数据驱动的无模型自适应轧机液压位置控制

关于轧机液压伺服系统性能优化的研究,针对轧机液压伺服位置系统是一个多变量、强耦合、非线性控制系统,传统PID需建立被控对象的精确数学模型,难以处理复杂非线性问题,神经网络控制又需要进行在线与离线系统辨识,且计算量大、适合阶数已知的被控系统,而数据驱动仅使用被控系统的在线与离线I/O数据便可完成控制器的设计,无需任何过程数学模型信息,因此提出基于无模型自适应的液压位置控制方法。在相同条件下,通过PID、神经网络、无模型自适应控制三种算法对液压位置系统进行仿真比较,结果表明,无模型自适应控制收敛快,稳态误差小,且提高对扰动和模型变化的鲁棒性。

轧机液压控制系统故障仿真研究

液压AGC系统是实现板材高质量不可缺少的技术手段。本文针对液压AGC系统的液压回路工作原理与控制系统原理,采用AMESim软件建立其仿真模型,讨论了应用仿真模型对轧机AGC系统进行性能分析的可行性。之后,对液压AGC系统的常见故障进行模拟,包括伺服阀前后止回阀的先导控制阀故障模拟,液压缸内泄漏故障模拟,伺服阀阀芯卡死卡滞故障模拟和伺服阀位移传感器增益异常故障模拟。研究对液压AGC系统的故障诊断具有指导意义。

Identification and optimization for hydraulic roll gap control in strip rolling mill

In order to improve the control performance of strip rolling mill, theoretical model of the hydraulic gap control(HGC) system was established. HGC system offline identification scheme was designed for a tandem cold strip mill, the system model parameters were identified by ARX model, and the identified model was verified. Taking the offline identified parameters as the initial values, online identification using recursive least square was carried out with model parameters changing. For the purpose of improving system robustness and decreasing the sensitivity due to model errors, the HGC system based on generalized predictive control(GPC) was designed, and simulation experiments for traditional controller and GPC controller were conducted. The results show that both controllers acquire good control effect with model matching. When the model mismatches, for the traditional controller, the overshot will increase to 76.7% and the rising time will increase to 165.7 ms, which cannot be accepted by HGC system; for the GPC controller, the overshot is less than 8.5%, and the rising time is less than 26 ms in any case.