基于SVM的无轨胶轮车全液压制动系统故障诊断研究

发布时间：2018-05-26 04:53

本文选题：无轨胶轮车 + 全液压制动系统　；参考：《山东科技大学》2017年硕士论文

【摘要】：经国内外实践经验表明,矿用无轨运输设备在现代化矿井的运作中发挥着重要作用。与传统的运输方式相比,矿用无轨运输设备具有载重量大,速度快,效率高等独特优势,因此,可以在很大程度上提高矿井运输能力、生产能力和社会经济效益。其中,无轨胶轮车因其较好的灵活性得到广泛使用。为了提高行车安全性和制动可靠性,无轨胶轮车一般采用全液压双回路制动系统。由于煤矿井下环境比较恶劣,无轨胶轮车要长期承受巨大的工作负荷,同时在大载荷下车辆还要进行频繁的启动与制动等操作,因此制动回路中的制动器、蓄能器、充液阀以及液压管路等部件会不可避免地出现一些故障,使车辆在运输过程中失去控制,严重时则可能造成矿井人员伤亡,带来不必要的人财损失。因此,针对无轨胶轮车全液压制动系统进行故障诊断研究,对于煤矿安全、高效地生产具有重大价值。首先,根据无轨胶轮车的全液压制动系统结构组成和重要工作参数,在AMESim环境下搭建了全液压制动系统仿真模型,得到前桥蓄能器、后桥蓄能器、前桥制动器、后桥制动器的压力输出曲线。结合全液压制动系统的工作原理,对系统的动态响应性能进行分析,验证了所建立AMESim仿真模型的合理性。同时,通过该仿真模型获得了前后桥蓄能器、前后桥制动器的压力输出数据,以及制动踏板压力输入数据,为下一步进行支持向量机故障诊断提供了可靠的原始数据样本。其次,基于SVM回归预测故障诊断原理,建立了 4个SVM回归预测模型,分别对前桥蓄能器压力、后桥蓄能器压力、前桥制动器压力以及后桥制动器压力进行了训练和校验,得到的建模误差和校验误差均在10-2～10-1数量级,验证了 SVM模型的推广性能。再通过所建立的SVM回归预测模型对故障数据进行预测,结果发现在故障数据下得到的残差值发生突变,有效诊断出了相应的故障及其发生时间,验证了 SVM在全液压制动系统故障诊断中的可行性。为了进一步增强SVM故障预测模型的诊断性能,有效提高故障诊断正确率,利用交叉验证、遗传算法以及粒子群算法分别对SVM故障预测模型的核参数g以及惩罚参数c进行优化。在最佳参数g和参数c下再次对SVM回归模型进行训练与预测,得到诊断效果更优的SVM故障预测模型,其预测精度由原来的10-3提升到10-5数量级,结果令人满意。最后,基于Labwindows/CVI软件设计了一套全液压制动系统状态监测与故障诊断系统,通过友好的人机界面对液压制动系统的状态进行实时监控,并利用ActiveX技术调用MATLAB支持向量机故障预测模型程序,实现了对全液压制动系统的状态监测与故障诊断。
[Abstract]:The practical experience at home and abroad shows that the trackless transport equipment plays an important role in the operation of modern mines. Compared with the traditional transportation mode, the mine trackless transportation equipment has the unique advantages of large load, high speed and high efficiency, so it can greatly improve the mine transportation capacity, production capacity and social and economic benefits. Among them, trackless rubber wheel car is widely used because of its good flexibility. In order to improve driving safety and braking reliability, trackless rubber wheel cars generally adopt full hydraulic double-loop braking system. Because the underground environment of coal mine is relatively bad, the trackless rubber wheel car has to bear a huge workload for a long time, and at the same time, the vehicle has to carry on frequent operation such as starting and braking under the heavy load, so the brake and accumulator in the brake circuit, Some faults will inevitably occur in the hydraulic valve and hydraulic pipeline, which will make the vehicle out of control during the transportation process, and may cause mine personnel casualties and unnecessary loss of human wealth when serious. Therefore, the research of fault diagnosis for the full hydraulic braking system of trackless rubber wheel car is of great value to the safe and efficient production of coal mine. First of all, according to the structure composition and important working parameters of the full hydraulic brake system of the trackless rubber wheel car, the simulation model of the full hydraulic braking system is built under the AMESim environment, and the front axle accumulator, the rear axle accumulator and the front axle brake are obtained. Pressure output curve of rear axle brake. Combined with the working principle of the full hydraulic braking system, the dynamic response performance of the system is analyzed, and the rationality of the established AMESim simulation model is verified. At the same time, the output data of the front and rear axle accumulator, the pressure output of the front and rear axle brake and the input data of the brake pedal pressure are obtained through the simulation model, which provides a reliable original data sample for the next step in the fault diagnosis of support vector machine. Secondly, based on the principle of SVM regression prediction fault diagnosis, four SVM regression prediction models are established. The pressure of front axle accumulator, rear axle accumulator, front axle brake and rear axle brake are trained and calibrated respectively. Both the modeling error and the calibration error are in the order of 10 ~ (-2) ~ 10 ~ (-1), which verifies the extended performance of the SVM model. Then the fault data are predicted by the established SVM regression prediction model. The results show that the residual value under the fault data has a sudden change, and the corresponding fault and its occurrence time are effectively diagnosed. The feasibility of SVM in fault diagnosis of full hydraulic braking system is verified. In order to further enhance the diagnosis performance of SVM fault prediction model and improve the accuracy of fault diagnosis effectively, the kernel parameters g and penalty parameter c of SVM fault prediction model are optimized by cross validation, genetic algorithm and particle swarm optimization algorithm, respectively. Under the optimal parameters g and c, the SVM regression model is trained and predicted again, and a SVM fault prediction model with better diagnostic effect is obtained. The prediction accuracy is improved from 10-3 to 10-5, and the results are satisfactory. Finally, a full hydraulic braking system condition monitoring and fault diagnosis system is designed based on Labwindows/CVI software. The condition of hydraulic braking system is monitored in real time through friendly man-machine interface. The condition monitoring and fault diagnosis of full hydraulic braking system are realized by calling the MATLAB support vector machine fault prediction model program with ActiveX technology.
【学位授予单位】：山东科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TD525

【参考文献】