A fault diagnosis method and device for a mine belt conveyor system

Abstract

The application provides a mine belt conveyor system fault diagnosis method and device, and relates to the technical field of fault diagnosis. The method comprises offline modeling and online fault diagnosis. For the offline modeling, a model of the belt conveyor system based on subspace identification technology is established; for the online fault diagnosis, initialization parameters are included, fault diagnosis residual values and fault detection thresholds are obtained; online calculation of fault diagnosis monitoring variables is performed, and whether the variable exceeds the fault detection threshold is judged; if the threshold is exceeded, a fault exists, and an alarm is given. The application can diagnose whether the belt conveyor system has a fault in real time, accurately provide alarm information for on-site staff, and provide a decision basis for operation adjustment and equipment operation and maintenance of the production process.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of fault diagnosis, in particular to a mine belt conveyor system fault diagnosis method and device. BACKGROUND

[0002] The belt conveyor is a kind of material conveying equipment, which transports materials in a continuous manner through friction drive. Belt conveyors are widely used in ports, mines, metallurgy, grain, papermaking and other industries. In the field of mining, belt conveyors are mainly used for ore transportation. Under complex geological conditions and harsh environments, mine transportation faces the severe challenges of variable working conditions and large randomness of load, so through the whole process management of mine transportation, the normal operation of the transportation system is ensured. In the process of transporting ore, the belt conveyor system is the most important link. If this link fails, it will directly lead to reduced equipment efficiency and transportation efficiency, and even system shutdown, production interruption, and impact on production. In real production, the belt conveyor often fails due to ore blocking and overloading. If online fault diagnosis of the belt conveyor system is realized to detect these abnormal conditions and prompt production operators and maintenance personnel to intervene in time, it is of great significance to ensure production and equipment management.

[0003] For the problem of ore blocking or overloading when the No. 1 belt conveyor transports ore to the No. 2 belt conveyor, the current detected during the operation of the No. 1 and No. 2 belt conveyors can reflect the amount of ore carried on the belt conveyor from the side, and the current data of the No. 1 and No. 2 belt conveyors can be collected by the electrical drive system of the equipment. Based on the consideration of the system mechanism characteristics, the current data of the No. 1 and No. 2 belt conveyors are fully analyzed to diagnose faults, which is a simple, effective and low-cost fault diagnosis method for belt conveyor systems. And timely alarm for blocking to avoid subsequent escalation of harm.

[0004] The invention patent with application number 202111347306.0 "Coal belt fault detection system" essentially completes the fault determination task based on different training sets by analyzing real-time belt images, and the neural network has no good interpretability, so the diagnosis result often lacks persuasiveness. This method needs at least two identical industrial cameras when implemented, and the device cost is high. The invention patent with application number 202011355753.6 "Vibration signal amplification device for monitoring belt fault" proposes to judge the abnormal vibration of the belt by physically amplifying the fault vibration frequency of the belt during operation, which requires the installation of a vibration sensor, increasing the cost of fault diagnosis; and needs to obtain the service life state parameter of the belt, which cannot be accurately obtained. The above two methods both use relatively complex methods to determine whether the belt conveyor is faulty. SUMMARY

[0005] The technical problem solved by the present application is to provide a mine belt conveyor system fault diagnosis method and device to solve the problems of the prior art, which can diagnose whether there is a fault in the conveying process of the mine belt conveyor in real time, accurately provide alarm information for the production personnel, and provide decision basis for the operation adjustment and equipment operation and maintenance of the production process.

[0006] To solve the above technical problems, the technical solution adopted by the present application is:

[0007] In one aspect, the present application provides a mine belt conveyor system fault diagnosis method, which comprises:

[0008] The current value of the No. 1 belt conveyor is set as the input variable of the subspace identification algorithm, and the current value of the No. 2 belt conveyor is set as the output variable, i.e. the predicted current value of the No. 2 belt conveyor. After initializing the parameters, the subspace identification model of the belt conveyor system is obtained by offline modeling. The extended state space model is established using the fault-free data. The extended observability matrix and Toeplitz matrix are calculated based on the subspace identification technology, so as to obtain the coefficient matrix estimate value.

[0009] The initialization parameters for online diagnosis include the model order, the modeling data set length, the expansion dimension of the data matrix, the input and output variable dimension, the confidence, the fault detection threshold, and the filtering coefficient of the input and output variable. The coefficient matrix estimate value calculated by offline modeling is loaded.

[0010] The current current value data of the No. 1 belt conveyor and the No. 2 belt conveyor are collected and saved online. The collected data is filtered and saved.

[0011] The coefficient matrix estimate value calculated by offline modeling is used to establish an online extended state space model, and the predicted current value and residual value of the No. 2 belt conveyor are calculated. The monitoring quantity is calculated using the residual variance matrix.

[0012] Fault detection is performed by judging whether the monitoring quantity exceeds the fault detection threshold, and an alarm signal is output.

[0013] Further, in offline modeling, the current value of the No. 1 belt conveyor is set as the input variable I1(k) of the subspace identification algorithm, and the predicted current value of the No. 2 belt conveyor is set as the output variable l x is the state vector, W, T, G, and Q are system matrices, k is the sampling time, and the discrete-time state space model is established as follows:

[0014] x(k+1)=Wx(k)+TI1(k)

[0015] ​

[0016] In the matrix sequence, p, f, n, N, I and m represent the number of past data, the number of future data, the system order, the extended dimension of data matrix, the dimension of input variable and the dimension of output variable respectively, where p≥f>n; the input variable Δ(k), the output variable E(k), the future Hankel input matrix Δ f , the past Hankel input matrix Δ p and the future Hankel output matrix E f are defined as follows respectively:

[0017] Δ(k)=[I1(k) I1(k+1) … I1(k+N-1)]∈R l×N

[0018]

[0019]

[0020]

[0021]

[0022] Under the condition of no fault, the extended state space model is expressed as,

[0023]

[0024] where X k is the state sequence and is defined as follows:

[0025] X k =[x(k)x(k+1) … x(k+N-1)]∈R n×N

[0026] where the extended observable matrix Ξ f and the Toeplitz matrix have the following forms respectively:

[0027]

[0028]

[0029] The estimated values of and are obtained based on the subspace identification technology; is the orthogonal space of Ξ f ; the processes of identifying and are as follows:

[0030] Firstly, collect data, and input-output data into an augmented data matrix V f ;

[0031]

[0032] SVD decomposition is performed on the augmented data matrix V f :

[0033]

[0034] where A consists of the first (lf+n) left singular vectors corresponding to the non-zero singular values in ∑, A ⊥ is the remaining (mf-n) left singular vectors corresponding to the zero singular values in ∑; similarly to A and A ⊥ , B and B ⊥ consist of the right singular vectors;

[0035] Let the first mf columns and the last lf columns of A ⊥T be A mf and A lf , respectively, then the extended observable matrix Ξ f and the Toeplitz matrix

[0036] Since Ξfand are composed of the system matrices W, T, G, Q, W, G are directly extracted from the structure of the extended observable matrix Ξ f ; after estimating the system matrices W and G, the system matrices T and Q are estimated from the Toeplitz matrix using the least squares method; thus, the estimated values of the system matrices W, T, G, Q in the state space model of the current relationship of the belt conveyor are obtained.

[0037] In online diagnosis, the current value of the No. 1 belt conveyor is taken as the input variable I1(k), and the extended state space model is established by the estimated values of the coefficient matrices W, T, G, Q, expressed as: The current prediction value of the No. 2 belt conveyor is obtained Thus, the expression of the output residual signal of the system is defined as:

[0038]

[0039] The variance matrix σ r of the residual is calculated, and the variance matrix of the residual is saved;

[0040] The monitoring quantity J is calculated through the obtained residual signal r(k), the monitoring quantity is compared with the fault detection threshold, and thus fault detection is realized, and the specific process is as follows:

[0041] Firstly, a chi-square distribution χ of the fault detection threshold is determined α And a fault detection threshold J is set th , J th = χ α / 2;

[0042] The monitoring quantity J is defined, and the monitoring quantity J is calculated through the residual signal r(k):

[0043]

[0044] Where σ r is the variance matrix of the residual;

[0045] The fault detection method is defined:

[0046]

[0047] The monitoring quantity is compared with the threshold value, so as to realize fault detection.

[0048] On the other hand, the application also provides a mine belt conveyor system fault diagnosis device, which comprises an offline modeling module and an online diagnosis module;

[0049] The offline modeling module is used for offline establishment of a subspace identification model of the belt conveyor system; after initialization of parameters of offline modeling, the module calculates an observable matrix and a Toeplitz matrix, and an estimated value of a coefficient matrix, and provides the estimated value of the coefficient matrix to the online diagnosis module;

[0050] The online diagnosis module is used for online diagnosis of a fault of the belt conveyor system, and receives various parameters output by the offline modeling module; the module online collects current value data of the No. 1 belt conveyor and the No. 2 belt conveyor, performs fault detection, and outputs an alarm prompt.

[0051] The online diagnosis module comprises an initialization module, a data acquisition and filtering module, a fault detection module, and a data storage module;

[0052] The initialization module is used for initialization of an order of the model, setting of a length of a modeling data set, expansion of a dimension of a data matrix, dimensions of input and output variables, a confidence level, a fault detection threshold value, filtering coefficients of the input and output variables, and loading of the estimated value of the coefficient matrix from the data storage module.

[0053] The data acquisition and filtering module is used for acquisition of current current value data of the No. 1 belt conveyor, acquisition of current current of the No. 2 belt conveyor, storage of the acquired data into the data storage module, filtering calculation of the acquired data, and saving of filtered output data into the data storage module;

[0054] The fault detection module is used for calculating the output No. 2 belt conveyor predicted current value, residual value and monitoring quantity; before calculation, historical acquisition data and historical prediction data are read from the data storage module, and the results of the predicted current, residual value and monitoring quantity calculated in the current period are output to the data storage module for storage; whether the calculated monitoring quantity is greater than the fault detection threshold value is judged, if not, it indicates that no fault occurs, and the next time data is acquired for a new round of calculation; if yes, it indicates that a fault occurs, and an alarm signal is output;

[0055] The data storage module is used for storing various variables of the initialization module: initialization model order, modeling data set length, expansion dimension of the data matrix, dimension of the input and output variables, confidence, threshold value, filtering coefficient of the input and output variables, estimated value of the coefficient matrix, storing the current of the No. 1 and No. 2 belt conveyors in data acquisition, filtering module output data and residual signal calculated each time, and providing historical filtering data, historical acquisition data and historical prediction data, monitoring quantity and residual value before calculation of the fault detection module.

[0056] The beneficial effects generated by the above technical scheme are that the mine belt conveyor system fault diagnosis method and device provided by the application can diagnose the whole system in real time based on the current data of the No. 1 and No. 2 belt conveyors, can accurately provide fault diagnosis information for production personnel, the diagnosis result has clear physical meaning, is strong in interpretation, and can provide important basis for production and operation and maintenance decision. The application fully utilizes the current data of the No. 1 and No. 2 belt conveyors to diagnose system faults on the basis of considering system mechanism characteristics, does not need to additionally install sensors, and is a simple, effective and low-cost fault diagnosis method. BRIEF DESCRIPTION OF DRAWINGS

[0057] Figure 1 The flowchart of the mine belt conveyor system fault diagnosis method provided for the embodiment of the application;

[0058] Figure 2 The mine belt conveyor system of the embodiment of the application;

[0059] Figure 3 The flowchart of the online fault diagnosis method provided for the embodiment of the application;

[0060] Figure 4 The structural schematic diagram of the mine belt conveyor conveying process fault diagnosis device provided for the embodiment of the application.

[0061] Figure 5The present application is an embodiment of a mine belt conveyor conveying process output current detection result in a fault-free state; wherein, Figure (5a) is a current value curve of the No. 1 belt conveyor in a fault-free state, Figure (5b) is a comparison curve of the current value and the predicted current value of the No. 2 belt conveyor in a fault-free state, Figure (5c) is a current prediction residual value curve of the No. 2 belt conveyor in a fault-free state, and Figure (5d) is a fault detection result in a fault-free state; Figure (5e) is a local enlarged view of Figure (5b);

[0062] Figure 6 The present application is an embodiment of a mine belt conveyor conveying process No. 1 belt conveyor blockage fault detection result; wherein, Figure (6a) is a current value curve of the No. 1 belt conveyor under the fault, Figure (6b) is a comparison curve of the current value and the predicted current value of the No. 2 belt conveyor under the fault, Figure (6c) is a current prediction residual value curve of the No. 2 belt conveyor under the fault, and Figure (6d) is a fault detection result under the fault.

[0063] Figure 7 The present application is an embodiment of a mine belt conveyor conveying process No. 2 belt conveyor overload fault detection result; wherein, Figure (7a) is a current value curve of the No. 1 belt conveyor under the fault, Figure (7b) is a comparison curve of the current value and the predicted current value of the No. 2 belt conveyor under the fault, Figure (7c) is a current prediction residual value curve of the No. 2 belt conveyor under the fault, and Figure (7d) is a fault detection result under the fault. DETAILED DESCRIPTION

[0064] The specific embodiments of the present application will be further described in detail below in combination with the drawings and examples. The following examples are used to illustrate the present application, but are not used to limit the scope of the present application.

[0065] As shown in Figure 1 , the present embodiment provides a mine belt conveyor system fault diagnosis method, which is specifically described as follows.

[0066] The mine belt conveyor system of the present embodiment is composed of a No. 1 belt conveyor and a No. 2 belt conveyor, as shown in Figure 2 The current of the belt conveyor is integrated into the factory DCS (Distributed Control System). The current value of the No. 1 belt conveyor is taken as an input variable of the subspace identification algorithm, which is denoted as I1(k) at time k, I1(k)∈R l The predicted current value of the No. 2 belt conveyor is taken as an output variable, which is denoted as I2(k) at time k, I2(k)∈R Since p≥f>n, p=10, f=5 are taken in the present embodiment, the system state x(k) is taken as n=2, l=m=1, and N=7.

[0067] Select 2-5 segment I1(k) fluctuation and system normal operation time series data, splicing as normal working condition data set. Each sample of the data set is {I1(k), I2(k)}, the length of the data set is L, and the effective data set length is L-f. Using the data set, according to steps S101-S102 in Figure 1 The subspace identification model of the belt conveyor system is established offline.

[0068] The subspace identification model of the belt conveyor system is established as follows:

[0069] The current value of the No. 1 belt conveyor is taken as the input variable I1(k) of the subspace identification algorithm l , and the predicted current value of the No. 2 belt conveyor is taken as the output variable x is the state vector, W, T, G, and Q are system matrices, k is the sampling time, and the discrete-time state space model is established as follows:

[0070] x(k+1)=Wx(k)+TI1(k)

[0071]

[0072] Wherein the input and output vectors and the past and future Hankel input and output matrices are:

[0073] Δ(k)=[I1(k)I1(k+1) … I1(k+7-1)]∈R 1×7

[0074] E(k)=[I2(k)I2(k+1) … I2(k+7-1)]∈R 1×7

[0075]

[0076]

[0077] The extended state space model is:

[0078]

[0079] Wherein,

[0080]

[0081]

[0082] Then the input and output data are formed into an augmented matrix V5:

[0083]

[0084] For augmented data matrix V f Perform SVD decomposition:

[0085]

[0086] Let A ⊥T If the first 5 columns and the last 5 columns are A5 and a5 respectively, then we have

[0087]

[0088]

[0089] Thus, identify and It is an orthogonal space of Ξ5, from which the estimated values ​​of the coefficient matrices W, T, G, and Q are obtained.

[0090] Therefore, by taking the current value I1(k) of belt conveyor No. 1 as input and using the identified state-space model, the predicted current output value of belt conveyor No. 2 can be obtained. Therefore, the expression for the system's output residual signal is defined as follows:

[0091]

[0092] See Figure 3 The flowchart of an online fault diagnosis method for a belt conveyor system is shown, including the following steps S501 to S507.

[0093] S501, Initialization, including initializing the model's order n, setting the modeling dataset length, the expanded dimension N of the data matrix, the dimensions l and m of the input and output variables, the confidence level α, the fault detection threshold, and the filter coefficients μ of the input and output variables. I1 μ I2 Load the coefficient matrices W, T, G, and Q.

[0094] In this embodiment, the confidence level α is set to 0.05, the input I1(k) is set, and the filter coefficient μ is set. I2 =0.8, μ I1 =0.9. Let p=10, f=5, system state x(k), n=2, l=m=1, N=7. Calculate the fault detection threshold J using the following formula. th :

[0095] J th =γ 0.05 / 2

[0096] S502, collect {I1(k), I2(k)} at current k moment from DCS, read {I1(k-1), I2(k-1)}, {I1(k-2), I2(k-2)}, {I1(k-3), I2(k-3)}, {I1(k-4), I2(k-4)} from DCS history database.

[0097] S503, calculate I2(k) filter value according to the following formula And store:

[0098]

[0099] Calculate I1(k) filter value according to the following formula And store:

[0100]

[0101] S504, calculate predicted output value according to the following formula That is, E' is obtained f And store it.

[0102]

[0103] S505, calculate output residual signal r(k) according to the following formula and store it.

[0104]

[0105] S506, calculate monitoring quantity J according to the following formula and store it.

[0106]

[0107] S507, judge whether the monitoring quantity J is greater than the fault detection threshold J th .

[0108] If not, jump to step S502.

[0109] If yes, alarm "belt conveyor fault", write back to DCS, and jump to step S502.

[0110] In the specific implementation process, if the above fault diagnosis analysis method process can continue, or end the process.

[0111] As described above, the above fault diagnosis method provided by the embodiment of the application can realize fault diagnosis of the belt conveyor system by using the current values of the first and second belt conveyors, and provides fault alarm for the operator.

[0112] The embodiment also provides a mine belt conveyor system fault diagnosis device, as shown in Figure 4As shown, the device comprises an offline modeling module and an online diagnosis module.

[0113] The offline modeling module is used to establish an online diagnosis model of the belt conveyor system. The module provides the coefficient matrix W, T, G, Q obtained by the model to the online diagnosis module.

[0114] The online diagnosis module is used to diagnose the fault of the belt conveyor system. The module receives the parameters output by the offline modeling module. The module collects DCS data online, performs fault detection, and writes the result back to the DCS.

[0115] The online diagnosis module comprises the following sub-modules: an initialization module, a data acquisition and filtering module, a fault detection module, and a data storage module.

[0116] The initialization module is used to initialize the model order n, the modeling data set length L, the parameters p, f, N, l, the confidence level a, load the coefficient matrix W, T, G, Q, and calculate the fault detection threshold J th and the monitoring quantity J.

[0117] The data acquisition and filtering module is used to communicate with the DCS, collect the current k time belt conveyor current value I1(k) from the DCS, collect the current time belt conveyor current value I2(k), and collect historical data. The communication protocol used in this embodiment can be OPC UA (Object Linking and Embedding for Process Control Unified Architecture) or other communication interfaces or APIs (Application Programming Interface) supported by the control system. The module performs filtering calculation on I1(k) and I2(k) to obtain The module reads the historical data from the data storage module before calculation and outputs the filtering result to the data storage module for storage.

[0118] The fault detection module is used to calculate the output 2 belt conveyor predicted current value, the residual value, and the monitoring quantity. The module reads the historical data from the data storage module before calculation and outputs the current cycle calculation result r(k) and the monitoring quantity J to the data storage module for storage; determines whether the monitoring quantity J is greater than the fault detection threshold J th , and writes the alarm information back to the control system.

[0119] The data storage module is used to store the calculated by the data acquisition and filtering module each time and provide ​It is also used to store the calculations made by the fault detection module each time. r(k) and monitoring quantity J are provided, and historical data is provided before this module calculates them.

[0120] Using the apparatus and method of this embodiment, fault detection was performed on the current data of belt conveyors No. 1 and No. 2 from the 14000T ore dressing control system of a certain mine, and the results were as follows: Figure 5 , Figure 6 , Figure 7 The three test results are as follows. The ore is transferred from belt conveyor No. 1 to belt conveyor No. 2, and the current trends of No. 1 and No. 2 should be consistent. If they are inconsistent, a malfunction may have occurred.

[0121] like Figure 5 As shown, this is the detection result under normal and fault-free conditions. Since the monitored quantity in Figure (5d) does not exceed the threshold line, it belongs to the case where the monitored quantity is less than the threshold, which is a fault-free condition.

[0122] like Figure 6 As shown in Figure (6d), the monitored quantity exceeded the threshold at time 4796, indicating a fault. Therefore, it is necessary to analyze the cause of the fault. Figure (6a) shows the current of belt conveyor No. 1. It can be seen that the current value is 265.76 at time 4794 and 279.88 at time 4795, indicating an increasing current. Since the current detected during the operation of belt conveyors No. 1 and No. 2 can reflect the amount of ore they carry, the more ore, the greater the current. Therefore, it can be analyzed that the amount of ore on belt conveyor No. 1 is increasing. Figure (6b) shows the current of belt conveyor No. 2. It can be seen that the current value is 95.18 at time 4794 and 90.36 at time 4795, indicating a decreasing current, indicating that the amount of ore on belt conveyor No. 2 is decreasing. Therefore, it can be concluded that belt conveyor No. 1 experienced a blockage.

[0123] like Figure 7As shown in the figure (7d), the monitoring quantity exceeds the threshold value at 4145, a fault occurs, and thus the cause of the fault needs to be analyzed: Figure (7a) is the current of the No. 1 belt conveyor, and from the figure, it can be seen that the current value is 131.88 at 4143 and 118 at 4144, and the current is in a falling stage. Since the currents detected during the operation of the No. 1 and No. 2 belt conveyors can reflect the amount of ore carried thereon from the side, the more the amount of ore, the greater the current. At this time, it can be analyzed that the amount of ore on the No. 1 belt conveyor is decreasing. Figure (7b) is the current of the No. 2 belt conveyor, and from the figure, it can be seen that the current value is 125.08 at 4143 and 130.78 at 4144, and the current is in a rising stage, and thus the amount of ore on the No. 2 belt conveyor is increasing. It is indicated that the No. 2 belt conveyor is in an overload condition.

[0124] The method and device of the embodiment can diagnose whether the belt conveyor system has a fault in real time, accurately provide alarm information for the on-site staff, and provide a decision basis for the operation adjustment and equipment operation and maintenance of the production process.

[0125] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, but not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for part or all of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope defined by the claims of the present application.

Claims

1. A method of fault diagnosis for a mining belt conveyor system, characterized by: The method comprises offline modeling and online diagnosis; First, set the current value of the No. 1 belt conveyor as the input variable of the subspace identification algorithm, and the current value of the No. 2 belt conveyor as the output variable, i.e. the predicted current value of the No. 2 belt conveyor; after initializing the parameters, obtain the subspace identification model of the belt conveyor system through offline modeling, establish an extended state space model using fault-free data, calculate the extended observability matrix and Toeplitz matrix based on the subspace identification technology, and thus obtain the estimated value of the coefficient matrix; Initialize the parameters for online diagnosis, and collect and save the current current values of the No. 1 belt conveyor and the No. 2 belt conveyor; filter and save the collected data; Using the estimated value of the coefficient matrix calculated through offline modeling, establish an online extended state space model, calculate the predicted current value and residual value of the No. 2 belt conveyor, and calculate the monitoring quantity using the variance matrix of the residual value; Perform fault detection by judging whether the monitoring quantity exceeds the fault detection threshold, and output an alarm signal.

2. The mining belt conveyor system fault diagnostic method of claim 1, wherein: The initialization parameters before offline modeling and the initialization parameters for online diagnosis both include the initialization model order, the set modeling data set length, the extended dimension of the data matrix, the input and output variable dimension, the confidence level, the fault detection threshold, and the filtering coefficients of the input and output variables; The initialization for online diagnosis also includes loading the estimated value of the coefficient matrix calculated through offline modeling.

3. The mining belt conveyor system fault diagnostic method of claim 2, wherein: In the offline modeling, the current value of the first belt conveyor is taken as the input variable I1(k)∈R of the subspace identification algorithm l The predicted current value of the second belt conveyor is taken as the output variable x is the state vector, W, T, G, Q are system matrices, k is the sampling time, and the discrete-time state-space model is established as follows: x(k+1)=Wx(k)+TI1(k) In the matrix sequence, p represents the number of past data, f represents the number of future data, n represents the order of the system, N represents the extended dimension of the data matrix, l is the dimension of the input variable, m is the dimension of the output variable, wherein p≥f>n; the input variable Δ(k), the output variable E(k), the future Hankel input matrix Δ f , the past Hankel input matrix Δ p and the future Hankel output matrix E f are defined as follows, respectively: Δ(k) = [I1(k) I1(k+1)... I1(k+N-1)] e R l×N Under the condition of no fault, the extended state space model is represented as, where X k is a sequence of states, defined as follows: X k = [x(k) x(k+1)... x(k+N-1)] e R n×N where the extended observable matrix Ξ f and Toeplitz matrix have the following forms, respectively: Based on subspace identification techniques and estimates; is the orthogonal space of Ξ f .

4. The mining belt conveyor system fault diagnostic method of claim 3, wherein: Based on the subspace identification technique, the estimates of and are obtained as follows: First, collect data, input and output data into an augmented data matrix V f ; The augmented data matrix V f SVD decomposition is performed: where A consists of the first (lf + n) left singular vectors corresponding to the non-zero singular values in ∑, A ⊥ is the remaining (mf - n) left singular vectors corresponding to the zero singular values in ∑; and similarly to A and A ⊥ , B and B ⊥ consist of the right singular vectors; Let the front mf column and the back lf column of A mf and A lf respectively, then the extended observable matrix Ξ f and Toeplitz matrix Since Ξ f and are composed of system matrices W, T, G, Q, W and G are directly extracted from the structure of the extended observation matrix Ξ f ; after estimating the system matrices W and G, the system matrices T and Q are estimated from the Toeplitz matrix by using the least square method; thus the estimated values of the system matrices W, T, G, Q in the state space model of the current relationship of the belt conveyor are obtained.

5. The mining belt conveyor system fault diagnostic method of claim 4, wherein: In the online diagnosis, the current value of the No. 1 belt conveyor is taken as an input variable I1(k), and an extended state space model is established by using the estimated values of the identified coefficient matrices W, T, G and Q, and is expressed as: The current prediction value of the No. 2 belt conveyor is obtained Thus, the output residual error signal expression of the system is defined as: a variance matrix σ of the computed residuals r and save the variance matrix of the residuals; Calculate the monitoring quantity J using the obtained residual signal r(k), compare the monitoring quantity with the fault detection threshold, and thus realize fault detection.

6. The mining belt conveyor system fault diagnostic method of claim 5, wherein: The specific process of the fault detection is as follows: First, the chi-square distribution χ of the fault detection threshold is determined α , and the fault detection threshold J is set th , J th = χ α / 2; Define the monitoring quantity J and calculate the monitoring quantity J using the residual signal r(k): where σ r is the variance matrix of the residuals; Define the fault detection method as follows: Compare the monitoring quantity with the threshold to obtain the fault detection result.

7. A mining belt conveyor system fault diagnosis device for implementing the mining belt conveyor system fault diagnosis method according to claim 1, characterized by: The device comprises an offline modeling module and an online diagnosis module; The offline modeling module is used to offline establish a subspace identification model of a belt conveyor system; after initializing the parameters for offline modeling, the module establishes an extended state space model using fault-free data, calculates the observability matrix and Toeplitz matrix based on the subspace identification technology, and obtains the estimated value of the coefficient matrix, which is provided to the online diagnosis module; The online diagnosis module is used to diagnose faults of the belt conveyor system online, and receives the parameters output by the offline modeling module; after initializing the parameters online, the module establishes an online extended state space model using the estimated value of the coefficient matrix calculated through offline modeling, calculates the predicted current value and residual value of the No. 2 belt conveyor, calculates the monitoring quantity using the variance matrix of the residual value, performs fault detection by judging whether the monitoring quantity exceeds the fault detection threshold, and outputs an alarm prompt.

8. The mining belt conveyor system fault diagnostic apparatus of claim 7, characterized in that: The online diagnosis module comprises an initialization module, a data collection and filtering module, a fault detection module, and a data storage module; The initialization module is used for initializing the order of the model, setting the length of the modeling data set, the extended dimension of the data matrix, the dimension of the input and output variables, the confidence, the fault detection threshold, the filtering coefficient of the input and output variables, and loading the estimated value of the coefficient matrix from the data storage module; The data acquisition and filtering module is used for acquiring the current value data of the No. 1 belt conveyor at the current time, and acquiring the current of the No. 2 belt conveyor at the current time, and storing the acquired data into the data storage module; The acquired data is filtered and calculated, and the filtered output data is saved into the data storage module; The fault detection module is used for calculating the predicted current value of the No. 2 belt conveyor, the residual value and the monitoring quantity; Before calculation, the historical acquisition data and the historical prediction data are read from the data storage module, and the results of the predicted current, the residual value and the monitoring quantity calculated in the current period are output to the data storage module for storage; It is judged whether the calculated monitoring quantity is greater than the fault detection threshold, if not, it means that no fault occurs, and the next time data is acquired for a new round of calculation; If yes, it means that a fault occurs, and an alarm signal is output; The data storage module is used for storing various variables of the initialization module: the order of the initialized model, the length of the modeling data set, the extended dimension of the data matrix, the dimension of the input and output variables, the confidence, the threshold, the filtering coefficient of the input and output variables, the estimated value of the coefficient matrix, the currents of the No. 1 and No. 2 belt conveyors in the data acquisition, the filtering module output data and the residual signal calculated each time, and providing the historical filtering data, the historical acquisition data and the historical prediction data, the monitoring quantity and the residual value before the calculation of the fault detection module.

Images