A method for diagnosing ventilation network faults

Abstract

A ventilation network fault diagnosis method proposed in the present application includes: collecting a fault data set of the ventilation network; establishing fault diagnosis models according to multiple machine learning algorithms; designing fault diagnosis learning algorithms for multiple fault diagnosis models according to the fault data set; performing fault diagnosis on multiple fault diagnosis models according to the fault data set and obtaining the fault diagnosis accuracy; selecting the optimal fault diagnosis model using the fault diagnosis accuracy as a selection index; and performing ventilation network fault diagnosis using the optimal fault diagnosis model. The advantages of the present application over the prior art are: selecting the optimal fault diagnosis model through the fault data set, thereby improving the accuracy of fault diagnosis, thereby minimizing casualties and property losses and ensuring the safety and reliability of the mine ventilation system.

Description

Technical Field

[0001] The present application relates to the technical field of tunnel fault diagnosis, and in particular to a ventilation network fault diagnosis method. Background Art

[0002] Changes in the equivalent wind resistance between tunnels caused by deformation, extension and scrapping, and damper switching in underground coal mine tunnels can cause abnormal changes in the air volume of the mine ventilation system. The current ventilation network fault diagnosis method cannot quickly and accurately identify the faulty tunnel, and cannot guarantee the safety and reliability of the mine ventilation system. Therefore, it is necessary to improve it. Summary of the invention

[0003] The present application aims to solve one of the technical problems in the related art at least to some extent.

[0004] To this end, the purpose of this application is to propose a ventilation network fault diagnosis method.

[0005] To achieve the above-mentioned purpose, the present application proposes a ventilation network fault diagnosis method, including: collecting a fault data set of the ventilation network; establishing fault diagnosis models according to multiple machine learning algorithms; multiple fault diagnosis models respectively design fault diagnosis learning algorithms according to the fault data set; multiple fault diagnosis models respectively perform fault diagnosis according to the fault data set and obtain the fault diagnosis accuracy; use the fault diagnosis accuracy as a selection index to select the optimal fault diagnosis model; and apply the optimal fault diagnosis model to perform ventilation network fault diagnosis.

[0006] The acquisition of the fault data set of the ventilation network includes: calibrating the actual fault lane number in the ventilation network; acquiring the air volume change data set corresponding to the actual fault lane number; and integrating the actual fault lane number and the air volume change data set to obtain the fault data set.

[0007] The collecting of the air volume change data set corresponding to the actual fault lane includes: obtaining the original air volume of the lane directly or indirectly connected to the actual fault lane; obtaining the current air volume of the lane directly or indirectly connected to the actual fault lane; calculating the difference between the current air volume and the original air volume, and calculating the ratio of the difference to the original air volume; if the ratio is not less than 10%, collecting the current air volume as the air volume change data corresponding to the actual fault lane; and integrating multiple air volume change data to obtain the air volume change data set.

[0008] The multiple fault diagnosis models are respectively designed with fault diagnosis learning algorithms according to the fault data set, including: dividing the fault data set into training sets; using the air volume change data set of the training set as the input feature of the fault diagnosis model; using the actual fault lane number of the training set as the output feature of the fault diagnosis model; and designing fault diagnosis learning algorithms respectively for the multiple fault diagnosis models.

[0009] Before the multiple fault diagnosis models perform fault diagnosis according to the fault data set respectively; multiple parameters of the multiple fault diagnosis models are selected respectively; multiple value ranges of the parameters are selected respectively; the fault data set is divided into a verification set; the air volume change data set of the verification set is used as the input feature of the fault diagnosis model; the actual fault lane number of the verification set is used as the output feature of the fault diagnosis model; the multiple fault diagnosis models are traversed using the grid search method according to the designed fault diagnosis learning algorithm and multiple parameters, and the cross-validation average score is used as the selection index to obtain the optimal parameters of the fault diagnosis model.

[0010] The multiple fault diagnosis models respectively perform fault diagnosis according to the fault data set, and obtain the fault diagnosis accuracy rate, including: dividing the fault data set into a test set; using the air volume change data set of the test set as the input feature of the fault diagnosis model; the multiple fault diagnosis models respectively perform fault diagnosis according to the designed fault diagnosis learning algorithm and optimal parameters, and obtain multiple diagnosis fault lane numbers; according to the diagnosis fault lane number and the actual fault lane number of the test set, obtain the fault diagnosis accuracy rate of the multiple fault diagnosis models.

[0011] The fault diagnosis models are established according to a plurality of machine learning algorithms respectively, including: establishing a support vector machine diagnosis model according to a support vector machine; establishing a random forest diagnosis model according to a random forest; and establishing a neural network diagnosis model according to a neural network.

[0012] The parameters of the support vector machine diagnostic model include: Gaussian kernel width and regularization parameter; the parameters of the random forest diagnostic model include: the number of decision trees and the randomness size of each tree; the parameters of the neural network diagnostic model include: the number of hidden layer nodes and regularization parameter.

[0013] The sum of the training set, the validation set and the test set is the fault data set, and the proportions of the training set, the validation set and the test set in the fault data set are 56.25%, 18.75% and 25% respectively.

[0014] The application of the optimal fault diagnosis model to perform ventilation network fault diagnosis includes: collecting the air volume change data set corresponding to the actual lane; using the air volume change data set corresponding to the actual lane as the input feature of the optimal fault diagnosis model; the optimal fault diagnosis model performs fault diagnosis according to the fault diagnosis learning algorithm and optimal parameters, and outputs the diagnosed fault lane number.

[0015] After adopting the above technical scheme, the advantages of the present application compared with the prior art are: through the fault diagnosis model, the faulty tunnel can be identified and judged in time, thereby improving the efficiency of ventilation network fault diagnosis, and the optimal fault diagnosis model is selected through the fault data set, thereby improving the accuracy of fault diagnosis, thereby minimizing casualties and property losses and ensuring the safety and reliability of the mine ventilation system.

[0016] Additional aspects and advantages of the present application will be given in part in the description below, and in part will become apparent from the description below, or will be learned through the practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The above and / or additional aspects and advantages of the present application will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:

[0018] Figure 1 It is a flow chart of a ventilation network fault diagnosis method proposed in one embodiment of the present application;

[0019] Figure 2 It is a schematic diagram of the ventilation network structure used in the ventilation network fault diagnosis method experiment proposed in one embodiment of the present application;

[0020] Figure 3 It is a line chart used in an experiment of a ventilation network fault diagnosis method proposed in an embodiment of the present application. DETAILED DESCRIPTION

[0021] Embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements with the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present application, and cannot be construed as limiting the present application. On the contrary, the embodiments of the present application include all changes, modifications and equivalents that fall within the spirit and connotation of the appended claims.

[0022] like Figure 1 As shown, the embodiment of the present application proposes a ventilation network fault diagnosis method, comprising:

[0023] S1: Collect the fault dataset of the ventilation network;

[0024] S2: Establish fault diagnosis models based on multiple machine learning algorithms;

[0025] S3: Multiple fault diagnosis models are designed with fault diagnosis learning algorithms based on fault data sets;

[0026] S4: Multiple fault diagnosis models perform fault diagnosis based on the fault data set respectively, and obtain the fault diagnosis accuracy;

[0027] S5: Using the fault diagnosis accuracy as the selection index, select the optimal fault diagnosis model;

[0028] S6: Apply the optimal fault diagnosis model to diagnose ventilation network faults.

[0029] It can be understood that through the fault diagnosis model, the faulty tunnel can be identified and judged in time, which improves the efficiency of ventilation network fault diagnosis, and the accuracy of fault diagnosis is improved by selecting the optimal fault diagnosis model through the fault data set, thereby minimizing casualties and property losses and ensuring the safety and reliability of the mine ventilation system.

[0030] When a lane fails and becomes a faulty lane, the air volume of lanes directly or indirectly connected to it is bound to be affected. Therefore, in some embodiments, the fault data set of the ventilation network collected in step S1 includes:

[0031] S11: calibrate the actual fault lane number in the ventilation network;

[0032] S12: Collecting the air volume change data set corresponding to the actual fault lane;

[0033] S13: synthesizing the actual fault lane number and air volume change data set to obtain a fault data set;

[0034] It can be understood that by collecting the actual fault lane number and the corresponding air volume change data set, the fault data set is made more comprehensive, thereby making the fault diagnosis learning algorithm designed by the fault diagnosis model more accurate, effectively improving the accuracy of fault diagnosis, and the lanes are numbered so that operators can quickly identify the faulty lanes.

[0035] In some embodiments, collecting the fault data set of the ventilation network in step S1 further includes:

[0036] S14: Standardize the air volume change data set of the fault data set.

[0037] It can be understood that the fault data set is standardized to ensure that the average value of the data features of the air volume change data set is 0 and the variance is 1, thereby reducing the error in the fault diagnosis model design fault diagnosis learning algorithm and further improving the accuracy of fault diagnosis.

[0038] Since a large number of lanes are affected by the faulty lanes, in some embodiments, step S12: collecting the wind volume change data set corresponding to the actual faulty lanes includes:

[0039] S121: Obtaining the original air volume of the lanes directly or indirectly connected to the actual fault lane;

[0040] S122: Obtaining the current air volume of the lanes directly or indirectly connected to the actual fault lane;

[0041] S123: Calculate the difference between the current air volume and the original air volume, and calculate the ratio of the difference to the original air volume;

[0042] S124: If the ratio is not less than 10%, the current air volume is collected as the air volume change data corresponding to the actual fault lane;

[0043] S125: Integrate a plurality of wind volume change data to obtain a wind volume change data set.

[0044] It can be understood that by screening at a 10% ratio, the lanes that are less affected by the actual fault lanes are removed, thereby simplifying the fault data set, which not only improves the efficiency of the fault diagnosis model design fault diagnosis learning algorithm, but also reduces the error in the fault diagnosis model design fault diagnosis learning algorithm, thereby improving the accuracy of fault diagnosis.

[0045] It should be noted that the actual faulty tunnel is the tunnel that has experienced failures such as falling deformation, extension and scrapping in the past. The original air volume is the air volume recorded before the failure, and the current air volume is the air volume recorded after the failure.

[0046] In some embodiments, step S2: establishing fault diagnosis models according to multiple machine learning algorithms respectively includes:

[0047] S21: establishing a support vector machine diagnosis model based on the support vector machine;

[0048] S22: Establish a random forest diagnostic model based on random forest;

[0049] S23: Establish a neural network diagnosis model based on the neural network.

[0050] It can be understood that by establishing the support vector machine diagnosis model, the random forest diagnosis model and the neural network diagnosis model, the basis for selecting the optimal fault diagnosis model is made broader, thereby improving the quality of selecting the optimal fault diagnosis model and further improving the accuracy of fault diagnosis.

[0051] In some embodiments, step S3: designing a fault diagnosis learning algorithm based on the fault data set for each of the multiple fault diagnosis models includes:

[0052] S31: Divide the fault data set into a training set;

[0053] S32: using the air volume change data set of the training set as input features of the fault diagnosis model;

[0054] S33: using the actual fault lane number of the training set as the output feature of the fault diagnosis model;

[0055] S34: Design fault diagnosis learning algorithms for multiple fault diagnosis models respectively.

[0056] It can be understood that multiple fault diagnosis models are respectively designed with fault diagnosis learning algorithms according to training sets, which ensures the accuracy of the fault diagnosis learning algorithms designed by the fault diagnosis models.

[0057] In some embodiments, before multiple fault diagnosis models perform fault diagnosis based on the fault data sets respectively,

[0058] S35: Selecting multiple parameters of multiple fault diagnosis models respectively;

[0059] S36: Selecting value ranges of multiple parameters respectively;

[0060] S37: Divide the fault data set into a validation set;

[0061] S38: using the wind volume change data set of the validation set as the input feature of the fault diagnosis model;

[0062] S39: using the actual fault lane number of the validation set as the output feature of the fault diagnosis model;

[0063] S310: Multiple fault diagnosis models are traversed using a grid search method according to their designed fault diagnosis learning algorithms and multiple parameters, and the cross-validation average score is used as a selection index to obtain the optimal parameters of the fault diagnosis model.

[0064] It can be understood that the fault diagnosis model selects the optimal parameters through the verification set, which ensures that multiple fault diagnosis models have high fault diagnosis accuracy before the optimal fault diagnosis model is selected, thereby improving the accuracy of fault diagnosis.

[0065] It should be noted that the parameter with the highest cross-validation average score is the optimal parameter.

[0066] In some embodiments, step S4: multiple fault diagnosis models perform fault diagnosis according to the fault data set respectively, and obtain the fault diagnosis accuracy rate, including:

[0067] S41: Divide the fault data set into a test set;

[0068] S42: using the wind volume change data set of the test set as the input feature of the fault diagnosis model;

[0069] S43: multiple fault diagnosis models perform fault diagnosis according to their designed fault diagnosis learning algorithms and optimal parameters, and obtain multiple diagnosis fault lane numbers;

[0070] S44: deriving the fault diagnosis accuracy of multiple fault diagnosis models according to the diagnosed fault lane number and the actual fault lane number of the test set.

[0071] It can be understood that by comparing the diagnosed fault lane number with the actual fault lane number of the test set, the fault diagnosis accuracy of the fault diagnosis model is made more accurate, ensuring the selection of the optimal fault diagnosis model.

[0072] In some embodiments, the parameters of the support vector machine diagnostic model include: Gaussian kernel width a and regularization parameter b, the parameters of the random forest diagnostic model include: the number of decision trees c and the randomness size d of each tree, and the parameters of the neural network diagnostic model include: the number of hidden layer nodes e and regularization parameter f.

[0073] It can be understood that the fault diagnosis accuracy of the support vector machine diagnosis model is guaranteed by the Gaussian kernel width and the regularization parameter, the fault diagnosis accuracy of the random forest diagnosis model is guaranteed by the number of decision trees and the randomness size of each tree, and the fault diagnosis accuracy of the neural network diagnosis model is guaranteed by the number of hidden layer nodes and the regularization parameter.

[0074] In some embodiments, the sum of the training set, the validation set and the test set is the fault data set, and the proportions of the training set, the validation set and the test set in the fault data set are 56.25%, 18.75% and 25% respectively.

[0075] It can be understood that the accuracy of the fault diagnosis learning algorithm in the fault diagnosis model design is guaranteed through the relatively large proportion of training sets, and the accurate selection of the optimal parameters and the selection of the optimal fault diagnosis model are guaranteed through the relatively small proportion of validation sets and test sets that still have sufficient data sets.

[0076] In some embodiments, step S6: applying the optimal fault diagnosis model to perform ventilation network fault diagnosis includes:

[0077] S61: Collecting the wind volume change data set corresponding to the actual laneway;

[0078] S62: using the air volume change data set corresponding to the actual laneway as the input feature of the optimal fault diagnosis model;

[0079] S63: The optimal fault diagnosis model performs fault diagnosis according to the fault diagnosis learning algorithm and the optimal parameters, and outputs the diagnosed fault lane number.

[0080] It can be understood that the ventilation fault diagnosis of the actual tunnel is carried out through the optimal fault diagnosis model, which ensures the accuracy of fault diagnosis, minimizes casualties and property losses, and ensures the safety and reliability of the mine ventilation system.

[0081] Based on the above ventilation network fault diagnosis method, the specific experiment is as follows:

[0082] like Figure 2 As shown, 300 sets of air volume change data of the faulty lanes were obtained through the ventilation network platform, and five lanes, e10, e55, e15, e8, and e16, were selected as output variables. Some of the air volume change data are shown in Table 1:

[0083] Table 1

[0084]

[0085] Among them, there are 5 groups of actual fault lanes in Table 1, and the calibrated actual fault lane numbers are e10, e55, e15, e8 and e16 respectively. Each group of actual fault lanes selects the current air volume of 18 lanes directly or indirectly connected, among which 2-1, 2-2, etc. represent the lanes directly or indirectly connected to the actual fault lanes.

[0086] It should be noted that the serial numbers 1-5, 6-10, 11-15... are based on different faults.

[0087] The above air volume change data are standardized, and the standardized partial air volume change data set is shown in Table 2:

[0088] Table 2

[0089]

[0090] Establish support vector machine diagnostic model, random forest diagnostic model and neural network diagnostic model.

[0091] The parameter a in the support vector machine diagnosis model takes the value of {10, 100, 1000, 10000, 100000, 1000000}, and the parameter b takes the value of {0.00001, 0.0001, 0.001, 0.01, 0.1, 1}. Therefore, there are 36 parameter combinations in the support vector machine diagnosis model.

[0092] The parameter c in the random forest diagnosis model takes the value of {5, 10, 15, 20, 25, 30, 35}, and the parameter d takes the value of {1, 2, 3, 4, 5}. Therefore, there are 35 parameter combinations in the random forest diagnosis model.

[0093] In the neural network diagnosis model, the value of e is {6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18}, and the value of f is {0.000001, 0.00001, 0.0001, 0.001, 0.01, 0, 1}. Therefore, there are 78 parameter combinations in the neural network diagnosis model.

[0094] Using the grid search method to traverse, the optimal parameters of the support vector machine diagnosis model are a = 10000, b = 0.1, the optimal parameters of the random forest diagnosis model are c = 15, d = 4, and the optimal parameters of the neural network diagnosis model are e = 14, f = 0.00001. The accuracy comparison of the three fault diagnosis models under the optimal parameters is shown in Table 3:

[0095] Table 3

[0096]

[0097] It can be seen that the neural network diagnosis model has the highest accuracy of 0.946 on the test set and has the best generalization ability;

[0098] like Figure 3 As shown, the vertical axis is the accuracy of the neural network diagnosis model, and the horizontal axis is the actual fault tunnel numbers e10, e55, e15, e8 and e16. It can be seen that the fault recognition accuracy of the neural network diagnosis model in each tunnel is above 0.9, and the accuracy in tunnels e10 and e8 even reaches 1, which verifies that the ventilation network fault diagnosis method can improve the accuracy of fault diagnosis, minimize casualties and property losses, and ensure the safety and reliability of the mine ventilation system.

[0099] It should be noted that, in the description of this application, the terms "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or implying relative importance. In addition, in the description of this application, unless otherwise specified, the meaning of "plurality" is two or more.

[0100] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, segment or portion of code that includes one or more executable instructions for implementing the steps of a specific logical function or process, and the scope of the preferred embodiments of the present application includes alternative implementations in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in the reverse order depending on the functions involved, which should be understood by technicians in the technical field to which the embodiments of the present application belong.

[0101] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0102] Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limitations on the present application. Ordinary technicians in this field can change, modify, replace and modify the above embodiments within the scope of the present application.

Claims

1. A ventilation network fault diagnosis method, characterized in that: include: collecting a fault data set of the ventilation network; Establish fault diagnosis models based on various machine learning algorithms; The plurality of fault diagnosis models respectively design fault diagnosis learning algorithms according to the fault data sets; The plurality of fault diagnosis models respectively perform fault diagnosis according to the fault data set and obtain a fault diagnosis accuracy rate; Using the fault diagnosis accuracy as a selection index, an optimal fault diagnosis model is selected; Applying the optimal fault diagnosis model to perform ventilation network fault diagnosis; Wherein, the collecting of the fault data set of the ventilation network includes: calibrating the actual fault lane number in the ventilation network; collecting the air volume change data set corresponding to the actual fault lane number; combining the actual fault lane number and the air volume change data set to obtain the fault data set; The collecting of the air volume change data set corresponding to the actual fault lane includes: obtaining the original air volume of the lane directly or indirectly connected to the actual fault lane; obtaining the current air volume of the lane directly or indirectly connected to the actual fault lane; calculating the difference between the current air volume and the original air volume, and calculating the ratio of the difference to the original air volume; if the ratio is not less than 10%, collecting the current air volume as the air volume change data corresponding to the actual fault lane; combining multiple air volume change data to obtain the air volume change data set; The multiple fault diagnosis models respectively design fault diagnosis learning algorithms according to the fault data set, including: dividing the fault data set into training sets; using the air volume change data set of the training set as the input feature of the fault diagnosis model; using the actual fault lane number of the training set as the output feature of the fault diagnosis model; and designing fault diagnosis learning algorithms respectively for the multiple fault diagnosis models; Before the multiple fault diagnosis models perform fault diagnosis according to the fault data set respectively; multiple parameters of the multiple fault diagnosis models are selected respectively; multiple value ranges of the parameters are selected respectively; the fault data set is divided into a validation set; the air volume change data set of the validation set is used as the input feature of the fault diagnosis model; the actual fault lane number of the validation set is used as the output feature of the fault diagnosis model; the multiple fault diagnosis models are traversed using the grid search method according to the designed fault diagnosis learning algorithm and multiple parameters, and the cross-validation average score is used as the selection index to obtain the optimal parameters of the fault diagnosis model; The multiple fault diagnosis models respectively perform fault diagnosis according to the fault data set, and obtain the fault diagnosis accuracy rate, including: dividing the fault data set into a test set; using the air volume change data set of the test set as the input feature of the fault diagnosis model; the multiple fault diagnosis models respectively perform fault diagnosis according to the designed fault diagnosis learning algorithm and optimal parameters, and obtain multiple diagnosis fault lane numbers; according to the diagnosis fault lane number and the actual fault lane number of the test set, obtain the fault diagnosis accuracy rate of the multiple fault diagnosis models; The fault diagnosis models established according to the multiple machine learning algorithms include: establishing a support vector machine diagnosis model according to a support vector machine; establishing a random forest diagnosis model according to a random forest; establishing a neural network diagnosis model according to a neural network; The parameters of the support vector machine diagnostic model include: Gaussian kernel width and regularization parameter; the parameters of the random forest diagnostic model include: the number of decision trees and the randomness size of each tree; the parameters of the neural network diagnostic model include: the number of hidden layer nodes and regularization parameter.

2. The ventilation network fault diagnosis method according to claim 1, characterized in that: The sum of the training set, the validation set and the test set is the fault data set, and the proportions of the training set, the validation set and the test set in the fault data set are 56.25%, 18.75% and 25% respectively.

3. The ventilation network fault diagnosis method according to claim 1, characterized in that: The application of the optimal fault diagnosis model to diagnose ventilation network faults includes: Collect the wind volume change data set corresponding to the actual laneway; Using the air volume change data set corresponding to the actual lane as the input feature of the optimal fault diagnosis model; The optimal fault diagnosis model performs fault diagnosis according to the fault diagnosis learning algorithm and optimal parameters, and outputs the diagnosis fault lane number.

Images