CN116502043A - Finish rolling motor state analysis method based on isolated forest algorithm - Google Patents
Finish rolling motor state analysis method based on isolated forest algorithm Download PDFInfo
- Publication number
- CN116502043A CN116502043A CN202310472163.9A CN202310472163A CN116502043A CN 116502043 A CN116502043 A CN 116502043A CN 202310472163 A CN202310472163 A CN 202310472163A CN 116502043 A CN116502043 A CN 116502043A
- Authority
- CN
- China
- Prior art keywords
- data
- value
- waveform
- motor
- free end
- 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.)
- Pending
Links
- 238000005096 rolling process Methods 0.000 title claims abstract description 43
- 238000004458 analytical method Methods 0.000 title claims abstract description 22
- 230000002159 abnormal effect Effects 0.000 claims abstract description 28
- 238000012423 maintenance Methods 0.000 claims abstract description 9
- 238000007781 pre-processing Methods 0.000 claims abstract description 8
- 238000012549 training Methods 0.000 claims abstract description 7
- 230000001133 acceleration Effects 0.000 claims description 66
- 238000000034 method Methods 0.000 claims description 30
- 238000010606 normalization Methods 0.000 claims description 9
- 238000012545 processing Methods 0.000 claims description 7
- 238000005070 sampling Methods 0.000 claims description 6
- 230000001351 cycling effect Effects 0.000 claims description 3
- 238000004364 calculation method Methods 0.000 abstract description 5
- 238000001514 detection method Methods 0.000 description 4
- 230000008569 process Effects 0.000 description 4
- 238000001228 spectrum Methods 0.000 description 3
- 238000012952 Resampling Methods 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 238000011156 evaluation Methods 0.000 description 2
- 230000000737 periodic effect Effects 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 238000002372 labelling Methods 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000013450 outlier detection Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/10—Pre-processing; Data cleansing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/18—Complex mathematical operations for evaluating statistical data, e.g. average values, frequency distributions, probability functions, regression analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
- G06F18/2433—Single-class perspective, e.g. one-against-all classification; Novelty detection; Outlier detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N5/00—Computing arrangements using knowledge-based models
- G06N5/01—Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Computational Mathematics (AREA)
- Mathematical Analysis (AREA)
- Mathematical Optimization (AREA)
- Pure & Applied Mathematics (AREA)
- Computational Linguistics (AREA)
- Computing Systems (AREA)
- Operations Research (AREA)
- Probability & Statistics with Applications (AREA)
- Algebra (AREA)
- Databases & Information Systems (AREA)
- Measurement Of Mechanical Vibrations Or Ultrasonic Waves (AREA)
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
Abstract
A finish rolling motor state analysis method based on an isolated forest algorithm comprises the following steps: preprocessing motor vibration history data; step b: constructing a sample set, performing isolated forest algorithm training, and calculating abnormal score values of all data; step c: identifying the state of the motor, and judging that the motor is in an abnormal state if the abnormal score value is greater than the state threshold value; otherwise, judging that the motor is in a normal state; step d: and feeding back a result, and feeding back the working state of the motor to an operator and a technical maintenance staff. The invention uses the unsupervised isolated forest algorithm to cluster the sample data without labels, the clustering result is more objective and accurate, the algorithm complexity is low, and the calculation speed is high. The invention can accurately and efficiently perform state analysis on the finish rolling motor and provide reliable information for correct maintenance of the motor.
Description
Technical Field
The invention relates to the technical field of motors, in particular to a motor running state analysis method.
Background
The motor is an important power driving device in the finishing mill group, monitors the running state (including normal state and abnormal state) and analyzes the dynamic performance, so that unnecessary loss in the operation process is avoided as much as possible, and the motor is a research subject which is needed urgently at present. Prior to mining finish rolling motor monitoring information, prior knowledge of the data is generally not available, which also makes motor state analysis very difficult.
At present, the analysis method of the running state of the motor mainly comprises the following steps:
(1) The motor running state is manually evaluated according to the actual working state of the finishing mill, and the method has the defects that professional and experienced technicians are needed, the evaluation result has certain subjectivity, and the working efficiency is low;
(2) The motor sampling data is trained and classified by using a supervised learning algorithm, the method needs to use marked data, and the marking process also needs to be manually judged and has subjectivity.
(3) The motor data is subjected to unsupervised clustering training by using an SOM network, and the motor state is analyzed, so that the SOM algorithm is complex, the calculation speed is low, local optimum is easily trapped, and the convergence speed is low or even is not converged.
Chinese patent application No. CN2019112291087, entitled "method for detecting periodic vibration impact signal during rotational speed change of hydraulic generator", discloses a method, which comprises the following steps: collecting initial vibration signal data; calculating an optimal band-pass filter according to the kurtosis of the fast envelope spectrum; performing band-pass filtering by adopting an optimal band-pass filter to obtain a vibration impact signal waveform; solving to obtain a vibration impact envelope of the vibration impact signal waveform; calculating a time interval between the plurality of pulses; estimating a polynomial rotation speed fitting coefficient; resampling the variable frequency of the vibration impact envelope to obtain an optimal value of the rotation speed fitting coefficient; constructing a fitting rotating speed polynomial, and resampling variable frequency of the vibration impact envelope to obtain a vibration impact envelope waveform sampled in a whole period; performing fast Fourier transform on the vibration impact envelope waveform sampled in the whole period to obtain a frequency spectrum; and carrying out fault analysis and evaluation according to the acquired frequency spectrum and the fitting rotating speed. However, the patent is directed to a hydro-generator whose rotational speed is relatively slow and stable, and this analysis is not applicable to rolling mill motors. In addition, the patent adopts a successive approximation mode to solve a polynomial fitting rotation speed change function, then resamples according to a time-varying rotation speed, and obtains the identification and detection of periodic vibration impact signals through the transformation and analysis of the vibration signals in the rotation speed varying process, and the method is mainly used for judging faults such as cracks of structural parts. In many cases, however, it is desirable to accurately and efficiently monitor and analyze the operation state of the finish rolling motor in real time to provide reliable information for proper maintenance and service of the motor.
Therefore, how to accurately and efficiently perform the state analysis on the finish rolling motor becomes a difficult problem for the personnel concerned.
Disclosure of Invention
The invention aims at solving the defects of the prior art and provides a finish rolling motor state analysis method based on an isolated forest algorithm, so as to accurately and efficiently analyze the state of the finish rolling motor and provide useful information for the correct maintenance and the service of the motor.
The invention aims at realizing the following technical scheme:
a finish rolling motor state analysis method based on an isolated forest algorithm, the method comprising the steps of:
a: preprocessing of motor vibration history data
The motor vibration waveform historical data collected according to a certain time interval is integrated, and the following six vibration data are obtained: the method comprises the steps of performing standardized processing on vibration data of each column, wherein the vibration data comprise a free end high-frequency acceleration total value, a free end low-frequency acceleration total value, a free end speed total value, a load end high-frequency acceleration total value, a load end low-frequency acceleration total value and a load end speed total value;
step b: constructing a sample set, performing isolated forest algorithm training, and calculating abnormal score values of all data
(1) Taking a group of vibration data which are collected simultaneously and are preprocessed as a sample to obtain a sample setN is the number of samples, +.>Represents the i-th sample, i=1, 2, …, n,wherein->、/>、/>、/>、/>、/>The method comprises the steps of respectively obtaining a free end high-frequency acceleration total value, a free end low-frequency acceleration total value, a free end speed total value, a load end high-frequency acceleration total value, a load end low-frequency acceleration total value and a load end speed total value in an ith sample;
(2) randomly selecting k samples from the sample set as sample subsets, and putting the k samples into a root node of a tree in an isolated forest algorithm;
(3) randomly designating a dimension of the sample, and randomly generating a cut value p, wherein the cut value p is positioned between the maximum value and the minimum value of the designated dimension in the current node data;
(4) dividing the current node data space into left and right sub-node spaces by a cutting value p: data with the appointed dimension value smaller than the cutting value p is placed in the left sub-node space, and data with the appointed dimension value larger than or equal to the cutting value p is placed in the right sub-node space;
(5) recursively steps (2) through (4) in the child node space until only one datum in the child node space or the datum in the child node space has reached a defined height, where the height refers to the number of longest simple path edges from the root node to the leaf nodes;
(6) cycling the steps (2) to (5) until t isolated trees iTree are generated;
(7) counting the path length h (x) of the sample point x for each sample point x in the sample set, wherein the path length of the sample point x is the number of edges passing from the root node to the leaf node of the tree;
(8) normalizing the average height of all the sample points, and calculating the anomaly score value of each sample point x: ；
wherein,,is an outlier value of sample point x, +.>For harmonizing the number, add>For a given number of samples n, E (h (x)) is the average of the sample point path lengths, which is the expectation of the sample path lengths in a collection of isolated trees;
step c: motor status identification
If the abnormal score value is greater than the state threshold value, judging that the motor is in an abnormal state; otherwise, judging that the motor is in a normal state;
step d: and feeding back a result, and feeding back the working state of the motor to an operator and a technical maintenance staff.
The method for analyzing the state of the finish rolling motor based on the isolated forest algorithm comprises the following specific steps of:
the motor vibration waveform historical data are vibration waveform data acquired by utilizing two sensors arranged at the free end and the load end of a finish rolling motor according to a certain time interval, the vibration waveform data are obtained by integrating acceleration waveforms, one line of data acquired simultaneously each time comprises a motor free end high-frequency acceleration waveform, a motor free end low-frequency acceleration waveform, a motor free end speed waveform, a load end high-frequency acceleration waveform, a load end low-frequency acceleration waveform and a load end speed waveform, and the mean square error of the waveform data is calculated to obtain the following six lines of vibration data: the method comprises the following steps of a free end high-frequency acceleration total value, a free end low-frequency acceleration total value, a free end speed total value, a load end high-frequency acceleration total value, a load end low-frequency acceleration total value and a load end speed total value.
According to the finish rolling motor state analysis method based on the isolated forest algorithm, the lower frequency limit of the free end high-frequency acceleration waveform and the load end high-frequency acceleration waveform is 2Hz, the upper frequency limit is 20000 Hz, the lower frequency limit of the free end low-frequency acceleration waveform and the load end low-frequency acceleration waveform is 2Hz, the upper frequency limit is 2000 Hz, the lower frequency limit of the free end speed waveform and the load end speed waveform is 2Hz, and the upper frequency limit is 1000 Hz.
According to the finish rolling motor state analysis method based on the isolated forest algorithm, the data length of the free end high-frequency acceleration waveform, the free end low-frequency acceleration waveform, the load end high-frequency acceleration waveform and the load end low-frequency acceleration waveform is 128K, and the data length of the free end speed waveform and the load end speed waveform is 16K.
According to the finish rolling motor state analysis method based on the isolated forest algorithm, in the preprocessing process of motor vibration history data, each column of vibration data is subjected to standardized processing according to the following steps:
wherein,,for normalization of pre-processed data, +.>For normalization of the processed data, +.>Mean value of a list of vibration data, +.>Is the variance of a list of vibration data.
In the above method for analyzing the state of the finish rolling motor based on the isolated forest algorithm, the value of the harmonic number H (n-1) is estimated as ln (n-1) +0.5772156649.
According to the finish rolling motor state analysis method based on the isolated forest algorithm, the state threshold value of the anomaly score value is set to be 0.
The beneficial effects of the invention are as follows:
the invention uses the unsupervised isolated forest algorithm to cluster the sample data without labels, the clustering result is more objective and accurate, the algorithm complexity is low, and the calculation speed is high; the method avoids subjectivity caused by manual marking of the data, and the isolated forest algorithm calculates the abnormal value of the data by utilizing the characteristics of small abnormal point number, large abnormal data characteristic value and large normal data difference, so that the method is more accurate; the complexity of the orphan forest algorithm depends on the size of the binary tree, and in the worst case, is O (n), where n represents the amount of data of the binary tree. Since the isolated forest algorithm is fast to calculate, it uses a very simple data structure compared to neural networks and does not require the use of recursive algorithms to calculate the ordering of the trees. This can greatly increase the computational efficiency, especially in cases where the hardware resources of the computer are limited. The invention can accurately and efficiently perform state analysis on the finish rolling motor, and can provide reliable information for correct maintenance of the motor in time.
Drawings
The invention is described in further detail below with reference to the accompanying drawings.
Fig. 1 is a flow chart of the present invention.
The symbols are respectively expressed as: d is a sample set; n is the number of samples;representing the i-th sample; />、/>、、/>、/>、/>The method comprises the steps of respectively obtaining a free end high-frequency acceleration total value, a free end low-frequency acceleration total value, a free end speed total value, a load end high-frequency acceleration total value, a load end low-frequency acceleration total value and a load end speed total value in an ith sample; p is a cut value; h (x) is the path length of sample point x; />An outlier value for sample point x; />The sum is the sum of the adjustment; />For a given number of samples n, the average value of the sample point path length; />For the expected value of the path length of the sample in a batch of isolated trees, +.>For normalization of pre-processed data, +.>For the purpose of normalizing the processed data,mean value of a list of vibration data, +.>Is the variance of a list of vibration data.
Detailed Description
The analysis method of the invention comprises the following steps:
(1) Data preprocessing: integrating and normalizing the vibration history data of the finish rolling mill motor;
(2) Training phase: training an isolated tree by using an anomaly detection algorithm, and calculating an anomaly score value of each sample;
(3) And (3) motor state identification: comparing the value of the abnormal score of the sample with a state threshold value to obtain a sample label, and judging the working state of the motor according to the label;
(4) And (3) feeding back a result: and feeding back the working condition of the motor to an operator, and judging the next operation of finish rolling in advance.
The method specifically comprises the following steps:
step a: preprocessing of motor vibration history data
Firstly, integrating vibration waveform historical data: the motor vibration waveform historical data are vibration data of the free end and the load end of the finish rolling motor, which are respectively acquired according to a certain time interval by using two sensors arranged at the free end and the load end of the finish rolling motor, wherein one row of data acquired each time comprises a motor free end high-frequency (2 Hz-20000 Hz) acceleration waveform, a free end low-frequency (2 Hz-2000 Hz) acceleration waveform, a free end speed waveform (2 Hz-1000 Hz), a load end high-frequency (2 Hz-20000 Hz) acceleration waveform, a load end low-frequency (2 Hz-2000 Hz) acceleration waveform and a load end speed waveform (2 Hz-1000 Hz), the data length of the acceleration waveforms is 128K (namely 128 multiplied by 1024 sampling points), and the data length of the speed waveforms is 16K (namely 16 multiplied by 1024 sampling points).
The mean square error of each waveform data was calculated to obtain the following six vibration data: a free end high-frequency acceleration total value, a free end low-frequency acceleration total value, a free end speed total value, a load end high-frequency acceleration total value, a load end low-frequency acceleration total value and a load end speed total value;
the vibration data of each column is then normalized as follows:
wherein,,for normalization of pre-processed data, +.>For normalization of the processed data, +.>Mean value of a list of vibration data, +.>Is the variance of a list of vibration data.
Step b: constructing a sample set, performing isolated forest algorithm training, and calculating abnormal score values of all data
(1) Taking a group of vibration data which are collected simultaneously after pretreatment as a sample to obtain a sample setN is the number of samples, +.>Represents the i-th sample, i=1, 2, …, n,，/>、/>、/>、/>、/>、/>the method comprises the steps of respectively obtaining a free end high-frequency acceleration total value, a free end low-frequency acceleration total value, a free end speed total value, a load end high-frequency acceleration total value, a load end low-frequency acceleration total value and a load end speed total value in an ith sample;
(2) randomly selecting k samples from the sample set as sample subsets, and putting the k samples into a root node of a tree in an isolated forest algorithm;
(3) randomly specifying one dimension of the sample, i.e. the characteristic or parameter, refers to (1)、/>、/>、/>、/>、/>Randomly generating a cutting value p in the current node data, wherein the cutting value p is positioned between the maximum value and the minimum value of the specified dimension in the current node data;
(4) generating a hyperplane by the cutting value p, and dividing the current node data space into a left sub-node space and a right sub-node space: data with the appointed dimension value smaller than the cutting value p is placed in the left sub-node space, and data with the appointed dimension value larger than or equal to the cutting value p is placed in the right sub-node space;
(5) recursively steps (2) through (4) in the sub-node space, continuously constructing new sub-nodes until only one data in the sub-node space can not be cut any more or the data in the sub-node space has reached a defined height, wherein the defined height refers to the number of the longest simple path edge from the root node to the leaf node, and the iTree construction is finished;
(6) cycling the steps (2) to (5) until t isolated trees iTree are generated, and constructing the isolated forest is finished;
(7) counting the path length h (x) of the sample point x for each sample point x in the sample set, wherein the path length of the sample point is the number of edges passing from the root node to the leaf node of the tree;
(8) and carrying out normalization processing on the average height of all sample points, and calculating an anomaly score value of each sample point x, wherein the calculation formula is as follows: ；
wherein,,is an outlier value of sample point x, +.>For the sum, this value can be estimated as +.>。/>For a given number of samples n, the average value of the sample point path length is used to normalize the path length +.>，/>A desire for path length of a sample in a collection of orphaned trees;
step c: motor status identification
Screening out data with an abnormal score value exceeding a state threshold according to an abnormal detection result of an isolated forest algorithm, wherein the state threshold is set to 0, a sample label with the abnormal score value being greater than 0 is set to-1, a sample label with the abnormal score value being less than or equal to 0 is set to-1, and the sample label is the abnormal value which is the abnormal value of-1, so that the working state of the motor is abnormal;
step d: and feeding back the working state of the motor to operators and technical maintenance staff to judge the next operation of the finish rolling mill accurately.
The invention has the following advantages:
(1) And an unsupervised algorithm is used for detecting abnormal values of sample data without labels, so that subjectivity of supervised learning adopting the labeled data and complexity of labeling work are avoided.
(2) And the working state of the motor is accurately identified by using the abnormal value detection result, so that technicians are helped to respond in time, and economic loss is reduced.
(3) The method has strong universality and robustness to various sampled data without other data processing algorithms.
(4) The isolated forest outlier detection algorithm has good capability and speed for processing big data, is different from K-means algorithm and the like, does not need to calculate indexes such as related distance, density and the like, can greatly improve the speed and reduce the system overhead.
Example 1
The engineering example is the motor vibration data of a finish rolling U6 rolling mill, and data of one week from 2023, 1 month, 2 days to 2023, 1 month, 8 days are selected for analysis. The sensor collects vibration data of the finish rolling motor every 30 seconds, and the total data is 2880 pieces per day.
The experiment is based on Python programming language and isolated forest algorithm programming, and aims to obtain abnormal values of vibration data of the finish rolling motor. Firstly, preprocessing data collected by history, and integrating three waveform data of a motor free end and a motor load end sensor into a matrix of 20160 row and 6 column; the independent forest algorithm takes the preprocessed 20160 group data as a sample set, samples the sample set, constructs an independent tree, tests each independent tree in the forest, records the path length, calculates an outlier score of each sample point according to an outlier score calculation formula, and calculates a sample label according to whether the outlier score is larger than a state threshold value or not.
And (3) obtaining each sample label through Python programming, wherein the label is 1 as a normal value, and the label is-1 as an abnormal value, so that the working state of the finish rolling motor is obtained, and the working state is fed back to finishing mill operators and technicians so as to respond to the next operation of the finish rolling mill in time.
The working times of the finish rolling motor in each working state in one week are shown in the table, taking No. 1 month 2 as an example, the working times of the finish rolling motor in the normal working state are 2840 times, and the working times are 98.61% of the total working times; the working times under the warning state (abnormal state) are 34 times, and account for 1.39% of the total working times; therefore, most of the finishing rolling motor works in a normal state, and when the finishing rolling motor is in an abnormal state, operators or technicians are informed of timely safety check in time, and maintenance is carried out if necessary.
Claims (7)
1. The finish rolling motor state analysis method based on the isolated forest algorithm is characterized by comprising the following steps of:
a: preprocessing of motor vibration history data
Measuring motor vibration waveform signals according to a certain time interval by using two acceleration sensors arranged at the free end and the load end of the finish rolling motor, thereby forming motor vibration waveform historical data, integrating the motor vibration waveform historical data, and obtaining the following six vibration data: the method comprises the steps of performing standardized processing on vibration data of each column, wherein the vibration data comprise a free end high-frequency acceleration total value, a free end low-frequency acceleration total value, a free end speed total value, a load end high-frequency acceleration total value, a load end low-frequency acceleration total value and a load end speed total value;
step b: constructing a sample set, performing isolated forest algorithm training, and calculating abnormal score values of all data:
(1) taking a group of vibration waveform data which are collected simultaneously and are preprocessed as a sample to obtain a sample setN is the number of samples, +.>Represents the i-th sample, i=1, 2, …, n,wherein->、/>、/>、/>、/>、/>The method comprises the steps of respectively obtaining a free end high-frequency acceleration total value, a free end low-frequency acceleration total value, a free end speed total value, a load end high-frequency acceleration total value, a load end low-frequency acceleration total value and a load end speed total value in an ith sample, wherein the total values are obtained by carrying out square operation on each acquired data point, summing the squares of all the data points, dividing the sum by the sampling points and taking the square root, and the root mean square value of the digital signal is obtained, namely the total value;
(2) randomly selecting k samples from the sample set as sample subsets, and putting the k samples into a root node of a tree in an isolated forest algorithm;
(3) randomly designating a dimension of the sample, and randomly generating a cut value p, wherein the cut value p is positioned between the maximum value and the minimum value of the designated dimension in the current node data;
(4) dividing the current node data space into a left sub-node space and a right sub-node space by using a cutting value p: data with the appointed dimension value smaller than the cutting value p is placed in the left sub-node space, and data with the appointed dimension value larger than or equal to the cutting value p is placed in the right sub-node space;
(5) recursively executing the steps (2) to (4) in the sub-node space, namely circularly executing the steps (2) to (4) until only one data in the sub-node space or the data in the sub-node space reaches a limited height;
(6) cycling the steps (2) to (5) until t isolated trees iTree are generated;
(7) counting the path length h (x) of the sample point x for each sample point x in the sample set, wherein the path length of the sample point x is the number of edges passing from the root node to the leaf node of the tree;
(8) normalizing the average height of all the sample points, and calculating the anomaly score value of each sample point x: ；
wherein,,is an outlier value of sample point x, +.>For harmonizing the number, add>For a given number of samples n, E (h (x)) is the average of the sample point path lengths, which is the expectation of the sample path lengths in a collection of isolated trees;
step c: motor status identification
If the abnormal score value is greater than the state threshold value, judging that the motor is in an abnormal state; otherwise, judging that the motor is in a normal state;
step d: and feeding back a result, and feeding back the working state of the motor to an operator and a technical maintenance staff.
2. The method for analyzing the state of a finish rolling motor based on an isolated forest algorithm according to claim 1, wherein the method for integrating the historical data of the vibration waveform of the motor acquired at certain time intervals is as follows:
the motor vibration waveform historical data are vibration waveform data acquired according to a certain time interval by using two sensors arranged at the free end and the load end of a finish rolling motor, and 128K sampling points of one line of data acquired at the same time each time comprise a motor free end high-frequency acceleration waveform, a free end low-frequency acceleration waveform, a free end speed waveform, a load end high-frequency acceleration waveform, a load end low-frequency acceleration waveform and a load end speed waveform, and the mean square error of the waveform data is calculated to obtain the following six columns of vibration data: the method comprises the steps of carrying out square operation on all data points of speed or acceleration, summing squares of all data points, dividing the sum by sampling points and taking square root to obtain a root mean square value of the digital signal, wherein the root mean square value is the total value.
3. The method for analyzing the state of the finish rolling motor based on the isolated forest algorithm according to claim 2, wherein the frequency lower limit of the free end high-frequency acceleration waveform and the load end high-frequency acceleration waveform is 2Hz, the frequency upper limit of the free end low-frequency acceleration waveform and the load end low-frequency acceleration waveform is 20000 Hz, the frequency lower limit of the free end low-frequency acceleration waveform and the load end low-frequency acceleration waveform is 2Hz, the frequency upper limit of the free end speed waveform is 2000 Hz, the frequency lower limit of the free end speed waveform and the frequency upper limit of the load end speed waveform is 1000 Hz.
4. The finish rolling motor state analysis method based on the isolated forest algorithm according to claim 3, wherein the data length of the free end high-frequency acceleration waveform, the free end low-frequency acceleration waveform, the load end high-frequency acceleration waveform and the load end low-frequency acceleration waveform is 128K, and the data length of the free end speed waveform and the load end speed waveform is 16K.
5. The method for analyzing the state of a finish rolling motor based on an isolated forest algorithm according to claim 4, wherein in the preprocessing of the motor vibration history data, the normalization processing is performed on each column of vibration data according to the following formula:
wherein,,for normalization of pre-processed data, +.>For normalization of the processed data, +.>Mean value of a list of vibration data, +.>Is the variance of a list of vibration data.
6. The method for analyzing the state of a finish rolling motor based on an isolated forest algorithm according to claim 5, wherein the harmonic number H (n-1) can be estimated as ln (n-1) +0.5772156649, where n is the number of samples.
7. The method for analyzing the state of a finish rolling motor based on an isolated forest algorithm according to claim 6, wherein the state threshold of the anomaly score value is set to 0.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310472163.9A CN116502043A (en) | 2023-04-27 | 2023-04-27 | Finish rolling motor state analysis method based on isolated forest algorithm |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310472163.9A CN116502043A (en) | 2023-04-27 | 2023-04-27 | Finish rolling motor state analysis method based on isolated forest algorithm |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116502043A true CN116502043A (en) | 2023-07-28 |
Family
ID=87321222
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310472163.9A Pending CN116502043A (en) | 2023-04-27 | 2023-04-27 | Finish rolling motor state analysis method based on isolated forest algorithm |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116502043A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117407822A (en) * | 2023-12-12 | 2024-01-16 | 江苏新希望生态科技有限公司 | Full-automatic bud seedling machine and control method thereof |
-
2023
- 2023-04-27 CN CN202310472163.9A patent/CN116502043A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117407822A (en) * | 2023-12-12 | 2024-01-16 | 江苏新希望生态科技有限公司 | Full-automatic bud seedling machine and control method thereof |
| CN117407822B (en) * | 2023-12-12 | 2024-02-20 | 江苏新希望生态科技有限公司 | Full-automatic bud seedling machine and control method thereof |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110414155B (en) | Fan component temperature abnormity detection and alarm method with single measuring point | |
| CN112414694B (en) | Equipment multistage abnormal state identification method and device based on multivariate state estimation technology | |
| CN115409131B (en) | Production line abnormity detection method based on SPC process control system | |
| CN110738255A (en) | device state monitoring method based on clustering algorithm | |
| CN111738308A (en) | Dynamic threshold detection method for monitoring index based on clustering and semi-supervised learning | |
| CN112861350B (en) | Temperature overheating defect early warning method for stator winding of water-cooled steam turbine generator | |
| CN113253037B (en) | Current ripple-based edge cloud cooperative equipment state monitoring method and system and medium | |
| CN111913443A (en) | Industrial equipment fault early warning method based on similarity | |
| CN116502043A (en) | Finish rolling motor state analysis method based on isolated forest algorithm | |
| CN113284004A (en) | Power data diagnosis treatment method based on isolated forest algorithm | |
| CN109141625B (en) | Online state monitoring method for ball screw pair | |
| CN111666978B (en) | Intelligent fault early warning system for IT system operation and maintenance big data | |
| CN111157850A (en) | Mean value clustering-based power grid line fault identification method | |
| CN115220396A (en) | Intelligent monitoring method and system for numerical control machine tool | |
| CN117318614A (en) | Photovoltaic inverter fault prediction method | |
| CN117150418B (en) | Transformer operation detection period formulation method and system based on state characteristic fault tree | |
| CN117350377A (en) | Knowledge graph driving-based equipment fault diagnosis method and device | |
| CN116467592A (en) | Production equipment fault intelligent monitoring method and system based on deep learning | |
| CN116146515A (en) | Abnormality detection method for centrifugal fan equipment of cement plant | |
| CN115310497A (en) | Abnormity identification method for bearing of numerical control machine tool | |
| CN112417966B (en) | Multi-dimensional machine fault detection method based on audio signal characteristics and AI algorithm | |
| CN115326393A (en) | Wind turbine generator bearing pair fault diagnosis method based on temperature information | |
| Yu et al. | Application of distance measure in KNN motor fault diagnosis | |
| CN111814325A (en) | Single variable alarm threshold value optimization method for non-steady operation of coal-fired power generation equipment | |
| CN109102001A (en) | A kind of gene improve the rotor on-line fault diagnosis method of neural network |
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 |