Bayesian discriminant analysis for prediction of coal and gas outbursts and application

Based on the principle of Bayesian discriminant analysis, we established a model of Bayesian discriminant analysis for predicting coal and gas outbursts. We selected five major indices which affect outbursts, i.e., initial speed of methane diffusion, a consistent coal coefficient, gas pressure, destructive style of coal and mining depth, as discriminating factors of the model. In our model, we divided the type of coal and gas outbursts into four grades regarded as four normal populations. We then obtained the corresponding discriminant functions through training a set of data from engineering examples as learning samples and evaluated their criteria by a back substitution method to verify the optimal properties of the model. Finally, we applied the model to the prediction of coal and gas outbursts in the Yunnan Enhong Mine. Our results coincided completely with the actual situation. These results show that a model of Bayesian discriminant analysis has excellent recognition performance, high prediction accuracy and a low error rate and is an effective method to predict coal and gas outbursts.

Quantitative Method of Classification and Discrimination of a Porous Carbonate Reservoir Integrating K-means Clustering and Bayesian Theory

Reservoir classification is a key link in reservoir evaluation.However,traditional manual means are inefficient,subjective,and classification standards are not uniform.Therefore,taking the Mishrif Formation of the Western Iraq as an example,a new reservoir classification and discrimination method is established by using the K-means clustering method and the Bayesian discrimination method.These methods are applied to non-cored wells to calculate the discrimination accuracy of the reservoir type,and thus the main reasons for low accuracy of reservoir discrimination are clarified.The results show that the discrimination accuracy of reservoir type based on K-means clustering and Bayesian stepwise discrimination is strongly related to the accuracy of the core data.The discrimination accuracy rate of TypeⅠ,TypeⅡ,and TypeⅤreservoirs is found to be significantly higher than that of TypeⅢand TypeⅣreservoirs using the method of combining K-means clustering and Bayesian theory based on logging data.Although the recognition accuracy of the new methodology for the TypeⅣreservoir is low,with average accuracy the new method has reached more than 82%in the entire study area,which lays a good foundation for rapid and accurate discrimination of reservoir types and the fine evaluation of a reservoir.

Log interpretation of carbonate rocks based on petrophysical facies constraints

The complex pore structure of carbonate reservoirs hinders the correlation between porosity and permeability.In view of the sedimentation,diagenesis,testing,and production characteristics of carbonate reservoirs in the study area,combined with the current trends and advances in well log interpretation techniques for carbonate reservoirs,a log interpretation technology route of“geological information constraint+deep learning”was developed.The principal component analysis(PCA)was employed to establish lithology identification criteria with an accuracy of 91%.The Bayesian stepwise discriminant method was used to construct a sedimentary microfacies identification method with an accuracy of 90.5%.Based on production data,the main lithologies and sedimentary microfacies of effective reservoirs were determined,and 10 petrophysical facies with effective reservoir characteristics were identified.Constrained by petrophysical facies,the mean interpretation error of porosity compared to core analysis results is 2.7%,and the ratio of interpreted permeability to core analysis is within one order of magnitude,averaging 3.6.The research results demonstrate that deep learning algorithms can uncover the correlation in carbonate reservoir well logging data.Integrating geological and production data and selecting appropriate machine learning algorithms can significantly improve the accuracy of well log interpretation for carbonate reservoirs.

Probabilistic Methods in Multi-Class Brain-Computer Interface

Abstract-Two probabilistic methods are extended to research multi-class motor imagery of brain-computer interface （BCI）： support vector machine （SVM） with posteriori probability （PSVM） and Bayesian linear discriminant analysis with probabilistic output （PBLDA）. A comparative evaluation of these two methods is conducted. The results shows that： 1） probabilistie information can improve the performance of BCI for subjects with high kappa coefficient, and 2） PSVM usually results in a stable kappa coefficient whereas PBLDA is more efficient in estimating the model parameters.

基于贝叶斯判别的激光除漆声学监测方法研究

为了解决激光除漆声学监测方法难以满足实际生产需要的问题,采用贝叶斯判别方法进行了理论分析和实验验证,将除漆过程分为正在清洗、清洗完成且基底无损伤、基底损伤3种类别,结合光声效应分析除漆声信号在清洗过程的变化,提取特征参量建立判别模型,实现了对激光除漆的定量判别。结果表明,训练样本准确率达到99%,测试样本准确率达到98.7%。该方法具有较高的准确性和实用性,可为激光清洗声学监测的研究提供借鉴。

A model for discrimination and prediction of mental workload of aircraft cockpit display interface

With respect to the ergonomic evaluation and optimization in the mental task design of the aircraft cockpit display interface, the experimental measurement and theoretical modeling of mental workload were carried out under flight simulation task conditions using the performance evaluation, subjective evaluation and physiological measurement methods. The experimental results show that with an increased mental workload, the detection accuracy of flight operation significantly reduced and the reaction time was significantly prolonged; the standard deviation of R-R intervals（SDNN） significantly decreased, while the mean heart rate exhibited little change; the score of NASA_TLX scale significantly increased. On this basis, the indexes sensitive to mental workload were screened, and an integrated model for the discrimination and prediction of mental workload of aircraft cockpit display interface was established based on the Bayesian Fisher discrimination and classification method. The original validation and cross-validation methods were employed to test the accuracy of the results of discrimination and prediction of the integrated model, and the average prediction accuracies determined by these two methods are both higher than 85%. Meanwhile, the integrated model shows a higher accuracy in discrimination and prediction of mental workload compared with single indexes. The model proposed in this paper exhibits a satisfactory coincidence with the measured data and could accurately reflect the variation characteristics of the mental workload of aircraft cockpit display interface, thus providing a basis for the ergonomic evaluation and optimization design of the aircraft cockpit display interface in the future.