没有合适的资源?快使用搜索试试~ 我知道了~
资源推荐
资源详情
资源评论
摘要
I
摘要
电动物流车整车质量随载重不同而产生较大变化,若仍使用传统的电动助力转向
系统(简称 EPS 系统)设计方法,不能保证驾驶员始终获得清晰路感,影响驾驶体验。
同时,电动物流车主要运行于城市工况,车辆行驶速度不高,且转向频繁,由此对转
向系统的轻便性提出了更高的要求。本文主要针对电动物流车的 EPS 系统展开研究,
建立适用于电动物流车的全质量助力特性,解决因质量变化而引起的转向手感丧失以
及轻便性不佳的问题,并对电动物流车 EPS 系统的控制策略进行研究。本文的主要研
究内容如下:
(1)建立电动物流车质量估计模型。将汽车纵向运动学模型中的坡度、速度与质
量解耦,运用跟踪微分器对加速度、驱动力数据进行预处理,而后使用卡尔曼滤波法
对整车质量进行估计。随后建立了全车质量覆盖范围下的曲线型助力特性,使汽车在
任意质量下、速度与方向盘转矩下均能获得理想的助力力矩。并建立 BP神经网络对助
力力矩进行求解,代替传统的汽车 ECU 中的查表方式,减少 ECU 存储空间使用量。
(2)对助力电机的控制策略进行研究。针对传统 PID 控制存在着对电机电流敏感
性大、易放大高频噪声以及调参困难等问题,本文使用了基于 BP神经网络的不完全微
分 PID 控制器对电机进行控制,并引入 RBF 神经网络作为系统辨识器,求解系统的
Jacobian 信息,以提升 BP 神经网络的学习效率。为使控制系统更适用于工业实际需求,
对BP神经网络进行了改进,增加单连接线性单元以及输出层偏置,使神经网络输出的
不完全微分 PID 控制参数具有理想的初值以及可调节的范围,进一步提升控制性能。
(3)使用 CARSIM 与 SIMULINK 进行联合仿真。通过在固定坡度路面以及变坡
度路面仿真,验证了质量估计模型的有效性。对车辆进行双纽线工况以及方向盘角阶
跃仿真,证明建立的全质量助力特性 EPS 系统满足轻便性要求,并提升了转向手感与
车辆的操纵稳定性。通过 PI 控制与基于 RBF 辨识的神经网络不完全微分 PID 控制、改
进后的基于 RBF 辨识的神经网络不完全微分 PID 控制进行控制效果对比,验证了本文
所使用控制策略具有良好控制性能,并证明改进后的 BP神经网络改善了控制性能,在
阶跃目标信号下,超调量由 2.02%降为了 0.36%,在正弦目标电流下具有很小的跟踪误
差。
(4)运用 EPS 台架系统与单片机搭建试验平台,并进行试验程序设计。对 PI 控
制与改进后的基于 RBF 辨识的神经网络不完全微分 PID 控制进行了阶跃目标信号与斜
坡目标信号下的控制性能对比。试验结果表明,本文使用控制策略比 PI 控制调节时间
少 0.26s,超调量降低了 0.5%,跟踪性能更加优越。
盐城工学院硕士学位论文
II
本文对电动物流车辆 EPS 系统展开研究,设计了质量估计模型与全质量助力特性
EPS 系统,并对 EPS 系统控制策略进行研究。仿真与试验结果证明了本文所设计质量
估计方法与全质量助力特性 EPS 系统的可行性,以及控制策略的良好控制性能。
关键词:EPS,质量估计,跟踪微分器,助力特性,神经网络
ABSTRACT
III
Abstract
The weight of electric logistics vehicles varies greatly with different loads. If
traditional Electric Power Steering (EPS) design methods are still used, it can’t guarantee
the driver always has a clear road feel and may impact driving experience. At the same
time, electric logistics cars mainly operate under city conditions at low speeds and
frequent turning, which raises higher requirements for the agility of steer assist system.
Focusing on the EPS system of electric logistics vehicles in this paper, it establishes a
full-weight power-assisted characteristic suitable for electric logistics vehicles, solves the
problem of steering loss and poor portability performance caused by weight changes, and
research on the control strategies of the EPS system of electric logistics vehicles. The
main research contents of this paper are as follows:
(1) A quality estimation model for electric logistics vehicles was established,
decoupling the slope, speed, and quality in the longitudinal vehicle dynamics model. The
acceleration and driving force data were preprocessed using a tracking differentiator, and
then using the Kalman filter method to estimate the overall vehicle quality. Subsequently,
a curve-type assist characteristic covering the entire vehicle quality range was established,
allowing the vehicle to obtain ideal assist torque at any quality, speed, and steering wheel
torque. Establishing a BP neural network to solve the assist torque to replace the table
lookup method in traditional automobile ECU and reduce ECU storage space usage.
(2) Research on the control strategy of the assist motor. Aiming at the problems of
traditional PID control such as high sensitivity to motor current, easy amplification of
high-frequency noise, and difficulty in parameter adjustment, this paper uses an
incomplete differential PID controller based on BP neural network to control the motor,
and introduces RBF neural network as a The system identifier solves the Jacobian
information of the system to improve the learning efficiency of the BP neural network. In
order to make the control system more suitable for the actual needs of the industry, the
BP neural network has been improved, adding single-connected linear units and output
layer bias, so that the incomplete differential PID control parameters output by the neural
network have ideal initial values and adjustable ranges , to further improve the control
performance.
(3) Co-simulation using CARSIM and MATLAB/SIMULINK. The effectiveness of
the mass estimation model is verified through simulations on both fixed and variable
盐城工学院硕士学位论文
IV
slope roads. Lemniscate conditions and steering wheel angle step simulations are
performed on the vehicle to demonstrate that the established full-quality assisted
characteristic EPS system meets the requirements for lightness and improves the steering
feel and vehicle handling stability. By comparing the control effects of PI control and
incomplete differential PID control based on RBF identification and neural network, as
well as improved incomplete differential PID control based on RBF identification and
neural network, the effectiveness of the control strategy in this paper was verified. The
improvement of the neural network was proved to improve the control performance, and
the overshoot was reduced from 2.02% to 0.36% under step target signal. It had a small
tracking error under sinusoidal target current.
(4) An experimental platform was built using the EPS bench system and single-chip
microcomputer, and the experimental program was designed. The control performance of
PI control and improved PID control based on RBF identification and incomplete
differential neural network was compared under step target signal and slope target signal.
The experimental results show that the control strategy used in this paper reduces the
regulation time by 0.26 seconds compared to PI control, lowers overshoot by 0.5%, and
exhibits superior tracking performance.
This paper studies the EPS system of electric logistics vehicles and designs a mass
estimation model and a full-mass power-assisted characteristic EPS system. The control
strategy of the EPS system is also studied. Simulation and experimental results prove the
feasibility of the mass estimation method and full-mass power-assisted characteristic EPS
system designed in this paper, as well as the good control performance of the control
strategy.
Key Words: EPS, mass estimation, tracking differentiator, assist characteristic,
Neural network
目 录
1 绪论 ......................................................................................................... 1
1.1 研究背景及意义 ............................................................................... 1
1.2 国内外研究现状 ............................................................................... 2
1.2.1 EPS 系统国外研究现状 .............................................................. 2
1.2.2 EPS 系统国内研究现状 .............................................................. 4
1.2.3 汽车质量参数估计国内外研究现状......................................... 6
1.3 本文主要研究内容 ........................................................................... 8
2 电动物流车质量估计 .......................................................................... 10
2.1 纵向动力学模型建立 ..................................................................... 10
2.2 质量估计模型设计 ......................................................................... 11
2.2.1 基于跟踪微分器的信号预处理 ............................................... 12
2.2.2 基于卡尔曼滤波的质量参数估计 ........................................... 13
2.2.3 整车质量的状态空间描述 ....................................................... 15
2.3 本章小结 ......................................................................................... 15
3 转向系统建模与全质量助力特性设计 .............................................. 16
3.1 EPS 系统模型建立 .......................................................................... 16
3.1.1 EPS 系统动力学模型 ................................................................ 16
3.1.2 车辆转向反馈力矩模型 ........................................................... 18
3.2 助力特性曲线选择 ......................................................................... 19
3.3 助力特性设计 ................................................................................. 20
3.3.1 原地转向时转向盘阻力矩 ....................................................... 20
3.3.2 行驶过程中的转向盘阻力矩 ................................................... 21
3.3.3 助力力矩求解 ........................................................................... 22
3.4 基于 BP 神经网络的助力特性拟合 .............................................. 25
3.4.1 BP 神经网络的正向传播 .......................................................... 26
3.4.2 BP 神经网络的反向传播 .......................................................... 27
3.4.3 模型建立及训练 ....................................................................... 28
3.5 本章小结 ......................................................................................... 31
4 EPS 系统控制模式与控制策略研究 ................................................... 32
4.1 EPS 系统控制模式 .......................................................................... 32
4.2 不完全微分 PID 控制 .................................................................... 34
4.3 基于 BP 神经网络的不完全微分 PID 参数整定 ......................... 35
4.4 BP 神经网络的改进 ........................................................................ 38
4.5 基于 RBF 神经网络的系统辨识 ................................................... 41
剩余84页未读,继续阅读
资源评论
2301_77550592
- 粉丝: 17
- 资源: 7163
上传资源 快速赚钱
- 我的内容管理 展开
- 我的资源 快来上传第一个资源
- 我的收益 登录查看自己的收益
- 我的积分 登录查看自己的积分
- 我的C币 登录后查看C币余额
- 我的收藏
- 我的下载
- 下载帮助
安全验证
文档复制为VIP权益,开通VIP直接复制
信息提交成功