On-line detection self-adaptive threshold noise reduction method for gliding oil metal chips in high-noise environment

Abstract

The invention belongs to the technical field of health management of aero-engine mechanical systems, and particularly relates to an online detection self-adaptive threshold noise reduction method for glide oil metal chips in a high-noise environment, and the method comprises the steps: carrying out the IIR low-pass filtering of all sampling points through an IIR digital filter, and obtaining the X-axis data and Y-axis data in a metal chip signal; amplitude values and phases of X-axis data and Y-axis data in the metal chip signals are obtained through trigonometric function transformation; and a search threshold is obtained by adopting a self-adaptive threshold waveform search algorithm, so that the search threshold is greater than the intensity of vibration noise in the metal chip signal, and the vibration noise is shielded. Through multi-stage signal processing and dynamic threshold optimization, the problems of missing detection and false detection of a traditional fixed threshold method in a high-noise scene are solved in a targeted manner, and the detection reliability and practicability are remarkably improved. By adopting an adaptive threshold noise reduction method, high-noise interference signals can be effectively filtered out, and the problems of missed alarm, false alarm and the like of a detection system are reduced.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of health management of aero-engine mechanical systems, and particularly relates to a method for online detection of oil metal dust in a high-noise environment and adaptive threshold noise reduction. BACKGROUND

[0002] Due to limitations of design and manufacturing technology, bearings of an aero-engine have always been weak links. By installing an oil dust sensor in an oil system pipeline, important parameters such as properties, quantity, and size of metal dust passing through the pipeline can be monitored in real time, early warning can be performed in the early stage of bearing failure, and a trend graph of cumulative increase of oil metal dust over time can be drawn. In a laboratory or other good working conditions, the detection accuracy of online oil metal dust detection is high. However, when the detection is performed on an aero-engine, false detection problems are prone to occur due to vibration interference, resulting in false alarms.

[0003] At present, domestic online oil metal dust detection generally uses a fixed threshold to filter out background noise caused by vibration in order to eliminate the vibration interference factor of the whole machine, which has the following disadvantages:

[0004] (1) When the fixed threshold is set to be large, real small particle signals in the oil metal dust are prone to be filtered out, resulting in missed detection.

[0005] (2) When the fixed threshold is set to be small, high-noise signals are prone to interfere with system acquisition, resulting in false detection.

[0006] Therefore, how to improve the accuracy of sensor particle detection is a problem to be solved. SUMMARY

[0007] To solve the above problems, the application provides a method for online detection of oil metal dust in a high-noise environment and adaptive threshold noise reduction, so as to solve the problems of missed detection or false detection of sensor particle detection in the prior art.

[0008] The technical scheme of the application is a method for online detection of oil metal dust in a high-noise environment and adaptive threshold noise reduction, comprising:

[0009] Collecting original data, calculating the average value of a certain number of collected data in each selection of original data as the sampling point of this time by mean filtering, and obtaining all sampling points in the original data;

[0010] Performing IIR low-pass filtering on all sampling points by an IIR digital filter to obtain X-axis data (V x ) and Y-axis data (V y ) in the metal dust signal;

[0011] X-axis data (V) from the metal chip signal is obtained through trigonometric function transformation. x ) and Y-axis data (V y The amplitude and phase of )

[0012] An adaptive threshold waveform search algorithm is used to obtain the search threshold, which is made to be greater than the intensity of vibration noise in the metal shavings signal, thus shielding the vibration noise.

[0013] Preferably, the average value of the collected data for:

[0014] ;

[0015] Where n is the number of selections each time.

[0016] Preferably, the IIR digital filter is represented by an nth-order difference equation. This represents the output after IIR filtering. The input signal is represented as:

[0017] ;

[0018] IIR low-pass filtering uses a first-order filter, and the calculation formula is:

[0019] ;

[0020] In the formula, and All of these are sampling coefficients.

[0021] Preferably, the X-axis data (V) in the metal chip signal x ) and Y-axis data (V y The amplitude and phase of ) are:

[0022] ;

[0023] Where V is the calculated amplitude. The calculated phase is represented by the subscripts x and y, which indicate the directions, respectively.

[0024] Preferably, the search threshold for:

[0025] ;

[0026] Where k is the threshold coefficient, v is the amplitude of the acquired signal, and 0-N is the number of sampling points. It is the root mean square.

[0027] Preferably, the root mean square (RMS) increases when the noise signal is strong and decreases when the noise signal is weak.

[0028] Preferably, if the vibration noise is a sinusoidal signal, the threshold coefficient is greater than or equal to 1.414; if the vibration noise is a square wave signal with a 10% duty cycle, the threshold coefficient is greater than or equal to 3.16.

[0029] The adaptive threshold noise reduction method for online detection of sliding metal chips in high-noise environments presented in this application has the following advantages:

[0030] Through multi-level signal processing and dynamic threshold optimization, the problem of missed detection and false detection in high-noise scenarios of traditional fixed threshold method is specifically solved, which significantly improves the reliability and practicality of detection.

[0031] The adaptive threshold noise reduction method can effectively filter out high noise interference signals and reduce the problems of missed alarms and false alarms in the detection system.

[0032] Through technological innovation of "multi-level noise reduction preprocessing → feature enhancement → dynamic adaptive threshold", the problems of "missing small particle signals due to excessively large fixed threshold" and "falsely detecting noise signals due to excessively small fixed threshold" have been systematically solved. Attached Figure Description

[0033] Figure 1 This is a schematic diagram of the lubricating oil metal shavings signal processing process in this application;

[0034] Figure 2 This is a schematic diagram illustrating the variation of lubricating oil metal chip signal noise with rotational speed in this application;

[0035] Figure 3 This is a schematic diagram of the overall process of this application;

[0036] Figure 4 This is a schematic diagram of the sinusoidal vibration noise signal waveform of this application;

[0037] Figure 5 This is a schematic diagram of the square wave vibration noise signal waveform with a 10% duty cycle according to this application. Detailed Implementation

[0038] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions in the embodiments of this application will be described in more detail below with reference to the accompanying drawings. In the drawings, the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The described embodiments are only some, not all, of the embodiments of this application. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application. The embodiments of this application will be described in detail below with reference to the accompanying drawings.

[0039] The lubricating oil metal shavings sensor employs a three-screw coil structure, identifying metal particles in lubricating oil through electromagnetic induction and eddy current principles. A high-frequency, high-voltage excitation signal is applied to both excitation coils, and a modulated induction signal is obtained from the secondary induction coil. This signal is then demodulated and filtered to acquire the metal shavings signal. Finally, an identification algorithm distinguishes between ferromagnetic and non-ferromagnetic particle signals, filtering out interference signals or false particles. The overall structure is as follows: Figure 1 As shown.

[0040] Mechanical vibration is one of the main causes of false signals from lubricating oil metal shavings. Figure 2 This shows the typical variation of lubricating oil metal shavings noise with engine speed during engine testing.

[0041] The lubricating oil metal shavings signal noise is related to engine speed, which is considered to be caused by changes in the environmental factors of the measurement system during engine operation. Considering that the environmental vibration generated by engine operation is also closely related to engine speed, the lubricating oil metal shavings signal noise is considered to be caused by system environmental vibration.

[0042] During engine operation, the contact resistance of the connectors between circuits changes under the influence of environmental vibration. If the contact resistance of the connectors that provide sensor excitation or feedback signals changes, it will modulate the signal. In addition, the contact noise caused by connector rubbing and slight displacement will also introduce interference signals. Furthermore, metal shavings also modulate the carrier signal. Therefore, the change in the contact resistance of the connectors caused by vibration will cause false alarm signals due to lubricating oil metal shavings.

[0043] Meanwhile, in vibrating environments, cable jitter can also cause false signals from lubricating oil and metal shavings. To reduce system size, existing sensors typically integrate excitation and feedback signals within a single connector. In many cases, the excitation and feedback signals are also within the same cable bundle, resulting in a close spatial distance between them. This causes crosstalk in the feedback signal due to the high voltage and high current excitation signal lines.

[0044] If the excitation signal wire and the feedback signal wire change relative displacement due to environmental vibration, the change in crosstalk signal amplitude will generate additional noise after demodulation.

[0045] Based on this, the first aspect of this application provides an adaptive threshold noise reduction method for online detection of sliding metal chips in high-noise environments, such as... Figure 3 It includes the following steps:

[0046] Step S100: Collect raw data, and calculate the average value of a certain number of collected data points in the raw data each time by means filtering, and use it as the sampling point for that time until all sampling points in the raw data are obtained.

[0047] In practical lubricating oil metal shavings signal processing, hardware limitations can add noise to the input signal, causing waveform errors, and the noise can even overwhelm the signal. Therefore, digital filtering is necessary in metal shavings signal processing. This system performs mean filtering and low-pass filtering on the acquired raw data sequentially.

[0048] The logic of mean filtering is to collect n numbers, take the average of the n numbers as the current sampling point, and reduce the sampling rate from A to A / n. The purpose of mean filtering is to reduce the impact of interference on signal glitches. Mean filtering is commonly used in engineering to reduce the impact of dynamic interference. However, its disadvantage is that some obviously unreasonable data (such as sudden signal changes caused by interference during the process) are also averaged in the result, reducing the accuracy of the data. Since the AD sampling rate is 200k / s, and considering the probability of lubricating oil metal shavings appearing, a value of n of 8 is selected, and the actual sampling rate is reduced by 1 / 8.

[0049] Average value of collected data for:

[0050] ;

[0051] Where n is the number of selections each time.

[0052] Step S200: Perform IIR low-pass filtering on all sampling points using an IIR digital filter to obtain the X-axis data (Vx) and Y-axis data (Vy) in the metal shavings signal.

[0053] This involves modifying certain frequency bands of a signal. When processing a signal, the desired signal is the one carrying the information, while noise within the useful signal should be filtered out as much as possible. Furthermore, the frequencies of different signals are different. Selective filters are typically used to filter the signal. The purpose of using filters is to preserve certain frequency bands of the signal without distortion, while minimizing or even completely eliminating unwanted frequency bands. Filters can be classified as low-pass filters, high-pass filters, and band-pass filters, etc. This application uses a low-pass filter.

[0054] Digital filtering is a type of programmed filtering that removes noise from useful signals, improves the signal-to-noise ratio, reduces random errors, and helps improve signal quality and measurement accuracy. Based on the time-domain characteristics of discrete systems, digital filters can be divided into FIR filters and IIR filters. Filters with piecewise constant frequency characteristics are typically designed using IIR filters, which, compared to FIR filters, have a lower order for the same specifications. Furthermore, for convenient and quick design, IIR filters are generally implemented using analog filters. The transformation from analog to digital filters can utilize the impulse invariance method and the bilinear transform method. This application primarily employs the bilinear transform method to implement IIR filtering.

[0055] An IIR digital filter can be represented by an nth-order difference equation, as detailed in the following equation, where... This represents the output after IIR filtering. This indicates the input signal.

[0056] ;

[0057] In the formula, and All of these are sampling coefficients.

[0058] Or the Z-domain function, see the following formula:

[0059] ;

[0060] After mean filtering, IIR low-pass filtering is performed. IIR low-pass filtering uses a first-order filter, and its calculation formula is as follows:

[0061] .

[0062] Step S300: Obtain the amplitude and phase of the X-axis data (Vx) and Y-axis data (Vy) in the metal chip signal through trigonometric function transformation.

[0063] After mean filtering and low-pass filtering, the data consists of two sets: X-axis data (Vx) and Y-axis data (Vy). The amplitude and phase of the metal chip signal are obtained through trigonometric function transformation, as shown in the following formula.

[0064] ;

[0065] Where V is the calculated amplitude. The calculated phase is represented by the subscripts x and y, which indicate the directions, respectively.

[0066] Step S400: An adaptive threshold waveform search algorithm is used to obtain a search threshold, which is made to be greater than the intensity of vibration noise in the metal shavings signal, thereby shielding the vibration noise.

[0067] Based on the principle of lubricating oil metal shavings sensors, ferromagnetic particles and non-ferromagnetic particles have a phase difference of approximately 90°. During hardware demodulation, the phase of the demodulated signal is adjusted to make the ferromagnetic particle signal... 45°, non-ferromagnetic particles The phase range is -45°. Since the phase of vibration interference is random and differs significantly from the phase of the metal shavings signal, the phase range of ferromagnetic particles is set to 45±10°, and the phase range of non-ferromagnetic particles is set to -45±10°. This can serve as a phase filter and suppress vibration interference to a certain extent.

[0068] The lubricating oil metal shavings signal is generated by abrasive particles in the lubricating oil passing through the sensor. The timing, location, and size of these particles are highly random, resulting in a small sample size for statistical analysis. Vibration interference, on the other hand, is generated by the engine's operating state, and its intensity varies with engine conditions. Vibration interference is strong under high-intensity conditions and weak under slow-moving conditions, but overall it is a constant quantity over a short period, constituting background noise and a continuous signal. Therefore, the randomness of the metal shavings signal and the continuity of vibration interference can be utilized to suppress vibration interference through adaptive threshold noise reduction, thereby reducing false alarms.

[0069] Threshold calculation involves calculating the root mean square (RMS) of the signal acquisition values ​​within the processing cycle, and then generating an adaptive threshold using a threshold coefficient. This adaptive threshold is used for waveform lookup. Since the threshold is related to the RMS of the signal acquisition values ​​within the processing cycle, it is significantly affected by background noise. Higher background noise results in a higher threshold, and lower background noise results in a lower threshold, thus reducing false alarms. Due to engine vibration, the metal shavings signal contains a large amount of continuous vibration interference. This interference is mostly low-frequency pulse signals with overlapping bandwidths and similar waveforms to the metal shavings signal. To eliminate the influence of vibration interference, an adaptive threshold waveform lookup algorithm is used to suppress it.

[0070] When searching for the waveform of metal shavings, the magnitude of the vibration noise is first calculated. Then, the effective value of the vibration noise is multiplied by a threshold coefficient to obtain the search threshold. This search threshold is made to be greater than the intensity of the vibration noise, thereby shielding the influence of the vibration noise. Since the vibration noise is related to the vibration characteristics of the engine, the threshold coefficient needs to be experimentally set on a test bench.

[0071] Preferably, the search threshold for:

[0072] ;

[0073] Where k is the threshold coefficient, v is the amplitude of the acquired signal, and 0-N is the number of sampling points. It is the root mean square.

[0074] When the noise signal is strong, the root mean square (RMS) will increase; when the noise signal is weak, the RMS will decrease. Different threshold systems are set according to different waveforms.

[0075] Different vibration noise characteristics result in different effective values, thus requiring different threshold coefficients. For example, if the vibration noise is a sinusoidal signal, the threshold coefficient must be at least 1.414 for the search threshold to exceed the vibration noise intensity. Figure 4 If the vibration noise is a square wave signal with a 10% duty cycle, the threshold coefficient must be at least 3.16 for the search threshold to exceed the vibration noise intensity. Figure 5 .

[0076] Based on the bench test results, the vibration and noise during engine operation are mostly low-frequency pulse signals with waveforms between sinusoidal signals and low duty cycle square wave signals. Therefore, the initial value of the threshold coefficient can be set within the range of (1.414, 3.16).

[0077] In summary, this application has the following advantages:

[0078] Through multi-level signal processing and dynamic threshold optimization, the problem of missed detection and false detection in high-noise scenarios of traditional fixed threshold method is specifically solved, which significantly improves the reliability and practicality of detection.

[0079] The adaptive threshold noise reduction method can effectively filter out high noise interference signals and reduce the problems of missed alarms and false alarms in the detection system.

[0080] Through technological innovation of "multi-level noise reduction preprocessing → feature enhancement → dynamic adaptive threshold", the problems of "missing small particle signals due to excessively large fixed threshold" and "falsely detecting noise signals due to excessively small fixed threshold" have been systematically solved.

[0081] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. An adaptive threshold noise reduction method for online detection of sliding metal debris in high-noise environments, characterized in that, include: Collect raw data, and calculate the average value of a certain number of collected data points in the raw data each time by mean filtering, and use it as the sampling point for that time until all sampling points in the raw data are obtained. IIR low-pass filtering is applied to all sampling points using an IIR digital filter to obtain the X-axis data (V) in the metal chip signal. x ) and Y-axis data (V y ); X-axis data (V) from the metal chip signal is obtained through trigonometric function transformation. x ) and Y-axis data (V y The amplitude and phase of ) An adaptive threshold waveform search algorithm is used to obtain the search threshold, which is made to be greater than the intensity of vibration noise in the metal shavings signal, thus shielding the vibration noise.

2. The adaptive threshold noise reduction method for online detection of sliding metal debris in high-noise environments as described in claim 1, characterized in that, The average value of the collected data for: ； Where n is the number of selections each time.

3. The adaptive threshold noise reduction method for online detection of sliding metal debris in high-noise environments as described in claim 2, characterized in that, The IIR digital filter is represented by an nth-order difference equation. This represents the output after IIR filtering. The input signal is represented as: ； IIR low-pass filtering uses a first-order filter, and the calculation formula is: ； In the formula, and All of these are sampling coefficients.

4. The adaptive threshold noise reduction method for online detection of sliding metal debris in high-noise environments as described in claim 3, characterized in that, The X-axis data (V) in the metal chip signal x ) and Y-axis data (V y The amplitude and phase of ) are: ； Where V is the calculated amplitude. The calculated phase is represented by the subscripts x and y, which indicate the directions, respectively.

5. The adaptive threshold noise reduction method for online detection of oil-slip metal debris in high-noise environments as described in claim 4, characterized in that, The search threshold for: ； Where k is the threshold coefficient, v is the amplitude of the acquired signal, and 0-N is the number of sampling points. It is the root mean square.

6. The adaptive threshold noise reduction method for online detection of sliding metal debris in high-noise environments as described in claim 5, characterized in that, When the noise signal is strong, the root mean square (RMS) will increase; when the noise signal is weak, the RMS will decrease.

7. The adaptive threshold noise reduction method for online detection of oil-slip metal debris in high-noise environments as described in claim 5, characterized in that, If the vibration noise is a sinusoidal signal, the threshold coefficient is greater than or equal to 1.414; if the vibration noise is a square wave signal with a 10% duty cycle, the threshold coefficient is greater than or equal to 3.16.

Images