A coal rock cutting medium efficient recognition and mechanical fault accurate diagnosis method of a coal mining machine

Abstract

The invention provides a coal cutting medium efficient recognition and mechanical fault accurate diagnosis method of coal winning machine. The vibration signal discriminant function model of the coal winning machine is constructed, and the high-dimensional feature set of the cutting medium type vibration signal and the high-dimensional feature set of the mechanical fault type vibration signal are efficiently distinguished. The vibration signal graph embedding model of the coal winning machine is created, effectively solving the problems of large volume, scattered characteristics and data redundancy of the vibration signal of the coal winning machine. And the vibration signal of the coal winning machine is efficiently and quickly reduced in dimension, improving the efficiency and accuracy of coal and rock medium recognition and fault diagnosis.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent and safe coal mining, and particularly relates to a method for efficient identification of coal and rock cutting medium and accurate diagnosis of mechanical failure of a coal mining machine. BACKGROUND

[0002] As one of the key equipment of the fully mechanized coal mining face, the coal mining machine realizes intelligent and unmanned mining, which is an important trend of the development of the coal mining machine. Meanwhile, the continuous, safe and reliable operation of the coal mining machine is an important guarantee for high yield and high efficiency of coal. The coal and rock cutting medium identification technology is the basis for realizing the intelligent and unmanned mining of the coal mining machine. At the same time, the mechanical failure prediction and diagnosis of the coal mining machine can effectively reduce the failure downtime of the coal mining machine.

[0003] During the operation of the coal mining machine, multi-source vibration signals are formed due to the rotation of the transmission system, the change of the coal and rock cutting medium and mechanical failure. Among them, the vibration caused by the rotation of the transmission system is the normal working vibration, and the vibration caused by the change of the cutting medium and mechanical failure is the abnormal vibration beyond the allowable vibration range. Existing researches show that different cutting coal and rock media correspond to different vibration characteristics, and the failure categories of the coal mining machine and the vibration signal characteristics have certain regularity. Through the processing and tracing analysis of the vibration signals of the coal mining machine, the working state and the equipment health state of the coal mining machine can be obtained, which is an important means to realize the identification of the coal and rock cutting medium and the diagnosis of the mechanical failure. However, the process of the coal and rock cutting medium identification and the mechanical failure diagnosis of the multi-source vibration signals of the coal mining machine currently exists mutual interference, which makes the analysis process complex and increases the difficulty of calculation and processing. In addition, the vibration signals of the coal mining machine are large in volume, the effective feature signals are scattered, and the information is redundant, which reduces the efficiency and accuracy of the cutting medium identification and mechanical failure diagnosis algorithm. At the same time, the existing researches on the vibration signals of the coal mining machine are mostly single-source researches, and the mutual interference between the cutting medium identification process and the mechanical failure diagnosis process during the operation of the coal mining machine is rarely considered comprehensively. Therefore, how to distinguish and separate the cutting medium type vibration signal and the mechanical failure type vibration signal, and then respectively realize the efficient identification of the coal and rock cutting medium and the accurate diagnosis of the mechanical failure is an important problem to be solved.

[0004] Therefore, it is urgent to develop a comprehensive method for efficient identification of the coal and rock cutting medium and accurate diagnosis of the mechanical failure of the coal mining machine under the complex mining conditions of the coal mine considering the mutual interference of the vibration signals and the data dimension reduction, which has great engineering value. SUMMARY

[0005] The present application aims to provide a method for efficient identification of the coal and rock cutting medium and accurate diagnosis of the mechanical failure of the coal mining machine to solve the problems in the prior art.

[0006] The technical scheme adopted for realizing the object of the present application is as follows: a coal cutter coal-rock cutting medium high-efficiency identification and mechanical fault accurate diagnosis method, comprising the following steps:

[0007] 1) in the debugging stage, the idle vibration signal distribution corresponding to different cutting speeds and different cutting positions of the coal cutter in the current fully-mechanized coal face is collected, and filtering and time-frequency domain feature extraction are performed. The time-frequency domain feature indexes extracted include the amplitude X ppv , the mean square value , the effective value X rms , the kurtosis index K v and the pulse index I. The vibration signal feature indexes are marked according to different cutting speeds and different cutting positions, and a vibration signal discrimination feature matrix E composed of initial idle vibration signal discrimination feature vectors is constructed.

[0008] 2) in the cutting coal-rock operation stage, the loaded vibration signal is collected in real time and filtering and time-frequency domain feature extraction are performed. The current working condition loaded vibration signal discrimination feature vector is constructed. When any one of the vibration signal feature indexes is greater than or equal to the corresponding feature index in the analysis threshold value P, it is determined that the monitored vibration signal is an abnormal vibration signal, and step 3) is started in response, otherwise step 2) is repeated. The analysis threshold value P is the minimum value of the response vibration signal. When the coal cutter vibration signal is less than the P value, the vibration is in the allowable range, and the vibration signal does not reflect the cutting medium change and mechanical fault.

[0009] 3) a high-dimensional feature set U of abnormal vibration signal is constructed. When the vibration signal is determined to be an abnormal vibration signal, 16 time domain feature indexes and 13 frequency domain feature indexes are obtained through further time-frequency domain analysis of the abnormal vibration signal. The 29 feature indexes (p1, p2,..., p 29 ) of the abnormal vibration signal of the current working condition loaded state in three directions are combined to obtain the high-dimensional feature set U of the abnormal vibration signal.

[0010] 4) a current working condition idle vibration signal discrimination feature vector is constructed. When the vibration signal is determined to be an abnormal vibration signal, the current working condition is kept unchanged, the electromechanical control system of the coal cutter is triggered to respond, the drum is controlled to change from the loaded state to the idle state, the vibration signal at the same position, in the same direction and at the same cutting speed is collected, and filtering and time-frequency domain feature extraction are performed. The current working condition idle vibration signal discrimination feature vector is constructed.

[0011] 5) the initial idle vibration signal discrimination vector in E with the same marks as and is called. and Substitute into the vibration signal discrimination function, according to the function value to distinguish abnormal vibration signal high-dimensional feature set U is cutting medium type vibration signal high-dimensional feature set U1 or mechanical fault type vibration signal high-dimensional feature set U2. When f function value is less than the threshold M, the abnormal vibration signal high-dimensional feature set U is cutting medium type vibration signal high-dimensional feature set U1. When f function value is greater than or equal to the threshold M, the abnormal vibration signal high-dimensional feature set U is mechanical fault type vibration signal high-dimensional feature set U2. Wherein, the threshold M is determined according to the type of coal mining machine and the working condition of coal mining. The vibration signal discrimination function is expressed as follows:

[0012]

[0013]

[0014]

[0015]

[0016] α+β+δ+μ+θ=1

[0017] In the formula, the numerator A1-A0 is a fault diagnosis factor, which is the change amount of the current no-load vibration state relative to the initial no-load vibration state. The denominator B1-A1 is a cutting medium factor, which is the change amount of the current load vibration state relative to the current no-load vibration state. α, β, δ, μ, θ are characteristic index weight coefficients, which are determined according to the actual situation and expert experience.

[0018] 6) According to the discrimination result in step 5), the corresponding identification diagnosis program is selected. If the abnormal vibration signal high-dimensional feature set U is the cutting medium type vibration signal high-dimensional feature set U1, the coal rock cutting medium identification program is called. Otherwise, the abnormal vibration signal high-dimensional feature set U is the mechanical fault type vibration signal high-dimensional feature set U2, and the shearer cutting part fault diagnosis program is called. The coal rock cutting medium identification program inputs the cutting medium type vibration signal high-dimensional feature set U1 into the shearer vibration signal graph embedding model for dimension reduction embedding to generate a low-dimensional sensitive feature vector, and then inputs the low-dimensional feature vector into the coal rock cutting medium KNN classifier to divide the coal rock cutting working condition into empty load working condition, coal cutting working condition, cutting gangue coal working condition and cutting rock working condition. The shearer cutting part fault diagnosis program inputs the mechanical fault type vibration signal high-dimensional feature set U2 into the shearer vibration signal graph embedding model for dimension reduction embedding to generate a low-dimensional sensitive feature vector, and then inputs the low-dimensional feature vector into the shearer cutting part fault diagnosis KNN classifier to identify the corresponding shearer fault category. The shearer cutting part fault type includes bearing fault and gear fault. The bearing fault includes inner ring fault, rolling body fault, outer ring fault and retainer fault. The gear fault includes gear fracture, tooth surface scratch, tooth surface pitting, gear wear, tooth surface fatigue and tooth surface gluing.

[0019] 7) Decision response based on classification diagnosis result. Different mining schemes are taken according to the coal rock cutting type. Sound and light alarm and maintenance management are carried out according to the fault type.

[0020] Further, in steps 1) and 2), the multi-channel vibration signal acquisition module is used for shearer vibration signal distribution acquisition, time-frequency domain analysis and signal transmission.

[0021] Further, in steps 1), 2) and 4), the spatial cutting position of the shearer drum is determined by digital spatial modeling of the fully mechanized working face. The midpoint of the working face floor is selected as the coordinate origin, the mining width direction is selected as the X-axis direction, and the mining height direction is selected as the Y direction to construct a X-Y digital model of the fully mechanized working face. The shearer cutting position is defined as (x, y). The initial empty load vibration signal discrimination feature matrix E, the initial empty load vibration signal discrimination feature vector The current working condition load vibration signal discrimination feature vector And the current working condition empty load vibration signal discrimination feature vector As follows:

[0022]

[0023]

[0024]

[0025]

[0026]

[0027] where x, y represent the cutting position of the shearer drum. The superscript v n represents different cutting speeds. The superscript i represents different vibration monitoring directions, where i = 1 represents the cutting height direction vibration signal. i = 2 represents the cutting width direction vibration signal. i = 3 represents the cutting depth direction vibration signal. ppv represents the passband amplitude. represents the mean square value. X rms represents the effective value. K v represents the kurtosis index, and I represents the pulse index.

[0028] Further, in step 2), the analysis threshold P is obtained according to actual conditions, historical data and expert experience.

[0029] Further, in step 3), the 16 time domain feature indexes include passband amplitude, maximum value, minimum value, mean value, variance, skewness, mean square value, effective value, square root amplitude, absolute average amplitude, kurtosis index, skewness index, margin index, pulse index, peak value index and waveform index. The 13 frequency domain feature indexes include one frequency, two frequency, half frequency, high frequency, average frequency, E0, E1, E2, E3, E4, E5, E6 and E7. Wherein E i is the filter signal uniformly divided into 8 sub-bands by 3-layer orthogonal wavelet packet decomposition using db4 wavelet packet function.

[0030] Further, in step 6), the shearer vibration signal graph embedding model construction method includes the following sub-steps:

[0031] 6.1) Normalizing the medium type vibration signal high-dimensional feature set U1 or the mechanical fault type vibration signal high-dimensional feature set U2, and dividing it into a training sample set X1 and a test sample set X2.

[0032] 6.2) Inputting the training sample set X1 into the boundary Fisher analysis algorithm for training to obtain a projection matrix A.

[0033] 6.3) Projecting and mapping X1, X2 using A to obtain the final low-dimensional sensitive feature subsets Y1, Y2. Inputting the low-dimensional sensitive feature subsets Y1, Y2 of the training sample and the test sample into the KNN classifier to obtain the cutting medium type or the mechanical fault type.

[0034] Further, in step 7), the mining scheme is obtained by analyzing historical coal mining data and expert evaluation.

[0035] Further, in step 7), the sound and light alarm displays the fault category in the form of sound and LED, reminding the maintenance personnel to repair in the first time.

[0036] Further, after step 7), there is also a related step of updating the shearer vibration information management database.

[0037] Further, the data of single monitoring analysis is updated to the vibration information management database. The vibration information management database includes a shearer basic information library, a shearer cutting working condition database, a shearer cutting fault database, an initial no-load vibration information database and a vibration history information database.

[0038] The technical effect of the present application is self-evident:

[0039] A. A shearer vibration signal discriminant function model is constructed. The present application aims at the problem that the shearer vibration signal is multi-source and complex and difficult to further analyze and utilize, constructs a vibration signal discriminant function model, defines a cutting medium type factor and a mechanical fault type factor, and distinguishes the high-dimensional feature set of the cutting medium type vibration signal and the high-dimensional feature set of the mechanical fault type vibration signal, which can solve the problem that the vibration monitoring technology is difficult to separate the complex information in the coal mine monitoring application;

[0040] B. A fully mechanized working X-Y plane digital model is constructed, the vibration information is marked from two dimensions of the cutting surface position and the cutting speed, and a plurality of shearer vibration signal matrices are defined, which provides a new idea for the digital development of the coal mining work;

[0041] C. A shearer vibration signal graph embedding model is created, which realizes efficient and rapid dimension reduction of the shearer vibration signal, effectively solves the problems of large volume, scattered characteristics and data redundancy of the shearer vibration signal, and improves the efficiency and accuracy of coal and rock medium recognition and fault diagnosis. BRIEF DESCRIPTION OF DRAWINGS

[0042] Figure 1 It is a flow chart of the efficient recognition of coal and rock cutting medium of the shearer and the accurate diagnosis method of mechanical fault;

[0043] Figure 2 It is a schematic diagram of the construction method of the X-Y digital model of the fully mechanized working face;

[0044] Figure 3 It is a schematic diagram of the principle of the boundary Fisher analysis algorithm in the graph embedding model;

[0045] Figure 4 It is a schematic diagram of the cutting part of the drum-type shearer. DETAILED DESCRIPTION

[0046] The application will be further described in connection with the following examples, which should not be construed as limiting the above subject matter of the application to the examples described below. Various substitutions and modifications will occur to those of ordinary skill in the art upon reading the description that follow, and such substitutions and modifications are intended to fall within the scope of the application. Although specific terms can be employed in the description, these terms are used in a generic and descriptive sense only and not for purposes of limitation.

[0047] Example 1

[0048] During the process of cutting coal and rock, the vibration response of the rocker arm excited by the cutting force will change with the change of the cutting medium, that is, different cutting conditions of coal and rock correspond to different vibration characteristics. Monitoring the vibration parameters in the coal mining process can realize the identification of the coal-rock interface in the cutting process. The mechanical fault of the cutting part of the coal mining machine will cause the change of the vibration signal in the working process. Research shows that there is a certain regularity between the fault category of the coal mining machine and the vibration signal characteristics. The vibration signal in the coal mining process can realize the diagnosis of the fault of the cutting part of the coal mining machine.

[0049] The embodiment provides a coal-rock cutting medium efficient identification and mechanical fault accurate diagnosis method of a coal mining machine, which comprises the following steps:

[0050] 1) Debugging phase. In the debugging phase, a basic information database of the coal mining machine is constructed, an analysis threshold P is determined, a discrimination threshold M is determined, a multi-channel vibration signal acquisition module is used to acquire the no-load vibration signal distribution of the rocker arm of the coal mining machine corresponding to different speeds and different spatial positions of the fully mechanized coal mining face, vibration signals are marked according to different speeds and different spatial positions, and an initial no-load vibration signal discrimination feature matrix E is constructed. The basic information database of the coal mining machine is used to store basic information such as the model of the coal mining machine, the coal mine, the coal seam and the fully mechanized coal mining face. The analysis threshold P is obtained according to the actual situation, historical data and expert experience. For example, in the past historical data, the minimum value set of the vibration signal characteristic index corresponding to the change of the cutting medium or the mechanical fault is P0, and P0 is determined as the value of the analysis threshold P. The discrimination threshold M is determined according to the actual situation, historical data and expert experience. For example, in the past historical data, the maximum value set of the vibration signal characteristic index corresponding to the change of the cutting medium or the mechanical fault is M0, and M0 is determined as the value of the discrimination threshold M. P0 is the minimum value set of abnormal vibration signal characteristic index, when any of the above five characteristic indexes of the vibration signal of the coal mining machine is less than the corresponding index value in P0, it is considered to be in the vibration allowable range, the vibration signal does not reflect the cutting medium change and mechanical failure, and belongs to normal vibration signal. When any characteristic index is greater than or equal to the corresponding index value in P, it is considered that the vibration exceeds the allowable range, the vibration signal reflects the cutting medium change and mechanical failure, and belongs to abnormal vibration signal, which needs to be further analyzed and processed to eliminate the abnormality. The value of the discrimination threshold M (M≥0) changes according to the model of the coal mining machine and the working conditions of the coal mining, in actual production, the value of M should be determined according to the actual situation, reference to historical data and expert experience. For example, according to the method of the application, when f The current working condition load vibration signal discrimination characteristic vector And the current working condition load vibration signal discrimination characteristic vector As follows:

[0051]

[0052]

[0053]

[0054]

[0055]

[0056] In the formula, x, y represent the cutting position of the drum of the coal mining machine, the superscript v n represents different cutting speeds, the superscript i represents different vibration monitoring directions, wherein i=1 represents the vibration signal in the mining height direction; i=2 represents the vibration signal in the mining width direction; i=3 represents the vibration signal in the mining depth direction, X ppv represents the amplitude of the pass frequency, represents the mean square value, X rms represents the effective value, Kv representing the kurtosis index, I represents the pulse index, when the working condition changes, the characteristic index in the different signal discrimination characteristic vector takes the value of the corresponding working condition.

[0057] Three discrimination characteristic vectors are defined in this embodiment by X ppv , the pass frequency amplitude, , the mean square value, X rms , the effective value, K v , the kurtosis index and the I pulse index. It is worth noting that relevant research shows that the pass frequency amplitude, the effective value, the mean square value can effectively reflect the vibration change of the cutting part of the coal mining machine, and the kurtosis index and the pulse index can directly reflect the vibration impact effect of cutting coal and rock. Therefore, when cutting medium type vibration signals and fault diagnosis type vibration signals are performed, the pass frequency amplitude, the effective value, the mean square value, the kurtosis index and the pulse index are selected as the discrimination characteristic indexes of the vibration signal discrimination function in this embodiment.

[0058] 2) Real-time acquisition of load vibration signals and preprocessing, judging whether the vibration signal is abnormal. In the cutting coal and rock operation stage, the multi-channel vibration signal acquisition module acquires vibration signals in real time and extracts the time-frequency domain characteristic indexes as described in 1), marks and constructs the current working condition load vibration signal discrimination characteristic vector When any one of the vibration signal characteristic indexes is greater than or equal to the corresponding characteristic index in the analysis threshold P, it is indicated that the monitored vibration signal is an abnormal vibration signal, which needs to be further processed, and step 3) is started in response. Otherwise, repeat step 2).

[0059] 3) Constructing an abnormal vibration signal high-dimensional feature set U. When the vibration signal is determined to be an abnormal vibration signal, 16 time domain characteristic indexes and 13 frequency domain characteristic indexes are obtained by further time-frequency domain analysis of the abnormal vibration signal. The abnormal vibration signal high-dimensional feature set U is obtained by combining the 29 characteristic indexes (p1, p2,..., p 29 ) of the three-direction abnormal vibration signals of the current working condition load state; the 16 time domain characteristic indexes include the pass frequency amplitude, the maximum value, the minimum value, the mean value, the variance, the skewness, the mean square value, the effective value, the square root amplitude, the absolute average amplitude, the kurtosis index, the skewness index, the margin index, the pulse index, the peak value index and the waveform index, and the calculation formulas of the indexes are shown in Table 1. The 13 frequency domain characteristic indexes include the first frequency, the second frequency, the half frequency, the high frequency, the average frequency, E0, E1, E2, E3, E4, E5, E6 and E7. Among them, E i is the filter signal uniformly divided into 8 sub-bands by 3-layer orthogonal wavelet packet decomposition using db4 wavelet packet function.

[0060] Table 1

[0061]

[0062] 4) Constructing the current working condition empty vibration signal discriminant feature vector When the vibration signal is determined as an abnormal vibration signal, the current working condition is kept unchanged, the mechanical and electrical control system of the coal mining machine is triggered to respond, the drum is controlled to change from the loaded state to the empty state, the vibration signal at the same position, the same direction and the same cutting speed is collected at this time, and the filtering and time-frequency domain feature index extraction as described in 1) are performed, the vibration signal is labeled, and the current working condition empty vibration signal discriminant feature vector is constructed

[0063] 5) Calling E in which and the initial empty vibration signal discriminant vector is labeled Substitute and into the vibration signal discriminant function, and according to the function value, the abnormal vibration signal high-dimensional feature set U is determined as the cutting medium type vibration signal high-dimensional feature set U1 or the mechanical fault type vibration signal high-dimensional feature set U2; when the f function value is less than the determination threshold M, the abnormal vibration signal high-dimensional feature set U is the cutting medium type vibration signal high-dimensional feature set U1; when the f function value is greater than or equal to the determination threshold M, the abnormal vibration signal high-dimensional feature set U is the mechanical fault type vibration signal high-dimensional feature set U2; wherein the vibration signal discriminant function is expressed as follows:

[0064]

[0065]

[0066]

[0067]

[0068] α+β+δ+μ+θ=1

[0069] Wherein, the numerator A1-A0 is a fault diagnosis factor, which is the change amount of the current empty vibration state relative to the initial empty vibration state; the denominator B1-A1 is a cutting medium factor, which is the change amount of the current loaded vibration state relative to the current empty vibration state; α, β, δ, μ, θ are feature index weight coefficients, which are determined according to actual conditions and expert experience.

[0070] 6) According to the discrimination result in step 5), the corresponding identification diagnosis program is selected. If the abnormal vibration signal high-dimensional feature set U is the cutting medium type vibration signal high-dimensional feature set U1, the coal rock cutting medium identification program is called; otherwise, the abnormal vibration signal high-dimensional feature set U is the mechanical fault type vibration signal high-dimensional feature set U2, and the shearer cutting part fault diagnosis program is called. The coal rock cutting medium identification program refers to inputting the cutting medium type vibration signal high-dimensional feature set U1 into the shearer vibration signal graph embedding model for dimension reduction embedding to generate a low-dimensional sensitive feature vector, and then inputting the low-dimensional feature vector into the coal rock cutting medium KNN classifier to divide the coal rock cutting working condition into empty load working condition, coal cutting working condition, cutting gangue coal working condition and cutting rock working condition. The shearer cutting part fault diagnosis program refers to inputting the mechanical fault type vibration signal high-dimensional feature set U2 into the shearer vibration signal graph embedding model for dimension reduction embedding to generate a low-dimensional sensitive feature vector, and then inputting the low-dimensional feature vector into the shearer cutting part fault diagnosis KNN classifier to identify the corresponding shearer fault category. The shearer cutting part fault type includes bearing fault and gear fault; the bearing fault includes inner ring fault, rolling body fault, outer ring fault and retainer fault; the gear fault includes gear fracture, tooth surface scratch, tooth surface pitting, gear wear, tooth surface fatigue and tooth surface gluing.

[0071] 7) Decision response based on classification diagnosis result. Different mining schemes are adopted according to the coal rock cutting type to realize safe and efficient coal mining of the shearer. According to the fault type, sound and light alarm and maintenance management are carried out.

[0072] 8) Update the shearer vibration information management database. The vibration information management database includes a shearer basic information library, a shearer cutting working condition database, a shearer cutting fault database, an initial empty load vibration information database and a vibration history information database. The shearer basic information library is mainly used to collect and store basic information such as coal seam depth, coal seam thickness and shearer type. The shearer cutting working condition database is used to store different coal rock working condition modes in the shearer cutting process and the reasonable operation scheme corresponding to each working condition. The coal rock working condition modes mainly include empty load working condition, coal cutting working condition, rock cutting working condition and cutting gangue coal working condition. The reasonable operation scheme includes the coordinated and efficient cutting speed and cutting height data obtained according to related research. The shearer cutting fault database is used to store different fault categories of the shearer and the corresponding measures. The construction of the shearer vibration information system is conducive to strengthening the management of the shearer work and providing a data basis for further research.

[0073] The embodiment comprehensively considers the coal-rock cutting medium identification process and the mechanical fault diagnosis process based on the mutual interference of vibration signals, discriminates the multi-source shearer vibration signals in two processes, respectively reduces the dimension of the high-dimensional shearer vibration signals by using a graph embedding model, respectively performs coal-rock cutting medium identification and mechanical fault diagnosis, and provides a new idea for shearer operation safety monitoring.

[0074] Embodiment 2:

[0075] The main steps of the embodiment are the same as those of embodiment 1), wherein in step 6), the shearer vibration signal graph embedding model is defined as follows:

[0076] Step 1: The medium type vibration signal high-dimensional feature set U1 or the mechanical fault type vibration signal high-dimensional feature set U2 in the above 3) is normalized and divided into a training sample set X1 and a test sample set X2.

[0077] Step 2: The training sample set X1 is input into the boundary Fisher analysis algorithm for training to obtain a projection matrix A.

[0078] Step 3: The A is used for feature projection mapping of X1 and X2 to obtain final low-dimensional sensitive feature subsets Y1 and Y2. The low-dimensional sensitive feature subsets Y1 and Y2 of the training sample and the test sample are input into a KNN classifier to obtain a cutting medium type or a mechanical fault type.

[0079] The boundary Fisher analysis (MFA) algorithm is defined as follows:

[0080] For each sample point in the training sample, its same-class intrinsic graph G B ={X,W B} and the different-class penalty graph G D ={X,W D} are constructed. Wherein X is the vertex of the two-class graph, representing the training sample data point; W is the edge of the two-class graph, representing the similarity matrix between the training sample data points. Secondly, the similarity matrix elements and If two sample points are neighbors in a certain similarity measure, they are connected by an edge, and the weight is given as 1, otherwise the weight is given as 0.

[0081]

[0082]

[0083] Wherein represent the same-class and different-class neighbors of X, respectively. With the above adjacency matrix, the local intra-class and inter-class similarity matrices can be defined as follows:

[0084] Φ(a)=aT X(D ω -F ω )X T a

[0085] Ψ(a)=a T X(D b -F b )X T a

[0086] where X(D ω -F ω )X T is the intra-class similarity matrix, and X(D b -F b )X T is the inter-class similarity matrix. The objective function is obtained by Fisher discriminant criterion:

[0087]

[0088] The calculation formula can be converted to:

[0089]

[0090] The objective function is transformed into a generalized eigenvalue problem by Lagrange multiplier method.

[0091] X(D b -F b )X T a=λa

[0092] MFA is to find a projection matrix A∈R D×d , so that X is projected by Y=A T X to get low-dimensional Y=[y1,y2,...,y n ]∈R d×n (d<D), then the projection vector that satisfies X(D b -F b )X T a=λa, only need to solve the first d largest eigenvalue corresponding to the eigenvector {a1,...,a d} of X(D b -F b )X T , record the projection matrix A=[a1,...,a d ].

[0093] Example 3:

[0094] The main steps of this embodiment are the same as those of Example 1), wherein in step 7), the mining scheme is obtained by analyzing historical coal mining data and expert evaluation.

[0095] Example 4:

[0096] The main steps of this embodiment are the same as those of Embodiment 1), wherein in step 7), the sound and light alarm displays the fault category in the form of sound and LED, reminding the maintenance personnel to perform maintenance in the first time.

Claims

1. A coal rock cutting medium efficient recognition and mechanical fault accurate diagnosis method of a coal mining machine, characterized in that, Comprising the following steps: 1) In the debugging phase, the idle vibration signal distribution corresponding to different cutting speeds and different cutting positions of the coal mining machine in the current fully mechanized working face is collected, and filtering and time-frequency domain feature extraction are performed; wherein the extracted time-frequency domain feature indexes include the amplitude X ppv , the mean square value , the effective value X rms , the kurtosis index K v and the pulse index I; according to different cutting speeds and different cutting positions, the vibration signal feature indexes are marked, and a vibration signal discrimination feature matrix E composed of initial idle vibration signal discrimination feature vectors is constructed; 2) Cutting coal rock operation stage, real-time acquisition of load vibration signal and filtering and time-frequency domain feature extraction; marking, building current working condition load vibration signal discriminant feature vector When When any one of the vibration signal feature indexes is greater than or equal to the corresponding feature index in the analysis threshold P, it is determined that the monitoring vibration signal is an abnormal vibration signal, and step 3) is started in response, otherwise step 2) is repeated; the analysis threshold P is the minimum value of the response vibration signal; when the vibration signal of the coal mining machine is less than the P value, the vibration is within the allowable range, and the vibration signal does not reflect the change of the cutting medium and mechanical failure; 3) Constructing high-dimensional feature set U of abnormal vibration signal; after determining the vibration signal as abnormal vibration signal, 16 time-domain characteristic indexes and 13 frequency-domain characteristic indexes are obtained through further time-frequency domain analysis of the abnormal vibration signal; 29 characteristic indexes (p1, p2,..., p 29 ) of abnormal vibration signals in three directions of current working condition load state are extracted to obtain high-dimensional feature set U of abnormal vibration signal; 4) Constructing the discriminant feature vector of the current working condition empty vibration signal When the vibration signal is determined as an abnormal vibration signal, the current working condition is kept unchanged, the electromechanical control system of the coal mining machine is triggered to respond, the roller is controlled to change from the loaded state to the empty state, the vibration signal at the same position, in the same direction and at the same cutting speed is collected at this time, and filtering and time-frequency domain feature extraction are performed; labeling is performed to construct the discriminant feature vector of the current working condition empty vibration signal 5) calling E in which and the initial empty load vibration signal discriminant vector will be and substituted into the vibration signal discriminant function, and according to the function value, the abnormal vibration signal high-dimensional feature set U is determined to be the cutting medium type vibration signal high-dimensional feature set U1 or the mechanical fault type vibration signal high-dimensional feature set U2; when the f function value is less than the determination threshold M, the abnormal vibration signal high-dimensional feature set U is the cutting medium type vibration signal high-dimensional feature set U1; when the f function value is greater than or equal to the determination threshold M, the abnormal vibration signal high-dimensional feature set U is the mechanical fault type vibration signal high-dimensional feature set U2; wherein the determination threshold M is determined according to the type of coal mining machine and the working condition of coal mining; the vibration signal discriminant function is expressed as follows: Alpha + beta + delta + mu + theta = 1 In the formula, the numerator A1-A0 is a fault diagnosis factor, which is the change of the current no-load vibration state relative to the initial no-load vibration state; the denominator B1-A1 is a cutting medium factor, which is the change of the current load vibration state relative to the current no-load vibration state; alpha, beta, delta, mu, theta are characteristic index weight coefficients, which are determined according to actual conditions and expert experience; 6) According to the discrimination result in step 5), the corresponding identification diagnosis program is selected; if the abnormal vibration signal high-dimensional feature set U is the cutting medium type vibration signal high-dimensional feature set U1, the coal and rock cutting medium identification program is called; otherwise, the abnormal vibration signal high-dimensional feature set U is the mechanical fault type vibration signal high-dimensional feature set U2, and the shearer cutting part fault diagnosis program is called; the coal and rock cutting medium identification program inputs the cutting medium type vibration signal high-dimensional feature set U1 into the shearer vibration signal graph embedding model for dimension reduction embedding to generate a low-dimensional sensitive feature vector, and then inputs the low-dimensional feature vector into a coal and rock cutting medium KNN classifier to divide the coal and rock cutting conditions into no-load working condition, coal cutting working condition, cutting gangue coal working condition and cutting rock working condition; the shearer cutting part fault diagnosis program inputs the mechanical fault type vibration signal high-dimensional feature set U2 into the shearer vibration signal graph embedding model for dimension reduction embedding to generate a low-dimensional sensitive feature vector, and then inputs the low-dimensional feature vector into a shearer cutting part fault diagnosis KNN classifier to identify the corresponding shearer fault category; the shearer cutting part fault type includes bearing fault and gear fault; the bearing fault includes inner ring fault, rolling element fault, outer ring fault and retainer fault; the gear fault includes gear fracture, tooth surface scratch, tooth surface pitting, gear wear, tooth surface fatigue and tooth surface gluing; 7) Decision response based on classification diagnosis result; different mining schemes are adopted according to the coal and rock cutting type; sound and light alarm and maintenance management are carried out according to the fault type.

2. The coal-rock cutting medium efficient identification and mechanical fault accurate diagnosis method of the coal winning machine according to claim 1, characterized in that: In steps 1) and 2), the multi-channel vibration signal acquisition module is used for shearer vibration signal distribution acquisition, time-frequency domain analysis and signal transmission.

3. The coal-rock cutting medium efficient identification and mechanical fault accurate diagnosis method of the coal winning machine according to claim 1, characterized in that: In step 1), step 2) and step 4), the spatial cutting position of the shearer drum is determined by digital spatial modeling of the fully-mechanized working face; the midpoint of the working face floor is selected as the coordinate origin, the mining width direction is the X-axis direction, and the mining height direction is the Y direction, so as to construct an X-Y digital model of the fully-mechanized working face; the shearer cutting position is defined as (x, y); the initial empty load vibration signal discriminant feature matrix E, the initial empty load vibration signal discriminant feature vector The current working condition load vibration signal discriminant feature vector And the current working condition empty load vibration signal discriminant feature vector As follows: In the formula, x, y represent the cutting position of the shearer drum; the superscript v n represents different cutting speeds; the superscript i represents different vibration monitoring directions, wherein i=1 represents the mining height direction vibration signal; i=2 represents the mining width direction vibration signal; i=3 represents the mining depth direction vibration signal; X ppv represents the pass frequency amplitude; represents the mean square value; X rms represents the effective value; K v represents the kurtosis index, and I represents the pulse index.

4. The coal-rock cutting medium efficient identification and mechanical fault accurate diagnosis method of the coal winning machine according to claim 1, characterized in that: In step 2), the analysis threshold P is obtained according to actual conditions, historical data and expert experience.

5. The coal-rock cutting medium efficient identification and mechanical fault accurate diagnosis method of a coal winning machine according to claim 1, characterized in that: In step 3), the 16 time-domain characteristic indexes include a pass frequency amplitude, a maximum value, a minimum value, a mean value, a variance, a skewness, a mean square value, an effective value, a square root amplitude, an absolute average amplitude, a kurtosis index, a skewness index, a margin index, a pulse index, a peak value index, and a waveform index; the 13 frequency-domain characteristic indexes include a one-fold frequency, a two-fold frequency, a half frequency, a high-fold frequency, an average frequency, E0, E1, E2, E3, E4, E5, E6, and E7; wherein E i The filter signal is obtained by performing 3-layer orthogonal wavelet packet decomposition on the db4 wavelet packet function to uniformly divide 8 sub-frequency bands.

6. The coal-rock cutting medium efficient identification and mechanical fault accurate diagnosis method of the coal winning machine according to claim 1, characterized in that, In step 6), the shearer vibration signal graph embedding model construction method includes the following sub-steps: 6.1) Normalizing the medium type vibration signal high-dimensional feature set U1 or the mechanical fault type vibration signal high-dimensional feature set U2 to divide them into training sample set X1 and test sample set X2; 6.2) Inputting the training sample set X1 into the boundary Fisher analysis algorithm for training to obtain the projection matrix A; 6.3) Using A to perform feature projection mapping on X1 and X2 to obtain the final low-dimensional sensitive feature subsets Y1 and Y2; inputting the low-dimensional sensitive feature subsets Y1 and Y2 of the training sample and the test sample into the KNN classifier to obtain the cutting medium type or the mechanical fault type.

7. The coal-rock cutting medium efficient identification and mechanical fault accurate diagnosis method of the coal winning machine according to claim 1, characterized in that: In step 7), the mining scheme is obtained by analyzing historical coal mining data and expert evaluation.

8. The coal-rock cutting medium efficient identification and mechanical fault accurate diagnosis method of the coal winning machine according to claim 1, characterized in that: In step 7), the sound and light alarm displays the fault category in the form of sound and LED to remind the operation and maintenance personnel to repair in the first time.

9. The coal-rock cutting medium efficient identification and mechanical fault accurate diagnosis method of a coal winning machine according to claim 1, characterized in that: After step 7), there is also a related step of updating the shearer vibration information management database.

10. The coal-rock cutting medium efficient identification and mechanical fault accurate diagnosis method of the coal winning machine according to claim 9, characterized in that: The data of the single monitoring analysis is updated to the vibration information management database; the vibration information management database comprises a shearer basic information database, a shearer cutting working condition database, a shearer cutting fault database, an initial no-load vibration information database and a vibration history information database.

Images