摘要

随着智能控制技术的发展，结合煤矿井下开采环境，实现工作面“少人”以及“无人化”开采是井下安全、高效生产的关键，同时也是煤矿智能开采的重要发展方向。刮板输送机作为综采工作面的主要运煤设备，承接采煤机开采过程中的落煤，且负载存在非线性、不稳定的特征，容易使刮板输送机因载荷突变而引起故障等问题。本文对刮板输送机电流进行特征幅值信息提取，引入卷积神经网络预测算法对刮板输送机负载进行预测，并以此作为刮板输送机速度控制的依据，利用模糊PID控制完成刮板输送机的速度调节。具体的研究工作如下：

（1）针对煤矿井下采运不平衡问题以及智能化的要求，分析刮板输送机的组成原理以及传动系统的特点，制定基于刮板输送机负载预测的调速控制总体方案；对浅层神经网络算法与深层神经网络算法进行分析，选定本文刮板输送机负载预测的算法。

（2）通过分析刮板输送机负载与电流之间的映射关系，以减速器齿轮啮合频率的幅值信息作为刮板输送机负载特性的表征；利用煤矿井下采集的电流数据进行去工频等处理，提取负载的特征幅值数据信息，建立负载幅值数据集，为刮板输送机负载预测模型提供数据支持。

（3）对卷积神经网络的网络架构以及特点进行分析，结合刮板输送机的短时特征性质，提出基于一维卷积的刮板输送机负载预测方法；建立基于卷积神经网络的刮板输送机负载预测模型，通过历史数据与预测数据的对比分析，验证模型的有效性，为煤矿井下刮板输送机负载预测研究提供实验室基础。

（4）在对刮板输送机负载进行预测的基础上，设计刮板输送机调速控制系统，利用模糊PID控制器幅值反馈的方式，以啮合频率幅值变化作为输入，实现刮板输送机的速度调节；利用Simulink软件仿真的方式对模糊PID控制器性能进行验证，并采用粒子群算法对其控制参数进行优化，使控制更加准确与高效。

ABSTRACT

With the development of intelligent control technology and the underground mining environment of the coal mine, the realization of "less people" and "unmanned" mining on the working face is the key to safe and efficient production, it is also an important development direction of coal mine intelligent mining. Scraper conveyor is the main coal conveying equipment in the fully mechanized coal mining face. It accepts coal falling during the mining process of the coal mining machine, and the load has nonlinear and unstable characteristics, which is easy to cause the scraper conveyor to fail due to sudden load changes. This paper extracts characteristic amplitude information of the scraper conveyor current, introduces a convolution neural network prediction algorithm to predict the load of the scraper conveyor, and uses this as the basis for the speed control of the scraper conveyor, using fuzzy PID control to complete the scraper Conveyor speed adjustment. The specific research work is as follows:

(1)Aiming at the problem of unbalanced mining and transportation in coal mines and the requirements of intelligence, the composition principle of the scraper conveyor and the characteristics of the transmission system are analyzed, and the overall scheme of speed control based on the load prediction of the scraper conveyor is formulated; the shallow neural network algorithm and The deep neural network algorithm was analyzed, and the load forecasting algorithm of the scraper conveyor was selected.

(2)By analyzing the mapping relationship between the load and current of the scraper conveyor, the amplitude information of the gear meshing frequency of the reducer is used as the characterization of the load characteristic of the scraper conveyor; the current data collected in the coal mine is used to remove the power frequency and other processing to extract the characteristic amplitude data information of the load, establish the load amplitude data set, and provide data support for the load prediction model of the scraper conveyor.

(3)The network architecture and characteristics of the convolution neural network are analyzed, combined with the short-term characteristic properties of the scraper conveyor, a load prediction method for the scraper conveyor based on one-dimensional convolution is proposed; A scraper conveyor based on the convolution neural network is established; The load forecasting model verifies the validity of the model through comparative analysis of historical data and forecasting data, and provides a laboratory basis for the load forecasting research of scraper conveyors in coal mine.

(4)Based on the load prediction of the scraper conveyor, the speed control system of the scraper conveyor is designed, and the amplitude feedback of the fuzzy PID controller is used, and the amplitude change of the meshing frequency is used as the input to realize the speed of the scraper conveyor; Simulink software is used to verify the performance of the fuzzy PID controller, and used particle swarm optimization to optimize its control parameters, making the control more accurate and efficient.