In order to accurately identify the dynamic health of shearer, reducing operating trouble and production accident of shearer and improving coal production efficiency further, a dynamic health assessment approach for shearer based on artificial immune algorithm was proposed. The key technologies such as system framework, selecting the indicators for shearer dynamic health assessment, and health assessment model were provided, and the flowchart of the proposed approach was designed. A simulation example, with an accuracy of 96%, based on the collected data from industrial production scene was provided. Furthermore, the comparison demonstrated that the proposed method exhibited higher classification accuracy than the classifiers based on back propagation-neural network (BP-NN) and support vector machine (SVM) methods. Finally, the proposed approach was applied in an engineering problem of shearer dynamic health assessment. The industrial application results showed that the paper research achievements could be used combining with shearer automation control system in fully mechanized coal face. The simulation and the application results indicated that the proposed method was feasible and outperforming others.
Due to the randomicity and complexity of underground geological conditions, assessment of shearer health condition would present the characteristics of complexity, fuzziness, and uncertainty, and this may affect the coal production or even endanger the operator’s life. Moreover, because of the poor mining environment and complex component structure of shearer, the shearer operator cannot accurately estimate the working status of shearer, which may lead to some problems of poor coal quality and low mining efficiency. Furthermore, an increasing number of safety accidents in collieries are caused frequently. Therefore, it is necessary to assess the dynamic health condition of shearer which has become a challenging and significant research subject [
Depending on the assessment of the health condition of shearer, this can reduce operating trouble and production accident of shearer and improve production efficiency further. In recent years, many researches have brought out some achievement on shearer health condition diagnosis. The multiple fault classifier based on the improved support vector machine theory is used to judge the fault types of coal shearer [
Dynamic health assessment was used in spacecraft primarily in the 1970s. At present, domestic and abroad researchers have worked on the modeling approaches for dynamic health assessment and proposed several solutions. The density-based spatial clustering of applications with noise has been used for bearings’ condition monitoring [
The first mathematical model in artificial immune system was proposed in 1974, which initiated subsequent researches and discussions. Artificial immune system (AIS), as a novel intelligent algorithm method, inspired from the biological immune system, is an effective means for prediction [
The remainder of this paper is organized as follows. Some related works are outlined in Section
Recent publications relevant to this paper are mainly concerned with two research streams: the dynamic health assessment methods and artificial immune algorithm. In this section, we try to summarize the relevant literatures.
For the dynamic health assessment problem, lots of research has been done since the last decades. In [
The artificial immune algorithm was firstly proposed by Farmer in 1986 [
According to the above researches, many health assessment methods, such as density-based spatial clustering and dynamic Bayesian networks, have been applied in the bearings’ condition monitoring, network device dynamic health monitoring, and so on. But there are still no relevant studies on the dynamic health assessment methods for shearer. Considering the superiority and universality of artificial immune algorithm, this paper prepares to use this AI algorithm to predict the dynamic health status of shearer. A simulation experiment and an application example are carried out and the proposed approach is proved to be feasible and efficient.
Some real-time running indicators of shearer are usually used to classify the health condition of shearer since the signals can describe its dynamic characteristics. In order to identify the dynamic health status of shearer, the following three processes are required. These processes are assessment indicators selecting, data acquisition and initialization, and multiclass classifiers training and testing. The proposed condition classification approach for shearer dynamic health state is shown in Figure
The framework of the proposed approach.
The system of shearer is made up by many subsystems. Establishing a scientific and reasonable evaluation system is the foundation of the health state evaluation for shearer. Depending on the actual operation situation of shearer and referencing other health assessment systems, the assessment consequences for shearer health can be divided into four typical modes: normal mode, transition mode, abnormal mode, and danger mode. The definition of each type of operation is given as follows.
By setting malfunction threshold value depending on operation situation, four modes of shearer health situation decrease progressively. Four different health modes can guide coal worker adopting corresponding operation, respectively.
The system of shearer is made up by many subsystems. The data from historical recording and real-time monitoring of the subsystems reflect the health status of shearer more or less. However, in practical application, we must choose the most effective indicators to assess the health situation of shearer and eliminate subordinate indicators, as excessive assessment indicators will reduce the impact of the main indicators, causing an incorrect result. According to the expert experience and actual working condition of shearer, the dynamic health condition depends on the real-time monitoring data. In this paper, the key content is the real-time health assessment of shearer. Thus, to assess the dynamic health situation of shearer, we choose nine real-time running indicators: the pulling speed
The indicators of dynamic health assessment model for shearer.
According to the nine indicators of shearer dynamic health assessment approach defined above, we can determine the unknown schema object
Immune classifier model of dynamic health assessment for shearer.
Before establishing the dynamic health assessment model based on artificial immune algorithm, the related definitions of multiclass classifiers are given as follows.
Each schema object can be represented as a
Detectors set
Receptor can recognize any one of the schema objects of a certain type. The degree of similarity between receptor and schema object can measure affinity:
The function value of affinity lies between 0 and 1. The more similar the value between schema object
In this section, the flows for establishing the dynamic health assessment model based on artificial immune algorithm are provided in detail. It mainly includes three steps.
This step can be regarded as a data preprocessing stage. Each schema object is represented as a 9-dimensional vector
The purpose of training stage is generating an effective detector for each schema object. The steps for generating detector are given as follows.
Take preprocessed
Generate random alternative detectors set
Calculate affinity between
Calculate affinity between
Delete the individual from
If
Using the same negative selection algorithm of generating detector, repeat calculation four times from class object
The generation process of a detector.
To distinguish self-class object and non-self-class object, all detectors are used for circulatory elimination for new sample in testing step. The flowchart of negative selection test for a new sample is shown in Figure
Flowchart of negative selection test for a new sample.
If schema object
If schema object
If schema object
In this section, some simulation examples were put forward to verify the feasibility and efficiency of the proposed approach.
The sample data were acquired from the shearer in 22210 fully mechanized coal face of Zhong Ping Energy Chemical Group No. 6 Mine. Depending on the assessment model of prediction approach of shearer dynamic health assessment, the acquired data were normalized so that the object data were represented as a 9-dimensional vector,
Normalized data of pattern objects for shearer.
Number |
|
|
|
|
|
|
|
|
|
Categories |
---|---|---|---|---|---|---|---|---|---|---|
1 | 0.226 | 0.515 | 0.485 | 0.559 | 0.539 | 0.455 | 0.568 | 0.298 | 0.338 |
|
2 | 0.194 | 0.540 | 0.485 | 0.557 | 0.535 | 0.455 | 0.565 | 0.287 | 0.328 |
|
3 | 0.168 | 0.577 | 0.494 | 0.486 | 0.480 | 0.458 | 0.565 | 0.291 | 0.335 |
|
4 | 0.167 | 0.562 | 0.485 | 0.513 | 0.496 | 0.458 | 0.565 | 0.290 | 0.338 |
|
⋮ | ⋮ | ⋮ | ⋮ | ⋮ | ⋮ | ⋮ | ⋮ | ⋮ | ⋮ | ⋮ |
185 | 0.257 | 0.509 | 0.494 | 0.461 | 0.489 | 0.458 | 0.565 | 0.296 | 0.343 |
|
186 | 0.258 | 0.519 | 0.506 | 0.471 | 0.497 | 0.455 | 0.565 | 0.295 | 0.343 |
|
187 | 0.000 | 0.485 | 0.497 | 0.444 | 0.432 | 0.458 | 0.570 | 0.309 | 0.329 |
|
188 | 0.227 | 0.503 | 0.488 | 0.417 | 0.422 | 0.458 | 0.568 | 0.281 | 0.327 |
|
⋮ | ⋮ | ⋮ | ⋮ | ⋮ | ⋮ | ⋮ | ⋮ | ⋮ | ⋮ | ⋮ |
354 | 0.350 | 0.519 | 0.491 | 0.459 | 0.450 | 0.458 | 0.565 | 0.284 | 0.329 |
|
355 | 0.858 | 0.552 | 0.821 | 0.531 | 0.704 | 0.458 | 0.570 | 0.301 | 0.350 |
|
356 | 0.000 | 0.485 | 0.497 | 0.444 | 0.432 | 0.458 | 0.570 | 0.309 | 0.329 |
|
357 | 0.408 | 0.522 | 0.497 | 0.455 | 0.445 | 0.458 | 0.568 | 0.280 | 0.327 |
|
⋮ | ⋮ | ⋮ | ⋮ | ⋮ | ⋮ | ⋮ | ⋮ | ⋮ | ⋮ | ⋮ |
587 | 0.000 | 0.485 | 0.497 | 0.444 | 0.432 | 0.458 | 0.570 | 0.309 | 0.329 |
|
588 | 0.773 | 0.556 | 0.990 | 0.659 | 0.653 | 0.458 | 0.568 | 0.304 | 0.348 |
|
589 | 0.850 | 0.503 | 0.907 | 0.630 | 0.619 | 0.458 | 0.568 | 0.301 | 0.348 |
|
590 | 0.854 | 0.540 | 0.861 | 0.587 | 0.635 | 0.458 | 0.568 | 0.303 | 0.349 |
|
⋮ | ⋮ | ⋮ | ⋮ | ⋮ | ⋮ | ⋮ | ⋮ | ⋮ | ⋮ | ⋮ |
753 | 0.943 | 0.506 | 0.509 | 0.816 | 0.783 | 0.461 | 0.570 | 0.311 | 0.359 |
|
754 | 0.950 | 0.509 | 0.707 | 0.630 | 0.619 | 0.458 | 0.568 | 0.301 | 0.348 |
|
755 | 0.000 | 0.481 | 0.497 | 0.445 | 0.432 | 0.458 | 0.570 | 0.313 | 0.352 |
|
756 | 0.000 | 0.494 | 0.497 | 0.449 | 0.431 | 0.458 | 0.570 | 0.313 | 0.352 |
|
⋮ | ⋮ | ⋮ | ⋮ | ⋮ | ⋮ | ⋮ | ⋮ | ⋮ | ⋮ | ⋮ |
893 | 0.042 | 0.491 | 0.497 | 0.386 | 0.329 | 0.458 | 0.570 | 0.294 | 0.339 |
|
894 | 0.151 | 0.494 | 0.491 | 0.405 | 0.419 | 0.458 | 0.570 | 0.290 | 0.342 |
|
895 | 0.950 | 0.494 | 0.503 | 0.708 | 0.724 | 0.461 | 0.570 | 0.296 | 0.354 |
|
896 | 0.399 | 0.488 | 0.491 | 0.389 | 0.393 | 0.458 | 0.570 | 0.288 | 0.338 |
|
⋮ | ⋮ | ⋮ | ⋮ | ⋮ | ⋮ | ⋮ | ⋮ | ⋮ | ⋮ | ⋮ |
1072 | 0.950 | 0.506 | 0.506 | 0.621 | 0.783 | 0.461 | 0.570 | 0.297 | 0.358 |
|
1073 | 0.948 | 0.506 | 0.500 | 0.695 | 0.658 | 0.461 | 0.570 | 0.298 | 0.358 |
|
1074 | 0.950 | 0.509 | 0.707 | 0.630 | 0.619 | 0.458 | 0.568 | 0.301 | 0.348 |
|
1075 | 0.854 | 0.506 | 0.920 | 0.587 | 0.635 | 0.458 | 0.568 | 0.303 | 0.349 |
|
⋮ | ⋮ | ⋮ | ⋮ | ⋮ | ⋮ | ⋮ | ⋮ | ⋮ | ⋮ | ⋮ |
1200 | 0.347 | 0.500 | 0.475 | 0.535 | 0.534 | 0.464 | 0.527 | 0.298 | 0.326 |
|
After the assessment model based on artificial immune algorithm was trained, the multiclass classifiers of the assessment system were constituted by four detectors, and each detector only could not recognize corresponding class object of particular assessment consequence mode. Actually, if the input schema object only could not be recognized by one detector, then the schema object belongs to this class object.
After the training phase, an assessment system could be obtained. In order to verify the accuracy of the model, the remaining 200 samples were utilized to test its performance. The prediction consequence was given as in Figure
Classification results of the classifier based on artificial immune algorithm.
In order to indicate the meliority of assessment model based on artificial immune algorithm, the assessment model based on back propagation-neural network (BP-NN) and support vector machine (SVM) was provided to solve the problem of the above example. The training samples and testing samples were the same. The configurations of simulation environment for three algorithms were uniform and the relevant parameters were in common with the above example. The prediction consequence of the assessment model based on BP-NN was given as in Figure
Classification results of the classifier based on BP-NN.
The prediction consequence of the assessment model based on SVM was given as in Figure
Classification results of the classifier based on SVM.
In order to further compare and analyze the overall performance of SVM, BP-NN, and AI, the same 1200 samples are experimented with. In this example, a certain number of samples, denoted by training size (
Figure
The changes of classification error rate with different training sizes.
From Figure
In this section, a system based on the proposed approach had been developed and applied in the field of shearer dynamic health assessment as shown in Figure
Hardware construction in fully mechanized coal face.
As Figure
In order to illustrate the application effect of the proposed approach, the shearer was running in fully mechanized coal face from 135.0 m to 150.0 m by the manual operation. The dynamic health assessment curve based on the proposed classifier was shown in Figure
The dynamic health assessment curve of shearer based on the proposed system.
The operational parameters curve of shearer with the manual operation.
The main contribution of this paper was that a methodology based on artificial immune algorithm for the assessment of shearer dynamic health status was presented. The detailed flows for the proposed approach were described, including three critical steps, that is, assessment indicators selecting, data acquisition and initialization, and multiclass classifiers training and testing. In order to verify the feasibility and efficiency of the proposed approach, a simulation example was provided and some comparisons with other algorithms were carried out. The simulation results showed that the proposed approach was outperforming others. Finally, the proposed approach was applied to an engineering problem of shearer dynamic health assessment. The industrial application results showed that the paper research achievements could be used combining with shearer automation control system in fully mechanized coal face and had obvious effectiveness on reducing operating trouble and production accident of shearer and improving coal production efficiency further. The artificial immune algorithm could obtain a relatively high accuracy to provide an effective support tool for dynamic health assessment for shearer.
In future studies, the authors plan to investigate some improvements for the proposed approach. Possible improvements may include the combination of artificial immune algorithm with other intelligent algorithms for better performance. In addition, the applications of the proposed approach in dynamic health assessment domain are worth further study from the authors.
The authors declare that there is no conflict of interests regarding the publication of this paper.
The support of Joint Funds of the National Natural Science Foundation of China (no. U1510117), National Key Basic Research Program of China: Key Fundamental Research on the Unmanned Mining Equipment in Deep Dangerous Coal Bed (no. 2014CB046301), the Priority Academic Program Development (PAPD) of Jiangsu Higher Education Institutions, and the Innovation Funds of Production and Research Cooperation Project in Jiangsu Province (BY2014107) in carrying out this research is gratefully acknowledged.