On-line monitoring and diagnosis method and diagnosis system for blockage approaching state of vibrating screen

Abstract

The invention provides an online monitoring and diagnosing method and system for the blockage approaching state of a vibrating screen. The method comprises the steps that vibration acceleration information of three different positions on the vibrating screen is measured through an acceleration sensor; torque information of the eccentric rotating shaft is measured through a torque sensor; the collected torque information is preprocessed, diagnosis is carried out through a torque diagnosis model, and a diagnosis result is obtained; establishing a vibration acceleration diagnosis model; inputting the collected vibration acceleration information of the three different positions into a vibration acceleration diagnosis model for diagnosis to obtain a diagnosis result; and comparing the results of the two diagnosis models, performing cross identification and judging to obtain the running state of the vibrating screen. The method effectively diagnoses the blockage approaching state of the vibrating screen, the diagnosis precision is low, and the monitoring and diagnosis capability of the system is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent agricultural machinery, and in particular to an online monitoring and diagnosis method and a diagnosis system for a blocking state of a vibrating screen. Background Art

[0002] As an important screening equipment, the vibrating screen plays an indispensable role in agriculture and a wider range of industrial fields. The vibrating screen is a device that uses vibration force to screen and grade materials. It transmits the vibration force to the screen box and the screen through the rotational motion generated by the vibrating motor, so that the material vibrates back and forth on the screen, thereby separating or classifying the particles by size. The working principle of the vibrating screen is based on physical vibration. The vibration of the screen makes the particles larger than the screen hole remain on the screen surface, and the particles smaller than the screen hole fall through the screen hole, thereby achieving the purpose of screening.

[0003] Vibrating screens are often used in agricultural machinery. As the input volume of the vibrating screen continues to increase, the clogging problem becomes more and more serious, greatly reducing the operating efficiency. According to relevant surveys, due to factors such as field crop humidity and density, feed volume changes, and vibrating screen vibration frequency, the conveying roller under the vibrating screen is often blocked due to increased feed volume or increased material humidity. Therefore, it is necessary to monitor the clogging state of the vibrating screen online and provide an online monitoring and diagnosis method for the clogging state of the vibrating screen.

[0004] At present, the monitoring of vibrating screens often adopts the method of directly measuring the acceleration and the amount of vibration screens, which can produce certain effects. However, the accuracy of the vibration screen blocking state diagnosed by acceleration and amount of vibration screens is low and has a certain hysteresis, which often cannot achieve the purpose of early warning. Summary of the invention

[0005] In response to the above technical problems, one of the purposes of one embodiment of the present invention is to provide an online monitoring and diagnosis method and diagnostic system for the blocking state of a vibrating screen. The advantage of the present invention is that it can monitor the parameter information of the vibrating screen under different working conditions online, solve the problem that the existing diagnostic system cannot effectively diagnose the blocking state of the vibrating screen and has low diagnostic accuracy, improve the monitoring and diagnostic capabilities of the system, and reduce the dependence on manual diagnostic experience.

[0006] The present invention achieves the above technical objectives through the following technical means.

[0007] An online monitoring and diagnosis method for the blocking state of a vibrating screen comprises the following steps:

[0008] The acceleration sensor is used to measure the vibration acceleration information at three different positions on the vibrating screen; the torque sensor is used to measure the torque information of the eccentric shaft;

[0009] The collected torque information is preprocessed and diagnosed through a torque diagnosis model to obtain a diagnosis result;

[0010] Establish vibration acceleration diagnostic model;

[0011] The vibration acceleration information collected at three different positions is input into the vibration acceleration diagnosis model to obtain the diagnosis result;

[0012] The results of the two diagnostic models are compared and cross-identified to determine the operating status of the vibrating screen.

[0013] Furthermore, the collected torque information is preprocessed, specifically:

[0014] The four characteristic parameters extracted from the torque information are: first-order difference ΔD c , relative torque ratio R c , difference accumulation sum ∑A c , second-order difference Δ 2 D c , specifically:

[0015]

[0016] Where: x c is the period of the c-th pulse period sequence; x c+1 is the period of the c+1th pulse period sequence; c is the pulse period sequence, c=1,2,...N, N is the total number of pulse period sequences;

[0017] ΔD c is the first-order difference value of the cth pulse period sequence. The first-order difference of the pulse period sequence describes the instantaneous change direction of the vibration screen torque;

[0018] R c is the relative torque ratio of the cth pulse cycle sequence;

[0019] x ref,c is the torque value measured in the cth pulse cycle sequence;

[0020] ∑A c is the first-order difference value ΔD of the c-th pulse cycle sequence c sum;

[0021] Δ 2 D c is the second-order difference value of the c-th pulse period sequence, Δ 2 D c The value represents the magnitude of the increase or decrease in the drum speed, and is used to reflect the direction of torque change;

[0022] ΔD c+1 is the first-order difference value of the c+1th pulse period sequence;

[0023] The four feature parameters are fused as follows:

[0024]

[0025] in:

[0026] I is the comprehensive information after information fusion of the four characteristic parameters;

[0027] a1 is the first-order difference ΔD c The weight of a2 is the relative speed ratio R c The weight of a3 is the difference cumulative sum ∑A c The weight of; a4 is the second-order difference Δ 2 D c The weight of

[0028] ΔD ref It is the maximum value of the first-order difference value when the vibrating screen is blocked;

[0029] R ref It is the maximum value of the relative speed ratio when the vibrating screen is blocked;

[0030] ∑A ref is the first-order difference ΔD when the vibrating screen is blocked c The maximum value of the sum of the terms of ;

[0031] Δ 2 D ref It is the maximum value of the second-order difference when the vibrating screen is blocked.

[0032] Furthermore, the torque diagnosis model is used to diagnose and obtain the diagnosis results, which are as follows:

[0033] Set the maximum value of comprehensive information I Max and minimum value I Min ; when I Min When I Min <I<I Max When I>I Max When , the vibrating screen is in a blocked state.

[0034] Furthermore, a vibration acceleration diagnostic model is established, specifically:

[0035] Three acceleration signals are obtained through experiments, each of which is a continuous signal and has a signal segment of a normal state, a signal segment of a jam state, and a signal segment of a jam state;

[0036] Fusion of the feature matrices of the three acceleration signals into a feature fusion matrix ​

[0037] The fusion matrix Each signal sample segment is manually labeled, and the labeled fusion matrix is ​​used as the input of the first BP neural network, and the output is the label, and the first BP neural network model is obtained through training;

[0038] The fusion matrix After the feature vector of the jth feature increases and decreases by 10% respectively, it is input into the first BP neural network model to obtain the predicted increase label value and predict reduced label values

[0039] Using predictions to increase label value and predict reduced label values By using the dimensionality reduction algorithm and the weighted average influence value algorithm, the eigenvector matrix containing n characteristic indicators is Reduce the dimension to a feature matrix containing only z feature indicators z <n;

[0040] The feature matrix It is divided into a test set and a training set, and used as the input layer of the second BP neural network. The number of neurons is set to z, and the number of neurons in the hidden layer is 2*(z+1) / 3. The model is trained with RMSE as the indicator until RMSE is less than 1e-04, and the vibration acceleration blocking diagnosis model is obtained.

[0041] Furthermore, the feature matrices of the three acceleration signals are fused into a feature fusion matrix Specifically:

[0042] The three acceleration signals are preprocessed and feature extracted to extract time domain features, frequency domain features and time-frequency domain features;

[0043] The time domain, frequency domain, and time-frequency domain of each acceleration signal are spliced ​​and fused to obtain the spliced ​​features; the spliced ​​feature of the first acceleration signal is O1, the spliced ​​feature of the second acceleration signal is O2, and the spliced ​​feature of the third acceleration signal is O3; the spliced ​​features are forward normalized to obtain forward normalized matrices A1, A2, and A3, where A1 is the matrix of O1 after forward normalization, A2 is the matrix of O2 after forward normalization, and A3 is the matrix of O3 after forward normalization, as shown below:

[0044]

[0045] Where: m represents the number of signal sample segments, n represents the number of features, represents the jth feature of the i-th signal sample segment; a represents the sensor number, a=(1, 2, 3), when a=1, represents the jth feature of the i-th signal sample segment in the first acceleration signal; when a=2, represents the jth feature of the i-th signal sample segment in the second acceleration signal; when a=3, represents the jth feature of the i-th signal sample segment in the third acceleration signal; i = 1, 2, ..., m; j = 1, 2, ..., n;

[0046] Calculate the proportion of the eigenvalue of the i-th signal sample fragment under the j-th feature of each signal to the sum of the eigenvectors

[0047] Calculate the lower uncertainty measure E of the j-th eigenvector of each signal according to the proportionality coefficient k a j :

[0048]

[0049] Where k>0;

[0050] The weight of the jth feature for:

[0051]

[0052] Fusion of the feature matrices of the three acceleration signals into a feature fusion matrix

[0053]

[0054] in The jth feature of the fused i-th signal sample segment, is the weight of the jth feature of the first acceleration signal; is the weight of the jth feature of the second acceleration signal; is the weight of the jth feature of the third acceleration signal.

[0055] Furthermore, by adopting the dimensionality reduction algorithm and the weighted average influence value algorithm, the eigenvector matrix containing n characteristic indicators is Reduce the dimension to a feature matrix containing only z feature indicators Specifically:

[0056] The weighted average impact value WMIV calculation formula is as follows:

[0057]

[0058] in, The predicted label is obtained by increasing the i-th sample under the j-th input variable by 10%; The predicted label is obtained after reducing the i-th sample by 10% under the j-th input variable; Represents the ratio of different predicted labels under the j-th input variable to the total number of labels of the feature column after all samples increase by 10%; Represents the ratio of different predicted labels under the j-th input variable to the total number of labels of the feature column after all samples are reduced by 10%;

[0059] Calculating WMIV j average value WMIV j is the weighted average impact value of the jth input variable;

[0060] According to the judgment formula: WMIV j ≥Mean, we get z eigenvectors, that is, we get z eigenvectors from n eigenvectors.

[0061] The eigenvector matrix containing n eigenindices Reduce the dimension to a feature matrix containing only z feature indicators

[0062]

[0063] Furthermore, the results of the two diagnostic models are compared and cross-identified to determine the operating status of the axial flow separation device, specifically:

[0064] If the results of the two diagnostic models are the same, the diagnostic result is output; if the results of the two diagnostic models are different, a prompt alarm sound is issued and manual auxiliary diagnosis is performed.

[0065] A system for online monitoring and diagnosis of the blocking state of a vibrating screen, comprising a collection module, a diagnosis module and an execution module;

[0066] The acquisition module includes a sensor and a signal acquisition unit. The sensor includes an acceleration sensor and a torque sensor. The acceleration sensor is installed on the vibrating screen to measure the vibration acceleration; the torque sensor is installed on the eccentric shaft of the vibrating screen to measure the torque of the shaft.

[0067] The diagnostic module includes a rotation speed diagnostic model and a vibration acceleration diagnostic model. The acquisition module inputs the acquired rotation speed information into the rotation speed diagnostic model, and the acquisition module inputs the acquired acceleration into the vibration acceleration diagnostic model. The operating state of the axial flow separation device is determined by comparing and cross-identifying the diagnostic results of the two models.

[0068] The execution module includes a human-machine interface and an acoustic and visual alarm device. The human-machine interface is used for visualizing real-time sensor signal data and storing and displaying diagnostic results. The acoustic and visual alarm device is used for issuing fault warnings.

[0069] The beneficial effects of the present invention are:

[0070] 1. The online monitoring and diagnosis method for the blocking state of the vibrating screen described in the present invention can monitor the parameter information of the vibrating screen under different working conditions in real time, including the torque of the vibrating screen, vibration information of each part, etc., and display the relevant parameter curve in the upper computer interface in real time. If a blocking fault occurs, the current state will be judged and an early warning will be issued.

[0071] 2. The online monitoring and diagnosis method of the blocking state of the vibrating screen described in the present invention solves the problem that the existing diagnostic system cannot effectively diagnose the blocking state of the vibrating screen and has low diagnostic accuracy, improves the monitoring and diagnostic capabilities of the system, and reduces the dependence on manual diagnostic experience.

[0072] 3. The online monitoring and diagnosis method of the blocking state of the vibrating screen described in the present invention obtains the result through the combined diagnosis and judgment of the torque information judgment method and the vibration acceleration diagnosis model, thereby further improving the accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0073] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. The drawings described below are some embodiments of the present invention. For ordinary technicians in this field, it is obvious that other drawings can be obtained based on these drawings without paying any creative work.

[0074] Figure 1 The present invention is a flowchart of the method for online monitoring and diagnosis of the blocking state of the vibrating screen.

[0075] Figure 2 This is a flow chart of the speed diagnosis model based on information fusion judgment described in the present invention.

[0076] Figure 3 This is a flow chart of the vibration acceleration signal fault diagnosis model described in the present invention.

[0077] Figure 4 The present invention is a block diagram of the structure of the online monitoring and diagnosis system for the blocking state of the vibrating screen. DETAILED DESCRIPTION

[0078] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present invention, and should not be construed as limiting the present invention.

[0079] In the description of the present invention, it should be understood that the terms "center", "longitudinal", "lateral", "length", "width", "thickness", "upper", "lower", "axial", "radial", "vertical", "horizontal", "inner", "outer" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first" and "second" are only used for descriptive purposes, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. In the description of the present invention, the meaning of "multiple" is two or more, unless otherwise clearly and specifically defined.

[0080] In the present invention, unless otherwise clearly specified and limited, the terms "installed", "connected", "connected", "fixed" and the like should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be an indirect connection through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0081] like Figure 1 As shown, the online monitoring and diagnosis method of the blocking state of the vibrating screen according to the present invention specifically comprises the following steps:

[0082] S01: The acceleration sensor is installed on the vibrating screen to measure the vibration acceleration; the torque sensor is installed on the eccentric shaft of the vibrating screen to measure the torque of the shaft. The signal acquisition module collects and stores the three acceleration signals and torque signals in real time, mainly including the following steps:

[0083] S1.1: Set the acquisition frequency, determine the communication method between the STM32 controller and the host computer, and use the installed acceleration sensor and torque sensor to collect real-time signals.

[0084] S1.2: The signal data is uploaded to the host computer through STM32 and stored in the database to facilitate subsequent signal processing.

[0085] S02: Figure 2 and Figure 3 As shown, the diagnostic module pre-processes and trains the collected torque information and acceleration information, and diagnoses through the torque diagnostic model and the vibration acceleration diagnostic model respectively. The operating status of the axial flow separation device is judged by comparing and cross-identifying the results of the two diagnostic models, which mainly includes the following steps:

[0086] S2.1: The four characteristic parameters extracted from the torque information by the torque diagnosis model are: first-order difference ΔD c , relative torque ratio R c , difference accumulation sum ∑A c , second-order difference Δ 2 D c , as shown below:

[0087]

[0088] Δ 2 D c =ΔD c+1 -ΔD c

[0089] Where: x c is the period of the c-th pulse period sequence; x c+1 is the period of the c+1th pulse period sequence; c is the pulse period sequence, c=1,2,...N, N is the total number of pulse period sequences;

[0090] ΔD c is the first-order difference value of the cth pulse period sequence. The first-order difference of the pulse period sequence describes the instantaneous change direction of the vibration screen torque.

[0091] R c is the relative torque ratio of the cth pulse cycle sequence;

[0092] x ref,c is the torque value measured in the cth pulse cycle sequence;

[0093] ∑A c is the first-order difference value ΔD of the c-th pulse cycle sequence c sum;

[0094] Δ 2 D c is the second-order difference value of the c-th pulse period sequence, Δ 2 D c The value represents the magnitude of the torque increase or decrease, and also reflects the direction of the torque change.

[0095] ΔD c+1is the first-order difference numerical value of the (c + 1)-th pulse period sequence.

[0096] S2.2: The fusion of the four characteristic parameters is as follows:

[0097]

[0098] Where:

[0099] I is the comprehensive information after fusing the four characteristic parameters;

[0100] a1 is the weight of the first-order difference ΔD c ; a2 is the weight of the relative speed ratio R c ; a3 is the weight of the difference cumulative sum ∑A c ; a4 is the weight of the second-order difference Δ 2 D c ;

[0101] ΔD ref is the maximum value of the first-order difference numerical value when the vibrating screen is in a blocked state;

[0102] R ref is the maximum value of the relative speed ratio when the vibrating screen is in a blocked state;

[0103] ∑A ref is the maximum value of the sum of each item of the first-order difference ΔD c when the vibrating screen is in a blocked state;

[0104] Δ 2 D ref is the maximum value of the second-order difference numerical value when the vibrating screen is in a blocked state;

[0105] S2.3: Set the maximum value I Max and the minimum value I Min of the comprehensive information I. When I < I Min , the vibrating screen is in a normal state; when I Min < I < I Max , the vibrating screen is in a state approaching blockage; when I > I Max , the vibrating screen is in a blocked state.

[0106] Example: The maximum value of the comprehensive information I is 1.2, and the minimum value of the comprehensive information I is 0.4. When I < 0.4, the axial flow threshing and separating device is in a normal state; when 0.4 < I < 1.2, the axial flow threshing and separating device is in a state approaching blockage; when I > 1.2, the axial flow threshing and separating device is in a blocked state.

[0107] S2.4: Establish a vibration acceleration diagnosis model, specifically:

[0108] S2.4.1: Three acceleration signals are obtained through experiments, each of which is a continuous signal and has a signal segment of a normal state, a signal segment of a congestion state, and a signal segment of a congestion state.

[0109] Experimental method:

[0110] Step 1: Turn on the motor, wait until it is stable, and then start collecting acceleration;

[0111] Step 2: The conveyor belt runs to transport materials. The rated material feed rate is 0.4kg / s at the beginning. At this time, the vibrating screen works normally.

[0112] Step 3: The conveyor belt starts to transport materials at a rate of 0.6, 0.7 or 0.8 kg / s, causing the vibrating screen to become abnormal and become blocked (0.6 and 0.7 kg / s) and blocked (0.7 and 0.8 kg / s).

[0113] Step 3: Get the acceleration signal from normal to congestion to congestion.

[0114] S2.4.2: Preprocess the three acceleration signals and perform feature extraction to extract 17 feature types in the time domain, frequency domain, and time-frequency domain, as shown in Table 1:

[0115] Table 1 17 types of features

[0116] Time Domain Frequency Domain Time-frequency domain Mean Center of gravity frequency Wavelet Energy Entropy Standard Deviation Mean square frequency Wavelet Singular Entropy Square root amplitude RMS frequency Total energy of wavelet subband RMS value Frequency variance Peak Frequency standard deviation Skewness Kurtosis Peak Index Margin Index

[0117] S2.4.3: Splice and fuse the time domain, frequency domain, and time-frequency domain of each acceleration signal to obtain the spliced ​​features; the spliced ​​feature of the first acceleration signal is O1, the spliced ​​feature of the second acceleration signal is O2, and the spliced ​​feature of the third acceleration signal is O3; forward normalize the spliced ​​features to obtain forward normalized matrices A1, A2, and A3, where A1 is the matrix of O1 after forward normalization, A2 is the matrix of O2 after forward normalization, and A3 is the matrix of O3 after forward normalization, as shown below:

[0118]

[0119] Where: m represents the number of signal sample segments, n represents the number of features, represents the jth feature of the i-th signal sample segment; a represents the sensor number, a=(1, 2, 3), when a=1, represents the jth feature of the i-th signal sample segment in the first acceleration signal; when a=2, represents the jth feature of the i-th signal sample segment in the second acceleration signal; when a=3, represents the jth feature of the i-th signal sample segment in the third acceleration signal; i = 1, 2, ..., m; j = 1, 2, ..., n;

[0120] S2.4.4: Calculate the proportion of the eigenvalue of the i-th signal sample segment under the j-th feature of each signal to the sum of the eigenvectors

[0121]

[0122] Calculate the proportionality factor k value,

[0123] Calculate the lower uncertainty measure E of the j-th eigenvector of each signal according to the proportionality coefficient k a j :

[0124]

[0125] Where k>0;

[0126] The weight of the jth feature for:

[0127]

[0128] S2.4.5: Fuse the feature matrices of the three acceleration signals into one feature fusion matrix

[0129]

[0130] in The jth feature of the fused i-th signal sample segment.

[0131] S2.4.6: Fusion Matrix Each signal sample segment is manually labeled with a sample label, denoted as y i (1≤i≤m), that is, the type of failure of vibrating screen blockage, as shown in the following table:

[0132] Fault type Tag value normal 1 Congestion 2 Blockage 3

[0133] The fusion matrix As the input of the first BP neural network, according to the above steps, the feature vector has n columns in total, so the input of the first BP neural network is n columns, and the number of samples m is divided into 70% and 30% as training set and test set, and the output is the label column, the label column is 1, the number of labels is m, and the hidden layer is set to 2*(n+1) / 3. The model is continuously trained with RMSE as the indicator until RMSE is less than 1e-04, and the BP neural network model training is completed.

[0134] S2.4.7: Increase and decrease the feature vector of the jth feature by 10% respectively, input it into the trained first BP neural network, and output the corresponding predicted label value and

[0135] By using the dimensionality reduction algorithm and the weighted mean influence value algorithm (WMIV), several features with large feature influence values ​​are screened out to achieve feature dimensionality reduction. The calculation formula is as follows:

[0136]

[0137] in, The predicted label is obtained by increasing the i-th sample under the j-th input variable by 10%; The predicted label is obtained after reducing the i-th sample by 10% under the j-th input variable; Represents the ratio of different predicted labels under the j-th input variable to the total number of labels of the feature column after all samples increase by 10%; Represents the ratio of different predicted labels under the j-th input variable to the total number of labels of the feature column after all samples are reduced by 10%;

[0138] Calculate the WMIV mean: WMIV j is the weighted average impact value of the jth input variable;

[0139] According to the judgment formula: WMIV j ≥Mean, z eigenvectors can be obtained;

[0140] The eigenvector matrix containing n eigenindices Reduce the dimension to a feature matrix containing only z feature indicators

[0141]

[0142] S2.4.8: Transform the feature matrix Divide into test set and training set, and use them as the input layer of the second BP neural network. Set the number of neurons to z, and the number of neurons in the hidden layer to 2*(z+1) / 3. Continuously train the model with RMSE as the indicator until RMSE is less than 1e-04, and then obtain the vibration acceleration blocking diagnosis model.

[0143] S2.5: Send the acceleration signals detected by the three acceleration sensors to the vibration acceleration diagnosis model to obtain a diagnosis result.

[0144] S2.6: The diagnostic results output by the vibration acceleration blocking diagnostic model are compared with the output results of the torque diagnostic model. If the results are the same, the diagnostic results are output; if the results are different, an alarm sound is prompted and the results are saved. Subsequently, the different results of multiple cross-validations are saved and compared with the actual working conditions of the actual vibrating screen, and the diagnostic results are manually summarized, that is, manual auxiliary diagnosis is performed.

[0145] like Figure 4 As shown, the online monitoring and diagnosis method system of the blocking state of the vibrating screen described in the present invention includes a collection module, a diagnosis module and an execution module.

[0146] The acquisition module includes a sensor and a signal acquisition unit, wherein the sensor includes an acceleration sensor and a torque sensor, and the signal acquisition unit includes an STM32 controller and a host computer. The acceleration sensor is installed on the vibrating screen to measure the vibration acceleration; the torque sensor is installed on the eccentric shaft of the vibrating screen to measure the torque of the shaft.

[0147] The diagnostic module includes a torque diagnostic model and a vibration acceleration diagnostic model. The host computer inputs the collected torque information into the torque diagnostic model, and the host computer inputs the collected acceleration into the vibration acceleration diagnostic model. The operating state of the axial flow separation device is determined by comparing and cross-identifying the diagnostic results of the two models.

[0148] The execution module includes a human-machine exchange interface and an audible and visual alarm device. The human-machine interface is used to realize the visualization of real-time signal data of each sensor, the display of blocking diagnosis results and the storage of data. The audible and visual alarm device is used to issue an early warning signal according to the operating status of the vibrating screen obtained by the diagnosis module. When in a blocking state, the audible and visual alarm device issues a fault warning.

[0149] Collect and store multiple acceleration signals and torque signals, pre-process the collected signals, extract characteristics of the torque signal, obtain thresholds, establish a torque diagnosis model, fuse and divide the processed three-way acceleration signals, train them at the same time, obtain a vibration acceleration diagnosis model, cross-validate the two, output the results and display them.

[0150] It should be understood that although this specification is described according to various embodiments, not every embodiment contains only one independent technical solution. This narrative method of the specification is only for the sake of clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each embodiment may also be appropriately combined to form other implementation methods that can be understood by those skilled in the art.

[0151] The series of detailed descriptions listed above are only specific descriptions of feasible embodiments of the present invention. They are not intended to limit the scope of protection of the present invention. All equivalent embodiments or changes that do not deviate from the technical spirit of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for online monitoring and diagnosis of the blocking state of a vibrating screen, characterized in that: The steps include: The acceleration sensor is used to measure the vibration acceleration information at three different positions on the vibrating screen; the torque sensor is used to measure the torque information of the eccentric shaft; The collected torque information is preprocessed and diagnosed through a torque diagnosis model to obtain a diagnosis result; Establish vibration acceleration diagnostic model; The vibration acceleration information collected at three different positions is input into the vibration acceleration diagnosis model to obtain the diagnosis result; The results of the two diagnostic models are compared and cross-identified to determine the operating status of the vibrating screen.

2. The method for online monitoring and diagnosis of the blocking state of a vibrating screen according to claim 1 is characterized in that: The collected torque information is preprocessed as follows: The four characteristic parameters extracted from the torque information are: first-order difference ΔD c , relative torque ratio R c , difference accumulation sum ∑A c , second-order difference Δ 2 D c , specifically: ΔD c =x c+1 -x c ∑A c =∑ΔD c D 2 D c =ΔD c+1 -ΔD c Where: x c is the period of the c-th pulse period sequence; x c+1 is the period of the c+1th pulse period sequence; c is the pulse period sequence, c=1,2,...N, N is the total number of pulse period sequences; ΔD c is the first-order difference value of the cth pulse period sequence. The first-order difference of the pulse period sequence describes the instantaneous change direction of the vibration screen torque; R c is the relative torque ratio of the cth pulse cycle sequence; x ref,c is the torque value measured in the cth pulse cycle sequence; ∑A c is the first-order difference value ΔD of the c-th pulse cycle sequence c sum; Δ 2 D c is the second-order difference value of the c-th pulse period sequence, Δ 2 D c The value represents the magnitude of the increase or decrease in the drum speed, and is used to reflect the direction of torque change; ΔD c+1 is the first-order difference value of the c+1th pulse period sequence; The four feature parameters are fused as follows: in: I is the comprehensive information after information fusion of the four characteristic parameters; a1 is the first-order difference ΔD c The weight of a2 is the relative speed ratio R c The weight of a3 is the difference cumulative sum ∑A c The weight of; a4 is the second-order difference Δ 2 D c The weight of ΔD ref It is the maximum value of the first-order difference value when the vibrating screen is blocked; R ref It is the maximum value of the relative speed ratio when the vibrating screen is blocked; ∑A ref is the first-order difference ΔD when the vibrating screen is blocked c The maximum value of the sum of the terms of ; Δ 2 D ref It is the maximum value of the second-order difference when the vibrating screen is blocked.

3. The method for online monitoring and diagnosis of the blocking state of a vibrating screen according to claim 2 is characterized in that: Through the torque diagnosis model diagnosis, the diagnosis results are obtained, specifically: Set the maximum value of comprehensive information I Max and minimum value I Min ; when I Min When I Min <I<I Max When I>I Max When , the vibrating screen is in a blocked state.​ 4. The method for online monitoring and diagnosis of the blocking state of a vibrating screen according to claim 1, characterized in that: Establish a vibration acceleration diagnostic model, specifically: Three acceleration signals are obtained through experiments, each of which is a continuous signal and has a signal segment of a normal state, a signal segment of a jam state, and a signal segment of a jam state; Fusion of the feature matrices of the three acceleration signals into a feature fusion matrix The fusion matrix Each signal sample segment is manually labeled, and the labeled fusion matrix is ​​used as the input of the first BP neural network, and the output is the label, and the first BP neural network model is obtained through training; The fusion matrix After the feature vector of the jth feature increases and decreases by 10% respectively, it is input into the first BP neural network model to obtain the predicted increase label value and predict reduced label values Using predictions to increase label value and predict reduced label values By using the dimensionality reduction algorithm and the weighted average influence value algorithm, the eigenvector matrix containing n characteristic indicators is Reduce the dimension to a feature matrix containing only z feature indicators z <n; The feature matrix It is divided into a test set and a training set, and used as the input layer of the second BP neural network. The number of neurons is set to z, and the number of neurons in the hidden layer is 2*(z+1) / 3. The model is trained with RMSE as the indicator until RMSE is less than 1e-04, and the vibration acceleration blocking diagnosis model is obtained.

5. The method for online monitoring and diagnosis of the blocking state of a vibrating screen according to claim 4, characterized in that: Fusion of the feature matrices of the three acceleration signals into a feature fusion matrix Specifically: The three acceleration signals are preprocessed and feature extracted to extract time domain features, frequency domain features and time-frequency domain features; The time domain, frequency domain, and time-frequency domain of each acceleration signal are spliced ​​and fused to obtain the spliced ​​features; the spliced ​​feature of the first acceleration signal is O1, the spliced ​​feature of the second acceleration signal is O2, and the spliced ​​feature of the third acceleration signal is O3; the spliced ​​features are forward normalized to obtain forward normalized matrices A1, A2, and A3, where A1 is the matrix of O1 after forward normalization, A2 is the matrix of O2 after forward normalization, and A3 is the matrix of O3 after forward normalization, as shown below: Where: m represents the number of signal sample segments, n represents the number of features, represents the jth feature of the i-th signal sample segment; a represents the sensor number, a=(1, 2, 3), when a=1, represents the jth feature of the i-th signal sample segment in the first acceleration signal; when a=2, represents the jth feature of the i-th signal sample segment in the second acceleration signal; when a=3, represents the jth feature of the i-th signal sample segment in the third acceleration signal; i = 1, 2, ..., m; j = 1, 2, ..., n; Calculate the proportion of the eigenvalue of the i-th signal sample fragment under the j-th feature of each signal to the sum of the eigenvectors Calculate the proportionality factor k value, Calculate the lower uncertainty measure E of the j-th eigenvector of each signal according to the proportionality coefficient k a j : Where k>0; The weight of the jth feature for: Fusion of the feature matrices of the three acceleration signals into a feature fusion matrix in The jth feature of the fused i-th signal sample segment, is the weight of the jth feature of the first acceleration signal; is the weight of the jth feature of the second acceleration signal; is the weight of the jth feature of the third acceleration signal.

6. The method for online monitoring and diagnosis of the blocking state of a vibrating screen according to claim 5, characterized in that: By using the dimensionality reduction algorithm and the weighted average influence value algorithm, the eigenvector matrix containing n characteristic indicators is Reduce the dimension to a feature matrix containing only z feature indicators Specifically: The weighted average impact value WMIV calculation formula is as follows: in, The predicted label is obtained by increasing the i-th sample under the j-th input variable by 10%; The predicted label is obtained after reducing the i-th sample by 10% under the j-th input variable; Represents the ratio of different predicted labels under the j-th input variable to the total number of labels of the feature column after all samples increase by 10%; Represents the ratio of different predicted labels under the j-th input variable to the total number of labels of the feature column after all samples are reduced by 10%; Calculating WMIV j average value WMIV j is the weighted average impact value of the jth input variable; According to the judgment formula: WMIV j ≥Man, we get z eigenvectors, that is, we get z eigenvectors from n eigenvectors. The eigenvector matrix containing n eigenindices Reduce the dimension to a feature matrix containing only z feature indicators 7. The method for online monitoring and diagnosis of the blocking state of a vibrating screen according to claim 1, characterized in that: The results of the two diagnostic models are compared and cross-identified to determine the operating status of the axial flow separation device, specifically: If the results of the two diagnostic models are the same, the diagnostic result is output; if the results of the two diagnostic models are different, a prompt alarm sound is issued and manual auxiliary diagnosis is performed.

8. A system for online monitoring and diagnosis of the blocking state of a vibrating screen according to claim 1, characterized in that: It includes acquisition module, diagnosis module and execution module; The acquisition module includes a sensor and a signal acquisition unit. The sensor includes an acceleration sensor and a torque sensor. The acceleration sensor is installed on the vibrating screen to measure the vibration acceleration; the torque sensor is installed on the eccentric shaft of the vibrating screen to measure the torque of the shaft. The diagnostic module includes a rotation speed diagnostic model and a vibration acceleration diagnostic model. The acquisition module inputs the acquired rotation speed information into the rotation speed diagnostic model, and the acquisition module inputs the acquired acceleration into the vibration acceleration diagnostic model. The operating state of the axial flow separation device is determined by comparing and cross-identifying the diagnostic results of the two models. The execution module includes a human-machine interface and an acoustic and visual alarm device. The human-machine interface is used for visualizing real-time sensor signal data and storing and displaying diagnostic results. The acoustic and visual alarm device is used for issuing fault warnings.

Images