences between the latest sampling point and other sampling points in the time window according to step 4, and compare the calculation result of step 4 with the judgment threshold determined in step 5 in real time. When the mean of the sum of square differences between the latest sampling point and other sampling points in the time window is greater than the judgment threshold, it is judged that surge occurs, that is, real-time judgment of surge is achieved.

[0048] Beneficial effects:

[0049] 1. The invention discloses an online detection method for turbocharger centrifugal compressor surge. When the pressure signal is detected under steady-state conditions, the noise generated by the test system follows the same distribution, and the speed is The influence can be ignored, but under surge conditions, the calculated results under different conditions are It will increase with the increase of the speed, that is, other speed surge identification can be performed based on the judgment threshold determined at the lowest speed, thereby expanding the application scope of the present invention.

[0050] 2. The present invention discloses a method for detecting surge of a turbocharger centrifugal compressor online. After the surge occurs, when the minimum value point in the first surge cycle is collected, the surge detection calculation result The maximum value is reached in the first surge cycle, exceeding the judgment threshold. There is no need to collect periodic pressure signals. The surge of the compressor can be judged within only half a surge cycle, and each time The calculation requires only a small amount of sampling signals and a small amount of calculation, which can significantly improve the real-time performance of surge detection. According to the surge online detection result, measures can be taken when surge occurs in the compressor to ensure the reliability of engine operation.

[0051] 3. The present invention discloses a turbocharger centrifugal compressor surge online detection method, which increases the signal-to-noise ratio of the signal to be detected and improves the accuracy of the detection result through singular point removal processing and wavelet filtering.

[0052] 4. In view of the problem that the manual method of judging surge in the turbocharger compressor performance test has inconsistent judgment on the surge boundary, the present invention discloses an online detection method for surge of a turbocharger centrifugal compressor, which replaces the manual judgment of surge in the turbocharger compressor performance test with automatic online surge detection, thereby ensuring the consistency of surge detection results. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 The present invention discloses a flow chart of an online detection method for surge of a turbocharger centrifugal compressor

[0054] Figure 2 This is the surge signal diagram of JP60 at 60000r / min

[0055] Figure 3 The three-layer Haar wavelet decomposition of the pressure signal of JP60 at 60000r / min

[0056] Figure 4 The pressure signal and the maximum value of the signal modulus of JP60 at 60000r / min

[0057] Figure 5The signal after removing the singular points and filtering for JP60 at 60000r / min

[0058] Figure 6 Calculated for JP60 at 60000r / min

[0059] Figure 7 This is the surge signal diagram of JP60 at 80000r / min

[0060] Figure 8 The three-layer Haar wavelet decomposition of the pressure signal of JP60 at 80000r / min

[0061] Fig. 9 The pressure signal and the maximum value of the signal modulus of JP60 at 80000r / min

[0062] Fig.10 The signal after removing the singular points and filtering for JP60 at 80000r / min

[0063] Fig.11 Calculated for JP60 at 80000r / min DETAILED DESCRIPTION

[0064] In order to enable those skilled in the art to better understand the solutions of the present invention, the contents of the present invention are further described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0065] Embodiment 1:

[0066] like Figure 1 As shown, the present embodiment discloses a method for online detection of turbocharger centrifugal compressor surge, the flow chart of which is as follows Figure 1 As shown, the following steps are included:

[0067] Step 1: Install a dynamic pressure sensor on the turbocharger compressor outlet pipe, and collect the compressor outlet dynamic pressure signal through the surge test. When collecting the signal, set the sampling frequency to 100Hz and the turbocharger speed to 60000r / min. During the test, gradually reduce the compressor outlet valve opening to make the compressor run to the surge state, and obtain the compressor outlet pressure signal under steady-state and surge conditions. The steady-state and surge conditions of the JP60 turbocharger compressor are shown in the figure. Figure 2 shown. Figure 2 In the example, at 5s, the compressor enters surge condition from steady state condition.

[0068] Step 2: Perform M-layer binary wavelet decomposition on the pressure signal, determine the location of the singular point using the wavelet transform modulus maximum and the Lipschitz index α calculated at the point, and remove the singular point in the signal;

[0069] Step 2.1: Perform M-layer wavelet decomposition on the pressure signal;

[0070] Select Haar wavelet, decompose the pressure signal into three layers of wavelet, and record the wavelet coefficients of each layer of decomposition, such as Figure 3 shown.

[0071] Step 2.2: In 2 M-1 , 2 M Find the maximum value of the signal modulus under the scale;

[0072] Take 2 2 , 2 3 The scale calculates the maximum value of the modulus of the signal, such as Figure 4 shown.

[0073] Step 2.3: Set the test threshold T 1 , when the maximum value of the wavelet coefficient modulus is greater than the test threshold T 1 When , the Lipschitz index α of the signal at the modulus maximum is calculated to remove the singular points of the signal;

[0074] According to the test threshold calculation formula,

[0075]

[0076] The W obtained in step 2.2 1 =0.1144,W 2 =0.5448, substitute into the formula to get the test threshold T 1 =0.33;

[0077] The calculated signal is at 4.28s and 2 2 , 2 3 The α under the scale is -0.4327. At this point, the Lipschitz index of the signal is less than 0. The value of this point is equal to the sampling value of the previous point, completing the detection and removal of the singular point of the pressure signal.

[0078] Step 3: Perform wavelet transform on the signal after removing the singular points, and keep 2 3 For all wavelet coefficients at the same scale, set the wavelet threshold to T for the wavelet coefficients at other scales. 2 =0, i.e. 2 1 , 2 2 The wavelet coefficients under the scale are set to 0, and finally the processed wavelet coefficients are reconstructed using inverse wavelet transform to restore the valid signal, such as Figure 5 shown.

[0079] Step 4: After each sampling, calculate the mean of the sum of squares of the differences between the latest sampling point and other sampling points for the signal filtered in step 3.

[0080] The calculation results are as follows Figure 6 shown.

[0081] Step 5: Determine the judgment threshold ΔS according to the calculation result of step 4;

[0082] According to the judgment threshold calculation formula,

[0083]

[0084] The S obtained in step 4 1 =0.00085, S 2 =0.011, and substituting it into the formula, we get the judgment threshold ΔS≈0.006.

[0085] Step 6: Collect the dynamic pressure sensor signal installed in step 1 in real time, use step 2 to remove the signal singularity points and step 3 to filter the signal to improve the signal-to-noise ratio of the signal, and calculate the mean of the sum of square differences between the latest sampling point and other sampling points in the time window according to step 4, and compare the calculation result of step 4 with the judgment threshold ΔS=0.006 determined in step 5 in real time. When the mean of the sum of square differences between the latest sampling point and other sampling points in the time window is greater than the judgment threshold, it is judged that surge occurs, that is, real-time judgment of surge is achieved.

[0086] Embodiment 2:

[0087] The present embodiment discloses a method for online detection of turbocharger centrifugal compressor surge, comprising the following steps:

[0088] Step 1: Install a dynamic pressure sensor on the turbocharger compressor outlet pipe, and collect the compressor outlet dynamic pressure signal through the surge test. When collecting the signal, set the sampling frequency to 100Hz and the turbocharger speed to 80000r / min. During the test, gradually reduce the compressor outlet valve opening to make the compressor run to the surge state, and obtain the compressor outlet pressure signal under steady-state and surge conditions. The steady-state and surge conditions of the JP60 turbocharger compressor are shown in the figure. Figure 7 shown. Figure 7 In the figure, at 5.1s, the compressor enters surge condition from steady state condition.

[0089] Step 2: Remove the singular points in the signal according to wavelet transform;

[0090] Step 2.1: Perform M-layer wavelet decomposition on the pressure signal;

[0091] Select Haar wavelet, decompose the pressure signal into three layers of wavelet, and record the wavelet coefficients of each layer of decomposition, such as Figure 8 shown.

[0092] Step 2.2: In 2 M-1 , 2 M Scale to find the maximum value of signal modulus;

[0093] Take 2 2 , 2 3 Calculate the maximum value of the signal modulus on the scale, such as Fig. 9 shown.

[0094] Step 2.3: Set the test threshold T 1 , when the maximum value of the wavelet coefficient modulus is greater than the test threshold T 1 When , the Lipschitz index α of the signal at the modulus maximum is calculated to remove the singular points of the signal;

[0095] Setting the test threshold is the same as step 2.3 of Example 1, T 1 =0.33.

[0096] The modulus maximum of the signal exceeds the test threshold at 5.25s. The calculated signal is at 5.25s and 2 2 , 2 3 The α under scale is 4.7, and this point is not a singular point.

[0097] Step 3: Perform wavelet transform on the signal after removing the singular points, and keep 2 3 For all wavelet coefficients at the same scale, the threshold is set to T for wavelet coefficients at other scales. 2 =0, i.e. 2 1 , 2 2 The wavelet coefficients under the scale are set to 0, and finally the processed wavelet coefficients are reconstructed using inverse wavelet transform to restore the valid signal, such as Fig.10 shown.

[0098] Step 4: After each sampling, calculate the mean of the sum of squares of the differences between the latest sampling point and other sampling points for the signal filtered in step 3.

[0099] The calculation results are as follows Fig.11 shown.

[0100] Step 5: Determine the judgment threshold ΔS according to the calculation result of step 4;

[0101] The determination threshold is the same as that in step 5 of embodiment 1, ΔS=0.006.

[0102] Step 6: Collect the dynamic pressure sensor signal installed in step 1 in real time, use step 2 to remove the signal singularity points and step 3 to filter the signal to improve the signal-to-noise ratio of the signal, and calculate the mean of the sum of square differences between the latest sampling point and other sampling points in the time window according to step 4, and compare the calculation result of step 4 with the judgment threshold ΔS=0.006 determined in step 5 in real time. When the mean of the sum of square differences between the latest sampling point and other sampling points in the time window is greater than the judgment threshold, it is judged that surge occurs, that is, real-time judgment of surge is achieved.

[0103] Some embodiments of the present invention described herein are intended to be used to explain the present application, but those skilled in the art who do not make creative efforts and come up with some implementation methods that do not change the substantial content of the invention should all fall within the scope of protection required by the present invention. Therefore, it should be understood that the attached claims are intended to cover these modifications and changes that are essentially the same as the substantial content of the present invention.

Claims

1. A method for online detection of surge of a turbocharger centrifugal compressor, characterized in that: The following steps are included: Step 1: Install a dynamic pressure sensor on the turbocharger compressor outlet pipe, collect the compressor outlet dynamic pressure signal through a surge test, and realize surge signal collection; Step 2: Perform M-layer binary wavelet decomposition on the pressure signal, determine the location of the singular point using the wavelet transform modulus maximum and the Lipschitz index α calculated at the point, and remove the singular point in the signal; Step 3: Perform wavelet transform on the signal after removing the singular points in step 2, retain all wavelet coefficients at large scales, set the wavelet threshold T2 for the wavelet coefficients at other scales, set the coefficients below the wavelet threshold T2 to 0, reconstruct the processed wavelet coefficients using inverse wavelet transform, and restore the valid signal; Step 4: After each sampling, calculate the mean of the sum of squares of the differences between the latest sampling point and other sampling points in the time window for the signal filtered in step 3. Step 5: Determine the judgment threshold ΔS according to the calculation result of step 4; In step 5, Select the lowest speed among the measured speeds, take the maximum value S1 of the mean of the sum of squares of the difference between the latest sampling point and other sampling points in the steady-state working condition time window and the maximum value S2 of the sum of squares of the difference between the latest sampling point and other sampling points in the first surge cycle time window, and take the average value of the two to determine the judgment threshold, that is, determine the judgment threshold ΔS, and the calculation formula is as follows: in, In the formula, Q k ,k=1,2, is the number of times the mean of the sum of squares of the differences between the latest sampling point and other sampling points in the time window is calculated, is the mean of the sum of squares of the differences between the latest sampling point and other sampling points in the time window obtained by the j-th calculation under steady-state conditions; Step 6: Collect the dynamic pressure sensor signal installed in step 1 in real time, use step 2 to remove the signal singularity points and step 3 to filter the signal to improve the signal-to-noise ratio of the signal, and calculate the mean of the sum of square differences between the latest sampling point and other sampling points in the time window according to step 4, and compare the calculation result of step 4 with the judgment threshold determined in step 5 in real time. When the mean of the sum of square differences between the latest sampling point and other sampling points in the time window is greater than the judgment threshold, it is judged that surge occurs, that is, real-time judgment of surge is achieved.

2. A turbocharger centrifugal compressor surge online detection method as claimed in claim 1, characterized in that: Step 2 is implemented as follows: Step 2.1: Perform M-layer dyadic wavelet decomposition on the pressure signal; Select the wavelet basis, decompose the pressure signal into different scales, and obtain the wavelet coefficients of each layer: Wf(2 1 ,b)、Wf(2 2 , b),…,Wf(2 M ,b),2 m is the wavelet transform scale, m=1,2,…,M, b is the displacement parameter; Step 2.2: In 2 M-1 , 2 M Scale to find the maximum value of the signal modulus; The method for extracting the maximum value of the wavelet sequence modulus is as follows: If |Wf(2 m ,b)|≤|Wf(2 m ,b0)|, m=M-1,M, then the wavelet transform is 2 m The modulus maximum is obtained at point b0 under the scale; Step 2.3: Set the test threshold T1. When the maximum value of the modulus of the wavelet coefficient is greater than the test threshold T1, calculate the Lipschitz index α of the signal at the maximum value of the modulus to remove the singular points of the signal. When 2 M The wavelet coefficient at the scale modulus maximum exceeds the test threshold T1 and 2 M-1 When the wavelet coefficient at the corresponding time under the scale is not 0, calculate the Lipschitz index of the signal. For 2 M-1 , 2 M scale, the Lipschitz index α at point b0 is calculated as follows If the Lipschitz index of the signal at the modulus maximum is less than 0, then the value of this point is equal to the sampling value of the previous point, and the value of the signal after 2 points is equal to the sampling value of the previous point. M -No more singularity detection is performed within 1 sampling period.

3. A turbocharger centrifugal compressor surge online detection method as claimed in claim 2, characterized in that: The method for setting the test threshold T1 in step 2.3 is as follows: When testing the performance of a turbocharger compressor, select the signal at any speed among the measured speeds. When a singular point appears, M The maximum value of the wavelet coefficient W1 under steady-state conditions and the wavelet coefficient W2 at the time when the singular point appears are taken from the wavelet coefficients under the scale. The wavelet threshold is the average of the two, that is, the test threshold T1 is determined. The calculation formula is as follows in, Where P is the steady-state condition2 M The number of wavelet coefficients at the scale, W2=Wf(2 M ,b s ) (4) Where b s is the moment when the singular point of the sampled signal appears.

4. A turbocharger centrifugal compressor surge online detection method as claimed in claim 1, characterized in that: Step 4 is implemented as follows: The calculation formula is as follows Where N is the number of sampling points in the time window, p i is the sampling value at other moments in the time window; p N is the latest sampling value in the time window; When the pressure signal p'(t) is disturbed by noise n(t) during the sampling process, the discrete expression of the signal obtained after the sampling system in the time domain is: p i =p′ i +n i (6) but Another way to express it is in, is the mean value of the pressure sampling data without noise interference, is the mean value of the pressure signal data collected by the sampling system, is the noise mean; The first term on the right side of the equal sign in formula (7) represents the fluctuation of the compressor outlet pressure signal; the second term is the covariance between the compressor outlet pressure signal and the noise signal, reflecting the degree of linearity between the two random variables, and this value is considered to be equal to 0; the third term is caused by noise and is the variance of the noise.

5. The method for online detection of turbocharger centrifugal compressor surge according to claim 1, characterized in that: In step 5, the judgment threshold obtained based on the lowest speed is determined by the maximum value S1 of the mean sum of square differences between the latest sampling point and other sampling points in the steady-state operating time window and the maximum value S2 of the mean sum of square differences between the latest sampling point and other sampling points in the first surge cycle time window.

Images