CN109740535B - Reciprocating motion signal extraction method based on multi-observer likelihood ratio detection method - Google Patents
Reciprocating motion signal extraction method based on multi-observer likelihood ratio detection method Download PDFInfo
- Publication number
- CN109740535B CN109740535B CN201910000343.0A CN201910000343A CN109740535B CN 109740535 B CN109740535 B CN 109740535B CN 201910000343 A CN201910000343 A CN 201910000343A CN 109740535 B CN109740535 B CN 109740535B
- Authority
- CN
- China
- Prior art keywords
- signal
- noise
- reciprocating motion
- frame
- reciprocating
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Abstract
The invention discloses a reciprocating motion signal extraction method based on a multi-observer likelihood ratio detection method, and belongs to the technical field of rotating machinery state monitoring. The invention adopts a multi-observer likelihood ratio detection method, does not need to install a key phase device, can find out the signal period consistent with actual reciprocation only by analyzing the test signal, and extracts the reciprocating motion signal. Experiments have been carried out on the industrial robot, and the experiments show that the method has high precision and stability and can accurately detect the reciprocating motion of the robot.
Description
Technical Field
The invention relates to a reciprocating motion signal extraction method based on a multi-observer likelihood ratio detection method, and belongs to the technical field of rotating machinery state monitoring.
Background
With the continuous improvement of the industrial automation degree, the joint type industrial robot has very wide application in various fields of automobile and automobile part manufacturing industry, machining industry, electronic and electrical industry and the like. The heart of the industrial robot transmission system is the RV reducer, accounting for over 1/3 of its total cost. As a main rotating member in an industrial robot, reliability gradually decreases and a failure rate increases as a service time increases. The assembly line is interrupted when the robot stops working, and the robot needs to be transferred away from the station when replacing the speed reducer, so that a large amount of time is consumed, and great economic loss is brought to enterprises. Because the robot has a large amplitude movement in the working process and causes structural vibration, the vibration intensity is far greater than the vibration caused by internal faults, and the frequency ranges of the robot and the structural vibration are close to each other, and fault signals are difficult to obtain. Therefore, it is necessary to separate a signal corresponding to the movement of the robot from continuously recorded data in a complete cycle, remove noise corresponding to the standstill, and obtain a signal corresponding to the reciprocating oscillation of the robot, so as to more accurately identify the failure of the robot rotary mechanical system.
Disclosure of Invention
The invention provides a reciprocating motion signal extraction method based on a multi-observer likelihood ratio detection method, which is used for realizing reciprocating motion signal extraction.
The technical scheme of the invention is as follows: a reciprocating motion signal extraction method based on a multi-observer likelihood ratio detection method comprises the following steps:
step1, dividing the test signal into M sub-frames, and dividing one of the M sub-frames into xlThe binary assumption is made for the reciprocating signal or noise:in the formula Sl、N、XlAre respectively signal framing xlDiscrete Fourier transform coefficient vector of contained motion signal, discrete Fourier transform coefficient vector of noise, signal framing xlThe length of each discrete Fourier transform coefficient vector is L, and the (k + 1) th elements in the three coefficient vectors are respectivelyNkAndl=1,2,3...M;k=0,1,2...L-1;
step2, dividing the signal into frames xlPerforming discrete Fourier transform to obtain XlAnd XlK elements of (a); taking the signal with the lowest energy as noise, carrying out discrete Fourier transform to obtain N,and k elements of N, and calculating the k +1 th element N of NkVariance λ ofN(k);
suppose thatBelonging to reciprocating motion signals, conditional probability density functions thereofExpressed as:
step4, obtaining the superposition relationship between the noise and the motion signalIs estimated asBy usingSubstitution formulaIn (1)Obtaining:
step5 based onLikelihood ratio ofObtaining signal frames xlGeometric mean of likelihood ratios across frequency bandsAnd defining a discrimination signal frame xlThe rules pertaining to noise or reciprocating signals are:by introducing and framing xlAdjacent 2m subframes { xl-m,...xl,...,xl+mImproving the quality of discrimination to obtain a discrimination function of the multi-observer likelihood ratio detection method comprising a plurality of sub-frames:in the formula: eta is a discrimination threshold value, and l represents the signal framing sequence number currently being analyzed;
step6, substituting the optimal discriminant threshold into the discriminant functionEach frame of the signal is classified into a reciprocating motion signal and noise according to a discriminant function, and the frames belonging to the reciprocating motion signal are extracted, i.e., the reciprocating motion signal is extracted from the test signal.
In Step6, the optimal discrimination threshold value obtaining method is as follows:
1) equally dividing the test signal into M sub-frames, wherein one signal sub-frame is xl,l∈[1,...,M];
2) Calculating the periodicity of the reciprocating motion: for a discrimination threshold η, frame xlBy discriminant functionTo judge that it belongs to the reciprocating motion signal H1Or noise H0(ii) a If each reciprocation period includes Z sub-frames, if frame xlIs H1Then frame xl+zShould also belong to H1(ii) a The periodicity of the reciprocating motion signal can be defined as:in the formula (I), the compound is shown in the specification,q is of signal H1The number of frames;
3) maximization of calculationSignal-to-noise ratio of the classified reciprocating motion signal to noise: for a discrimination threshold η, the signal-to-noise ratio is defined as the ratio of the power of the reciprocating signal frame to the power of the noise frame: in the formulaFor a set of frame-wise fractions of the reciprocating signal,for a noise framing set, rms represents the root mean square value of the computed signal;
4) combining the periodicity of the reciprocating motion and the signal-to-noise ratio of the reciprocating motion signal and the noise after maximum classification, and defining an optimized objective function of a discrimination threshold as follows:search for optimal discrimination thresholds over a range, i.e.And the eta corresponding to the maximum time is the optimal discrimination threshold.
The invention has the beneficial effects that: the invention adopts a multi-observer likelihood ratio detection method, does not need to install a key phase device, can find out the signal period consistent with actual reciprocation only by analyzing the test signal, and extracts the reciprocating motion signal. Experiments have been carried out on the industrial robot, and the experiments show that the method has high precision and stability and can accurately detect the reciprocating motion of the robot.
Drawings
FIG. 1 is a flow chart of the method steps of the present invention;
FIG. 2 is an example of a test signal for a reciprocating motion of a robot;
FIG. 3 is a relationship and fit curve of an objective function and an optimized objective function for a discrimination threshold;
FIG. 4 is a discriminant function of a measurement signal;
fig. 5 is a graph of experimental reciprocation signals versus noise classification results.
Detailed Description
The invention will be further described with reference to the following figures and examples, without however restricting the scope of the invention thereto.
In the running process of the reciprocating motion equipment, sensors such as vibration, sound and acoustic emission are used for picking up running state signals, and the running state signals are converted into digital signals through data acquisition equipment. The invention distinguishes the reciprocating motion signal and the noise in the test signal through analyzing the test signal to obtain the motion period signal in the signal, thereby laying a foundation for further deep analysis. The reciprocating period is only the noise part + the overlapping part of the noise and the action signal, and the overlapping part of the noise and the action signal is the reciprocating signal
Example 1: as shown in fig. 1 to 5, a reciprocating motion signal extraction method based on the multi-observer likelihood ratio detection method, the present example illustrates the implementation of the present invention with an actual test signal on the welding robot joint of a certain automobile factory:
step1, dividing the test signal into 469 sub-frames with length of 0.0256s, and dividing one signal into x sub-frameslThe binary assumption is made for the reciprocating signal or noise:in the formula Sl、N、XlAre respectively signal framing xlDiscrete Fourier transform coefficient vector of motion signal, discrete Fourier transform coefficient vector of noise, and signal frame xlThe length of each discrete Fourier transform coefficient vector is L, and the (k + 1) th elements in the three coefficient vectors are respectivelyNkAnd1,2,3.. 469; l-1, 0,1, 2; the test signal is shown in fig. 2 (time (seconds) on the abscissa and amplitude (mv) on the ordinate).
Step2, dividing the signal into frames xlPerforming discrete Fourier transform to obtain XlAnd XlK elements of (a); taking the signal frame with the lowest energy as noise, carrying out discrete Fourier transform on the signal frame to obtain N and k elements of the N, and calculating the k +1 th element N of the NkVariance λ ofN(k);
suppose thatBelonging to reciprocating motion signals, conditional probability density functions thereofExpressed as:
step4, obtaining the superposition relationship between the noise and the motion signalIs estimated asBy usingSubstitution formulaIn (1)Obtaining:obtained in step2Substitution intoCan obtain
Step5 based onLikelihood ratio ofObtaining signal frames xlGeometric mean of likelihood ratios in the 0 to L-1 frequency bandsAnd defines the frame x of the discrimination signallThe rules pertaining to noise or reciprocating signals are:by introducing and framing xlAdjacent 2m subframes { xl-m,...xl,...,xl+mImproving the quality of discrimination to obtain a discrimination function of the multi-observer likelihood ratio detection method comprising a plurality of sub-frames:in the formula: eta is a discrimination threshold value, and l represents the signal framing sequence number currently being analyzed;
step6, obtaining an optimal discrimination threshold value:
1) equally dividing the test signal into M sub-frames, wherein one signal sub-frame is xl,l∈[1,...,M];
2) Calculating the periodicity of the reciprocating motion: for a discrimination threshold η, frame xlBy discriminant functionTo judge that it belongs to the reciprocating motion signal H1Or noise H0(ii) a If each reciprocation period includes Z sub-frames, if frame xlIs H1Then frame xl+ZShould also belong to H1(ii) a The periodicity of the reciprocating motion signal can be defined as:in the formula (I), the compound is shown in the specification,q is of signal H1The number of frames;
3) calculating the signal-to-noise ratio of the reciprocating motion signal and the noise after maximum classification: for a discrimination threshold η, the signal-to-noise ratio is defined as the ratio of the power of the reciprocating signal frame to the power of the noise frame: in the formulaFor a set of frame-wise fractions of the reciprocating signal,for a noise framing set, rms represents the root mean square value of the computed signal;
4) combining the periodicity of the reciprocating motion and the signal-to-noise ratio of the reciprocating motion signal and the noise after maximum classification, and defining an optimized objective function of a discrimination threshold as follows:setting the search range of the discriminant threshold (1 ≦ eta ≦ 14) as the bottom portion of the discriminant function, calculating the target functionAnd (3) performing 4-order curve fitting on the data, and finding out a value corresponding to the highest point on a fitting curve, namely 3.3 serving as an optimal discrimination threshold eta. The results are shown in FIG. 3, where the discontinuities are the correspondences calculated for different values of ηThe continuous curve is a fourth order polynomial fitted curve of the discontinuity point.
The discrimination threshold η is substituted into Step5 by 3.3 to obtain a discrimination functionObtaining the final discriminant functionThe calculated discriminant function is shown in fig. 4.
The discriminant function is substituted into the test signal, and each frame of the test signal is classified into a noise or a reciprocating signal, as shown in fig. 5, a dotted line indicates a classification result, corresponding to the right coordinate axis, 0 is a noise, and 1 is a portion where the noise overlaps with the motion signal, i.e., the reciprocating signal. The solid line represents the test signal. It can be seen that the higher intensity motion signals can be separated from the noise.
While the present invention has been described in detail with reference to the embodiments shown in the drawings, the present invention is not limited to the embodiments, and various changes can be made without departing from the spirit of the present invention within the knowledge of those skilled in the art.
Claims (2)
1. A reciprocating motion signal extraction method based on a multi-observer likelihood ratio detection method is characterized by comprising the following steps: the method comprises the following steps:
step1, dividing the test signal into M sub-frames, and dividing one of the M sub-frames into xlThe binary assumption is made for the reciprocating signal or noise:in the formula Sl、N、XlAre respectively signal framing xlDiscrete Fourier transform coefficient vector of motion signal, discrete Fourier transform coefficient vector of noise, and signal frame xlDiscrete Fourier ofThe transform coefficient vectors are all L in length, and the k +1 th elements in the three coefficient vectors are respectivelyNkAndl=1,2,3...M;k=0,1,2...L-1;
step2, dividing the signal into frames xlPerforming discrete Fourier transform to obtain XlAnd XlK elements of (a); taking the signal frame with the lowest energy as noise, carrying out discrete Fourier transform on the signal frame to obtain N and k elements of the N, and calculating the k +1 th element N of the NkVariance λ ofN(k);
suppose thatBelonging to reciprocating motion signals, conditional probability density functions thereofExpressed as:
step4, obtaining the superposition relationship between the noise and the motion signalIs estimated asBy usingSubstitution formulaIn (1)Obtaining:
step5 based onLikelihood ratio ofObtaining signal frames xlGeometric mean of likelihood ratios across frequency bandsAnd defines the frame x of the discrimination signallThe rules pertaining to noise or reciprocating signals are:by introducing and framing xlAdjacent 2m subframes { xl-m,...xl,...,xl+mImproving the quality of discrimination to obtain a discrimination function of the multi-observer likelihood ratio detection method comprising a plurality of sub-frames:in the formula: eta is a discrimination threshold value, and l represents the signal framing sequence number currently being analyzed;
step6, substituting the optimal discriminant threshold into the discriminant functionEach frame of the signal is classified into a reciprocating signal and noise according to a discriminant function, and the frames belonging to the reciprocating signal are extracted, i.e., the reciprocating signal is extracted from the test signal.
2. The method of extracting a reciprocating motion signal based on the multi-observer likelihood ratio detection method according to claim 1, characterized in that: in Step6, the optimal discrimination threshold value obtaining method is as follows:
1) equally dividing the test signal into M sub-frames, wherein one signal sub-frame is xl,l∈[1,...,M];
2) Calculating the periodicity of the reciprocating motion: for a discrimination threshold η, frame xlBy discriminant functionTo judge that it belongs to the reciprocating motion signal H1Or noise H0(ii) a If each reciprocation period includes Z sub-frames, if frame xlIs H1Then frame xl+zShould also belong to H1(ii) a The periodicity of the reciprocating motion signal can be defined as:in the formula (I), the compound is shown in the specification,q is of signal H1The number of frames;
3) calculating the signal-to-noise ratio of the reciprocating motion signal and the noise after maximum classification: for a discrimination threshold η, the signal-to-noise ratio is defined as the ratio of the power of the reciprocating signal frame to the power of the noise frame: in the formulaFor a set of frame-wise fractions of the reciprocating signal,for noise framing, rms represents the mean square of the computed signalA root value;
4) combining the periodicity of the reciprocating motion and the signal-to-noise ratio of the reciprocating motion signal and the noise after maximum classification, and defining an optimized objective function of a discrimination threshold as follows:search for optimal discrimination thresholds over a range, i.e.And the eta corresponding to the maximum time is the optimal discrimination threshold.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910000343.0A CN109740535B (en) | 2019-01-02 | 2019-01-02 | Reciprocating motion signal extraction method based on multi-observer likelihood ratio detection method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910000343.0A CN109740535B (en) | 2019-01-02 | 2019-01-02 | Reciprocating motion signal extraction method based on multi-observer likelihood ratio detection method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109740535A CN109740535A (en) | 2019-05-10 |
CN109740535B true CN109740535B (en) | 2022-07-05 |
Family
ID=66363142
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910000343.0A Active CN109740535B (en) | 2019-01-02 | 2019-01-02 | Reciprocating motion signal extraction method based on multi-observer likelihood ratio detection method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109740535B (en) |
Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107478417A (en) * | 2017-06-23 | 2017-12-15 | 昆明理工大学 | A kind of reciprocation period extracting method based on frequency correction and principle of correlation analysis |
Family Cites Families (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP1192482A4 (en) * | 2000-05-08 | 2009-11-11 | Schlumberger Holdings | Digital signal receiver for measurement while drilling system having noise cancellation |
US20100106458A1 (en) * | 2008-10-28 | 2010-04-29 | Leu Ming C | Computer program and method for detecting and predicting valve failure in a reciprocating compressor |
-
2019
- 2019-01-02 CN CN201910000343.0A patent/CN109740535B/en active Active
Patent Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107478417A (en) * | 2017-06-23 | 2017-12-15 | 昆明理工大学 | A kind of reciprocation period extracting method based on frequency correction and principle of correlation analysis |
Non-Patent Citations (4)
Title |
---|
Impact analysis for Variable-Speed bearing fault detection;Xiaoqin Liu et al.;《Proceedings 2011 TMEE》;20120514;全文 * |
典型复杂微动目标的建模和检测;陈磊等;《电讯技术》;20161128(第11期);全文 * |
基于MLE阈值规则的小波特征提取技术在气阀故障诊断中的应用;魏中青等;《振动与冲击》;20110125(第01期);全文 * |
往复压缩机十字头滑块故障的小波包敏感特征提取研究;李震等;《噪声与振动控制》;20170831;全文 * |
Also Published As
Publication number | Publication date |
---|---|
CN109740535A (en) | 2019-05-10 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Xu et al. | An incorrect data detection method for big data cleaning of machinery condition monitoring | |
Cheng et al. | Adaptive multipoint optimal minimum entropy deconvolution adjusted and application to fault diagnosis of rolling element bearings | |
CN105628381B (en) | A kind of reciprocating compressor Method for Bearing Fault Diagnosis decomposed based on improved local mean value | |
CN102507230B (en) | Method for diagnosing fault of automatic ammunition supply and transportation device | |
Khan et al. | Automated bearing fault diagnosis using 2D analysis of vibration acceleration signals under variable speed conditions | |
CN112232366A (en) | Electrical equipment fault early warning method and system based on RFID monitoring | |
CN109190598B (en) | Rotating machinery monitoring data noise point detection method based on SES-LOF | |
Su et al. | Fault diagnosis of rotating machinery based on wavelet domain denoising and metric distance | |
CN111852837B (en) | Clustering-based reciprocating compressor self-adaptive fault monitoring method | |
CN113269169A (en) | Bearing fault detection method and device | |
Jiang et al. | A novel antibody population optimization based artificial immune system for rotating equipment anomaly detection | |
CN109934136B (en) | Rolling bearing fault diagnosis method based on Duffing vibrator and eigen mode component | |
CN113569990A (en) | Performance equipment fault diagnosis model construction method oriented to strong noise interference environment | |
CN107766780A (en) | Characteristics information extraction method when oil pumping system based on electric work figure is run | |
CN109740535B (en) | Reciprocating motion signal extraction method based on multi-observer likelihood ratio detection method | |
CN103674550A (en) | Real-time mixing denoising method for rolling bearing static monitoring signal | |
CN112177865B (en) | Method for solving marking noise and insufficient marks in fan fault detection | |
Xue et al. | Diagnosis method based on hidden Markov model and Weibull mixture model for mechanical faults of in-wheel motors | |
CN112100577A (en) | Long-range correlation-based equipment operation stability online monitoring method and system | |
Kahlen et al. | Improving Machine Learning Diagnostic Systems with Model-Based Data Augmentation―Part B: Application | |
CN116738367A (en) | Rolling bearing fault diagnosis method based on multi-scale map transducer | |
Pandarakone et al. | A study on machine learning and artificial intelligence methods in detecting the minor outer-raceway bearing fault | |
Talmoudi et al. | An iot-based failure prediction solution using machine sound data | |
Boukra et al. | Identifying new prognostic features for remaining useful life prediction | |
Yan et al. | Deep Transfer Learning Based Multi-source Heterogeneous data Fusion with Application to Cross-scenario Tool Wear monitoring |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |