CN103671190A - Intelligent early stage on-line fault diagnosis system of mine fan - Google Patents

Abstract

An intelligent early mining fan online fault diagnosis system belongs to the field of early fault diagnosis of equipment. Aiming at the importance of the mine ventilator to coal mine safety and the complexity of early fault diagnosis, the present invention can not only accurately extract the operating state characteristics of the mine ventilator, but also identify subtle changes in the operating state of the mine ventilator. Because the current mining fan fault diagnosis device can only realize fault early warning, but cannot make early fault diagnosis. The invention uses clustering and singular value decomposition methods to preprocess the state information extracted by sensors and transmitters and extract the state characteristics of the mine ventilator without being disturbed by noise. Use the support vector machine model in fault identification and classification, and use the data of each diagnosis to continuously enrich and update the training and learning samples of the support vector machine, so that the model contains more information and achieves the accuracy of early faults of mine ventilators , fast and smart on purpose.

Description

An Intelligent Early Online Fault Diagnosis System for Mine Ventilators

technical field

The invention relates to an online fault diagnosis system for an intelligent early mine ventilator, which belongs to the field of equipment fault diagnosis. It diagnoses and identifies the state and state changes of a mine ventilator by analyzing and calculating the state signal of a mine ventilator, especially for the mine ventilator. The fault diagnosis of mine ventilator in harsh environment belongs to the field of early fault diagnosis of mine ventilator. the

Background technique

Mine ventilator is an important equipment for continuously conveying fresh air, diluting dust, and toxic and harmful gases to the mine. casualties. In order to ensure the safe, reliable and stable operation of mine ventilators, early fault diagnosis of mine ventilators is required. the

Due to the harsh working environment and long-term operation of mining fans, they are prone to failures, with many types of failures, weak characteristic signals and strong background noise. Appropriate fault diagnosis method and hardware platform. The traditional monitoring system of mine ventilator can only detect the middle and late faults of mine ventilator, but cannot diagnose early faults, and the misdiagnosis rate is high and the reliability is poor. The earlier the fault is found, the lower the maintenance cost of mine ventilator. Mine ventilators need an accurate and objective fault diagnosis system to realize reliable early fault diagnosis of mine ventilators and ensure the safety of mine ventilators. At present, there is no mature early fault diagnosis system for mining ventilators in China, so it is of great practical significance to research and develop an intelligent online fault diagnosis system for early mining ventilators. the

Contents of the invention

The purpose of the present invention is to develop an intelligent monitoring system aimed at the deficiency that the current monitoring system of mining fans can only detect the middle and late failures of mining fans, but cannot diagnose the early failures of mining fans. An online fault diagnosis system for early mine ventilators, which is used for early fault diagnosis of mine ventilators. The state signal of the mine ventilator is measured by the sensor, and the state signal of the equipment is transmitted from the lower computer PLC to the upper computer. The state signal is preprocessed through the clustering algorithm, the noise signal is eliminated, and the singular value decomposition is used to extract the mine ventilator. State features, signal features are input into the support vector machine model to identify and diagnose early failures of mine ventilators. the

An intelligent online fault diagnosis system for early mining ventilators uses clustering and singular value decomposition as signal feature extraction methods, and uses support vector machines to identify and diagnose faults. The signal feature extraction method includes the following steps:

S1. Use the clustering method to process the data collected by the data acquisition module (known) x(θ)(k)(k=1,2...N), N sampling points, θ is the signal code, representing temperature, vibration, Negative pressure, carbon monoxide concentration, methane concentration, voltage, current, θ=1,2...ss is the total number of signals, set the number of clusters n, after analysis, the cluster containing the most signals is the desired signal, and the signals contained in the remaining clusters As an outlier, remove the outlier in the signal to obtain the cluster K _i (θ)(x) containing the most data, K _i (θ)(x)=[x _i1 , x _i2 ,… x _im ], i is the ordinal number of the divided clusters, i=1,2...n, m is the number of signal points of the clusters containing the most data, m≤N;

S2. Construct Hankel matrix A for K _i (θ)(x) in S1,

$\mathrm{<span\; class="notranslate">\; A</span>}<span\; class="notranslate">\; =</span>\left[\begin{array}{cccc}{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}\mathrm{<span\; class="notranslate">\; 1</span>}}& {\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}\mathrm{<span\; class="notranslate">\; 2</span>}}& <span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span>& {\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>}\\ {\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}\mathrm{<span\; class="notranslate">\; 2</span>}}& {\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}\mathrm{<span\; class="notranslate">\; 3</span>}}& <span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span>& {\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; ij</span>}}\\ <span\; class="notranslate">\; .</span>& <span\; class="notranslate">\; .</span>& & <span\; class="notranslate">\; .</span>\\ <span\; class="notranslate">\; .</span>& <span\; class="notranslate">\; .</span>& <span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span>& <span\; class="notranslate">\; .</span>\\ <span\; class="notranslate">\; .</span>& <span\; class="notranslate">\; .</span>& & <span\; class="notranslate">\; .</span>\\ {\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; l</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; )</span>}& {\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; l</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>}& <span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span>& {\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>}\end{array}\right]<span\; class="notranslate">\; ;</span>$

S3. Singular value decomposition is performed on the matrix A in S2, and the first singular value λ(θ) is taken as the signal feature;

S4. Construct X={λ(1),λ(2)...λ(s)} in S3 as the input of the support vector machine, output Y={y ₁ ,y ₂ ...y _s }, Where y _θ is equal to 0 or θ, if there is a fault in a certain position, the corresponding code θ will be output, if it is normal, 0 will be output, and the fault location will be determined according to the code.

S5. Record the training input and output of the support vector machine model in the form of a table. The horizontal row of the table represents the output of the support vector machine, and the vertical row represents the serial number of the sample. The diagnostic output after the support vector machine training is completed must be compared with the table. Compare, if the output is the same as the table, the state of the mine ventilator is the same as the sample;

An intelligent online fault diagnosis system for early mining ventilators, including a host computer, a slave computer, a signal acquisition module, a temperature sensor, a vibration sensor, a negative pressure sensor, a methane sensor, a carbon monoxide sensor, and an ACR series network multifunctional power meter; The two sensors at the measuring point are connected to the analog-to-digital conversion module SM331 through shielded wires, and SM331 converts the analog quantity into a digital quantity and inputs it into the CPU314; the voltage and current of the mine fan motor are measured by the ACR series network multifunctional power meter, It is input into CPU314 through RS485; the hardware of the early fault diagnosis system for each mining fan is two sets, which are mutually redundant; CPU314 inputs the measured signal to the upper computer, and the upper computer calculates the two sensors of the same measured point The average value is used as the signal of the measuring point; the two PLCs monitor each other, and the PLC of the working fan periodically sends instructions to the other PLC to reset the timer. If it cannot be reset, a fault alarm is issued; the mining fan is abnormal or According to the status of the fan, the corresponding control strategy is made to control the start and stop of the fan, automatic shutdown, and the opening and closing of the damper; the upper computer uses WinCC as the configuration software to complete the human-computer interaction, which can display the status parameters of the fan and display Temperature, motor voltage and current, vibration, negative pressure, air volume, carbon monoxide content, methane content, performance curve of mine fan, damper status and oil station status, and can realize control input, input temperature, vibration, negative pressure, air volume , methane content, and carbon monoxide content alarm thresholds; control commands can be input on the upper computer WinCC through the keyboard to control the opening and closing of the damper and the start and stop of the fan; the temperature sensor Pt100 is installed on both sides of the bearing seat and next to the stator coil of the motor , All bearing housings are installed with integrated vibration transmitters in the horizontal and vertical directions; the output point of CPU314 is connected to the relay, and the relay controls the power switch of the damper motor, fan motor and oil station motor. the

This system integrates the fault diagnosis algorithm in WinCC, and only one configuration software can complete the fault diagnosis and control of mine ventilator. the

The intelligent early mine fan fault online diagnosis system proposed by the present invention has the following advantages:

1. Realize the online real-time intelligent early fault diagnosis of mine ventilator, which can accurately extract the state characteristics of mine ventilator, and can diagnose as early as possible whether there is a fault in the mine ventilator, the type and location of the fault. the

2. In view of the harsh working environment and high reliability requirements of mine ventilators, all hardware of the fault diagnosis system adopts dual redundancy measures to ensure the reliable and accurate operation of the early fault diagnosis system of mine ventilators. the

3. All fault diagnosis algorithms are integrated in the upper computer configuration software WinCC, and one software completes the diagnosis and control of mine fan faults, which increases the reliability of the system. the

Description of drawings

Fig. 1 schematic diagram of the hardware of this system;

Fig. 2 is the fault diagnosis flow chart of this system;

Detailed ways

The present invention is described in detail below in conjunction with accompanying drawing and example:

The hardware structure of the system is shown in Figure 1. It is mainly composed of upper computer, lower computer, data acquisition module, control input module, control output module and sensors. for redundancy. Profibus protocol is used to connect between the upper computer and the lower computer, and between the lower computer and the lower computer. the

The fault diagnosis process of the system is shown in Figure 2. Set the number of clusters n for clustering and classification, set the window length l of the singular value, and input the signal collected by the data acquisition module to the PLC of the lower computer, and then transmit it to the upper computer. The average value of the two signals of the node is used as the signal of the measuring point, the abnormal points are removed by clustering, the state characteristics of the signal are extracted by singular value decomposition, and the result of the singular value decomposition is used as the training and diagnosis input of the support vector machine model, and the output The state of the mine ventilator, if there is a fault, it will alarm and take corresponding control automatically. Taking a mining fan as an example, the basic process is as follows:

1. Set the addresses of No. 1 and No. 2 PLCs to 1, 2, and the communication addresses of No. 1 upper computer and No. 2 upper computer to 3 and 4 respectively;

2. Install two Pt100 temperature sensors on each bearing and motor stator of the mining fan, and install two integrated vibration transmitters as vibration sensors on the horizontal and vertical directions of each bearing seat respectively. Install two carbon monoxide sensors and two methane sensors at the air outlet, respectively install two B0300 industrial-grade micro-pressure transmitters at the air inlet and outlet of the fan as negative pressure sensors, and install them on the PT distribution cabinet of the mine fan Two pieces of ACR series network multifunctional power meters measure the voltage and current of the mine fan;

3. Except for the two ACR series network multifunctional power meters, all the sensors are connected to Siemens’ modulus replacement module SM331 through shielded wires, and the addresses of the two ACR series network multifunctional power meters are set to 10 and 11 respectively. Input to P0 port of Siemens CPU314 through RS485 communication protocol;

4. Connect No. 1 PLC and No. 1 host computer, No. 2 PLC and No. 2 host computer, No. 1 PLC and No. 2 PLC through DP line;

5. On the host computer, enter the threshold values of mine fan temperature, vibration, negative pressure, air volume, carbon monoxide content, methane content, motor voltage and motor current into WinCC through the keyboard, and immediately alarm if the threshold is exceeded;

6. Set the number of divisions to 3, the collected signal is x(θ)(k), θ is the signal code, the θ of bearing 1 temperature is 1, the θ of motor stator temperature is 2, and the θ of bearing 3 temperature is 3. The θ of the temperature of bearing 4 is 4, the θ of the temperature of bearing 5 is 5, the θ of the horizontal and vertical vibration of bearing 1 is 6 and 7, the θ of the horizontal and vertical vibration of bearing 2 is 8 and 9, the level of bearing 3 The θ of the horizontal and vertical vibration of the bearing 3 is 10 and 11, the θ of the horizontal and vertical vibration of the bearing 3 is 12 and 13, the θ of the horizontal and vertical vibration of the bearing 4 is 14 and 15, the negative pressure of the air inlet and the negative pressure of the air outlet θ are respectively 16 and 17, theta of carbon monoxide concentration is 18, theta of methane concentration is 19, theta of voltage is 20, and theta of current is 21, the number of sampling points k=1,2...512, using cluster analysis x(θ) (k), get the cluster K _i , K _i =[x _i1 ,x _i2 ,…x _im ], i=1,2,3, m≤N;

7. Use SVD to analyze the cluster K _i that contains the most data, and construct the Hankel matrix A of the cluster K _i :

$\mathrm{<span\; class="notranslate">\; A</span>}<span\; class="notranslate">\; =</span>\left[\begin{array}{cccc}{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}\mathrm{<span\; class="notranslate">\; 1</span>}}& {\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}\mathrm{<span\; class="notranslate">\; 2</span>}}& <span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span>& {\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>}\\ {\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}\mathrm{<span\; class="notranslate">\; 2</span>}}& {\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}\mathrm{<span\; class="notranslate">\; 3</span>}}& <span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span>& {\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; ij</span>}}\\ <span\; class="notranslate">\; .</span>& <span\; class="notranslate">\; .</span>& & <span\; class="notranslate">\; .</span>\\ <span\; class="notranslate">\; .</span>& <span\; class="notranslate">\; .</span>& <span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span>& <span\; class="notranslate">\; .</span>\\ <span\; class="notranslate">\; .</span>& <span\; class="notranslate">\; .</span>& & <span\; class="notranslate">\; .</span>\\ {\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; l</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; )</span>}& {\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; l</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>}& <span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span>& {\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>}\end{array}\right]$

Among them, l is the window length, and 1<l<m, the order of matrix A (k=N-l+1). After obtaining the trajectory matrix A, it is necessary to find the singular value of A. X=AA ^T , then X is an l×l matrix, and the eigenvalues of X are obtained: λ ₁ , λ ₂ , λ ₃ ...λ _d , take the square root That is, the singular value (i≤d) of the trajectory matrix A is used as the signal feature. If none of the eigenvalues of the matrix X is zero, then d=l.

8. Take the first singular value λ(θ) ₁ as the training input of the support vector machine model, the input vector is X={λ(1) ₁ ,λ(2) ₁ ,…λ(21) ₁ }, a certain If there is a fault in the position, the corresponding signal code will be output, and if it is normal, 0 will be output to establish a support vector machine model;

9. Record the input and output of the support vector machine in the form of a table. When diagnosing, compare the output result with the table to ensure the reliability of the diagnosis result. The ordinate of the table is the training sample number, and the abscissa is the output of the support vector machine model. For an SVM with only one non-zero output, a failure is determined. If there are two or more outputs that are not 0, compared with the table, the state represented by the sample with the same output in the table is the state of the mine ventilator. The test output 1004000000000000000000 is the same as the output 100400000000000000000 of the bearing 4 failure in the table, so it is determined to be the bearing 4 failure;

10. According to the diagnosis result, the status of the mine ventilator is displayed on the upper computer. If there is any abnormal status or failure, it will immediately call the police and take control measures. the

Claims (3)

1. the early stage mine fan online system failure diagnosis of intelligence, usings cluster and singular value decomposition as the extracting method of signal characteristic, utilizes support vector machine identification and tracing trouble, it is characterized in that: signal characteristic extracting methods wherein comprises the following steps:

S1. utilize data x (θ) that the method processing data acquisition module of cluster gathers (k) (k=1,2 ... N), N sampled point, θ is signal code, representation temperature, vibration, negative pressure, carbonomonoxide concentration, methane concentration, voltage, electric current, θ=1,2 ... s s is signal sum, set number of clusters n, after analysis, comprise signal maximum bunch for required signal, remaining bunch concentrates the signal comprising as being abnormity point, remove the abnormity point in signal, obtain comprising bunch collection K that data are maximum _i(θ) (x), K _i(θ) (x)=[x _i1, x _i2... x _im], i is the ordinal number of divided bunch collection, i=1,2 ... n, m is that the signal that comprises bunch collection that data are maximum is counted, m≤N;

S2. to the K in S1 _i(θ) (x) structure Hankel matrix A,

S3. the matrix A in S2 is done to singular value decomposition, getting first singular value λ (θ) is signal characteristic;

S4. with the λ in S3 (θ) structure X={ λ (1), λ (2) ... λ (s) } be the input of support vector machine, output Y={y ₁, y ₂y _s, y wherein _θequal 0 or θ, breaking down in certain position, exports corresponding code θ, normally exports 0, according to code, determines location of fault.

S5. with the form of form, record the training input and output of supporting vector machine model, form walk crosswise the output that expresses support for vector machine, the sequence number of perpendicular line display sample, support vector machine after having trained diagnosis output all to compare with this form, if export identically with form, the state of mine fan is identical with sample.

2. the early stage mine fan online system failure diagnosis of intelligence, its feature: comprise upper-position unit, lower-position unit, signal acquisition module, temperature transducer, vibration transducer, B/P EGR Back Pressure Transducer EGR, methane transducer, carbon monoxide transducer, ACR series network multifunctional power meter; Two sensors of same measured point are connected to analog-to-digital conversion module SM331 by shielding wire, and SM331 is converted into digital quantity analog amount, is input in CPU314; The voltage and current of mine fan motor is measured by ACR series network multifunctional power meter, by RS485, is input in CPU314; The hardware of every mine fan Incipient Fault Diagnosis system is two covers, each other redundancy; CPU314 is input to upper-position unit by measured signal, and upper-position unit calculates the mean value of two sensors of same measured point as the signal of this measuring point; Two PLC monitor mutually, and the PLC of working fan periodically sends instruction to another PLC, and reset timer, if can not reset, sends malfunction alarm; There is the state of abnormal or fault simulation blower fan and make corresponding control strategy in mine fan, start and stop, automatic pouring machine, the air door of controlling blower fan open and close; Upper-position unit utilizes WinCC as configuration software, finishing man-machine interaction, both can show the status parameter of blower fan, the performance curve of the electric current and voltage of showing temperature, motor, vibration, negative pressure, air quantity, carbon monoxide content, methane content, mine fan, air door state and petrol station state, can realize control inputs again, the alarm threshold value of input temp, vibration, negative pressure, air quantity, methane content, carbon monoxide content; By keyboard input control order on upper station of WinCC, control the switching of air door, the start and stop of blower fan; Temperature transducer Pt100 is arranged on respectively the both sides of bearing support and the stator coil side of motor, and all bearing supports are all installed integral type vibration transmitters in the horizontal and vertical directions; CPU314 output point connects relay, Control damper motor, fan motor and petrol station electric power switch.

3. the early stage mine fan online system failure diagnosis of intelligence, is characterized in that: fault diagnosis algorithm is integrated in WinCC, and only configuration software can complete fault diagnosis and the control of mine fan.

Images