摘要
针对采用理想化信号模拟滚筒受载进行采煤机调高液压系统性能分析准确性差、基于传统优化控制算法难以实现对采煤机滚筒调高的自适应控制、响应速度和跟踪性能不好等问题,提出一种基于深度确定性策略梯度算法DDPG(Deep Deterministic Policy Gradient)的采煤机滚筒自适应调高控制策略,并利用虚拟样机技术、深层卷积神经网络(DCNN)与深度强化学习等机器学习算法搭建了采煤机自适应调高机-液-控一体化系统。利用Pro/E及RecurDyn建立采煤机调高系统刚柔耦合动力学仿真模型,根据某采煤工作面实际赋存条件,利用EDEM建立离散元煤壁模型,基于DEM-MFBD接口构建EDEM-RecurDyn双向耦合调高机构机械系统模型,基于AMEsim建立调高机构液压系统模型,利用Simulink搭建了集信号处理模块(Signal processing)、时频谱图生成模块(Continuous wavelet transform System)、数据样本扩充模块(Fancy PCA System)、截割状态识别模块(Alexnet Transfer Learning System)、调高控制决策模块(Height Control decision)和DDPG高度调节模型模块(DDPG Height Adjustment Model)6个模块于一体的采煤机自适应调高控制系统模型,基于接口技术搭建EDEM-RecurDyn-AMEsim-Simulink多领域协同仿真的采煤机自适应调高机-液-控一体化系统模型。利用该系统模型进行仿真并对其调高性能分析,研究结果表明:基于连续小波变换、Fancy PCA和Alexnet网络迁移学习能够实现煤岩截割状态的精准识别,识别准确率可达95.58%,所搭建系统的仿真过程能够更真实地模拟采煤机截割煤岩破碎过程,系统仅经0.6 s左右即能感知到截割工况的变化,且能够快速识别出煤岩截割状态并准确地将滚筒调整至目标高度,响应速度快,能够根据工况变化自适应调节活塞运动速度;相比于模糊PID控制,基于DDPG控制的采煤机自适应调高系统的活塞缩回位移稳态误差最大仅为0.002 1 mm,为前者的0.6
In view of the poor accuracy of performance analysis of shearer drum height adjustment hydraulic system by using idealized signal to simulate drum load,it is difficult to realize self-adaptive control of shearer drum height adjustment,poor response speed and tracking performance based on traditional optimization control algorithm.A selfadaptive height adjustment control strategy of shearer drum based on depth deterministic gradient algorithm DDPG is proposed,and using virtual prototype technology,deep convolutional neural network and deep reinforcement learning and other machine learning algorithms to build an integrated hydraulic control system for shearer self-adaptive height adjustment.The rigid flexible coupling dynamic simulation model of shearer height adjustment system is established by using Pro/E and RecurDyn.According to the actual occurrence conditions of a coal face,the discrete coal wall model is established by using EDEM,the mechanical system model of EDEM-RecurDyn bi-directional coupling height adjustment mechanism is constructed based on DEM-MFBD interface,and the hydraulic system model of height adjustment mechanism is established based on AMESim.A self-adaptive height adjustment control system model of shearer is built by using Simulink,which integrates six modules:signal processing module( Signal processing),time spectrum diagram generation module( Continuous wavelet transform System),data sample expansion module( Fancy PCA System),cutting state identification module(Alexnet Transfer Learning System),height adjustment control decision module( Height Control decision) and DDPG height adjustment model module( DDPG Height Adjustment Model).Based on the interface technology,the integrated hydraulic control system model of shearer selfadaptive height regulator based on EDEM-RecurDyn-AMESim-Simulink multi domain collaborative simulation is built.The system model is used to simulate and analyze its height adjustment performance.The research results show that the accurate identification of coal and roc
作者
王雅东
赵丽娟
张美晨
WANG Yadong;ZHAO Lijuan;ZHANG Meichen(School of Mechanical Engineering,Liaoning Technical University,Fuxin 123000,China;Liaoning Provincial Key Laboratory of Large-Scale Mining Equipment,Fuxin 123000,China)
出处
《煤炭学报》
EI
CAS
CSCD
北大核心
2022年第9期3505-3522,共18页
Journal of China Coal Society
基金
国家自然科学基金资助项目(51674134)
教育部科技发展中心教育技术研究基金资助项目(2018A04025)
辽宁省教育厅基础资助项目(LJ2019JL024)。
关键词
自适应调高
DDPG控制
机-液-控一体化
迁移学习
Alexnet网络
多领域建模与协同仿真
采煤机
self-adaptive height adjustment
DDPG control
mechanical-hydraulic-control integration
transfer-learning
Alexnet network
multi-domain modeling and collaborative simulation
shearer
