带式输送机作为旋转运输机械，广泛适用于采矿、海关以及物料工厂，由于连续运转时间长、负载大等极易发生内部损坏故障。对于煤矿行业，采用的带式输送机通常可以达到2-3km，其故障的排除依赖于人工巡检，该方法存在实际巡检难度大、故障排除不及时等弊端。为此，设计开发巡检机器人对于带式输送机的平稳运转以及企业的安全生产有着长远的意义和生产实践效益。该研究的具体工作如下： 首先，通过对带式输送机整体结构、物料运输原理以及故障发生机理等方面的研究，对带式输送机巡检机器人的实体机构展开三维建模。所设计的带式输送机巡检机器人能量供给方式采用安全高效的无线供电方式，避免电线的错乱排布。通过在装置上安装多类信息传感器，获取带式输送机工作状态及周围环境参数。 其次，针对带式输送机托辊故障时有发生且排除难度大等问题，以托辊从正常到故障过程为研究对象，分析其失效因素及最终发生故障后的时频特性；在巡检机器人动态采集的基础上，提出一种基于音频采集的无损检测方案，该方案采取经验模态-小波包分解方法对托辊故障信号中脉冲激励频率进行提取，并通过BP神经网络模型对故障样本进行分类训练，试验结果表明该分类方法具有较高的故障分辨率。 最后，采取模块化的设计原则，搭建LabVIEW实时检测系统，将采集的音频进行模型去噪、能谱计算、BP训练及测试，最终对故障特征做出识别；利用TCP/IP通信协议，实现应用软件与信息采集装置WIFI模块之间的通信；通过对无线供电机器人安装调试实现自动采集、智能处理以及基本分析等要求，最终满足自动化智能巡检要求。 本研究中，带式输送机巡检机器人将无损检测技术与智能技术相结合，提高了带式输送机故障诊断机制的灵活性和可靠性，对智能化井下巡检具有较高的实践工程意义。

Belt conveyor, as a rotating transport machine, is widely used in mining, customs and material factories. Due to long continuous operation time and heavy load, it is easy to cause internal damage. For the coal mine industry, the belt conveyor usually can reach 2-3km, and its troubleshooting depends on manual inspection. This method has the disadvantages of great difficulty in actual inspection and untimely troubleshooting. Therefore, the design and development of inspection robot has long-term significance and practical benefits for the smooth operation of belt conveyor and the safe production of enterprises. The specific works are as follows: Firstly, by researching the overall structure of the belt conveyor, the principle of material transportation and the mechanism of failure, the solid mechanism of the belt conveyor inspection robot is modeled in three dimensions. The energy supply mode of the designed belt conveyor inspection robot adopts a safe and efficient Wireless power transfer mode to avoid disordered distribution of wires. Moreover, by installing various types of information sensors on the device, the working state and surrounding environment parameters of the belt conveyor can be obtained. Secondly, aiming at the problems such as the occurrence and difficulty in eliminating the roll failure of belt conveyor, taking the process of roll failure from normal to failure as the research object, the failure factors and the time-frequency characteristics after the final failure are analyzed. Based on the dynamic collection of inspection robot, a nondestructive testing scheme based on audio collection is proposed. The empirical mode-wavelet packet decomposition method is adopted to extract the pulse excitation frequency in the roller fault signal, and the BP neural network model is used to classify and train the fault samples. Experimental results show that the classification method has high fault resolution. Finally, based on the modular design principle, a LabVIEW real-time detection system is built to carry out model denoising, energy spectrum calculation, BP training and testing on the collected audio, and finally identify the fault features. The communication between application software and WIFI module of information acquisition device is realized with TCP/IP communication protocol. Through the installation and debugging of Wireless power transfer robots, the requirements of automatic acquisition, intelligent processing and basic analysis are realized. The requirements of automatic intelligent inspection are met ultimately. In this study, the non-destructive inspection technology is combined with intelligent technology by the inspection robot, which improves the flexibility and reliability of fault diagnosis mechanism of belt conveyor. Therefore, it has higher practical engineering significance for intelligent underground inspection.