- 积分
- 992
- 注册时间
- 2014-12-4
- 最后登录
- 1970-1-1
- 打卡等级:偶尔看看
- 打卡总天数:105
- 打卡总奖励:319
|
楼主 |
发表于 2018-4-13 11:51:03
|
显示全部楼层
本帖最后由 PEN 于 2018-4-14 02:54 编辑
我通常可以在5%以内估计模型。大多数时候模型更接近。 20%是一个极端的建模错误。唯一不好的是负载变化很大的时候。那么最好用一点数学来估计负载变化并计算新的控制器增益。
以下80%的误差仅为0.2mm。当同步大型液压缸时,这通常足够好。
而且,实际位置很快到达指令位置。当同步多个气缸时,一个液压缸的重量可能会增加23%,并且只会有0.2mm的跟随误差。另外,请看速度,加速度和减速度。他们不像其他例子那样慢。
为了减少加速和减速过程中的误差,必须减小加速度和减速度。另一种选择是将闭环磁极在s平面中向左移动。积分器增益将迅速增加。
有一个实际限制。即使使用64位浮点数,系统也会变得不稳定。在此之前,反馈分辨率会导致系统不稳定,除非使用某些控制技巧来精确估计速度和加速度。
我希望你能问一下,如果P-only控制器的增益稳定在4.18%/ mm,闭环增益如此之高。
I can usually estimate the model within 5%. Most of the time the model is much closer. 20% is an extreme modeling error. The only time it is that bad is when the load changes a lot. Then it is better to use a little math to estimate the load change and calculate new controller gains.
The following error at 80% is only 0.2mm. That is usually good enough when synchronizing big hydraulic cylinders.
Also, the actual position reaches the command position quickly. When synchronizing multiple cylinders, one hydraulic cylinder can have 23% more weight and there would only be a 0.2mm following error. Also, look at the speed, acceleration and deceleration rates. They are not slow like other examples.
To reduce errors during acceleration and deceleration the acceleration and deceleration rates must be reduced. The other option is to move the closed loop poles farther to the left in the s-plane. The integrator gain will increase rapidly.
There is a practical limit. Even with 64 bit floating point number the system will go unstable. Before that happens the feed back resolution will cause the system to go unstable unless some control tricks are used to estimate the velocity and acceleration accurately.
I expected you to ask how can the closed loop gains be so high when a P only controller gain will be marginally stable at 4.18 %/mm.
增加气缸直径将减少以下错误,但这会花费额外的资金。 增加气缸直径会增加固有频率。 气缸直径加倍使自然频率加倍。 这也允许在理论上有更高的收益,但反馈分辨率是问题。
七阶运动轮廓应该有助于理论上的理由,因为在斜坡的开始和结束时目标加速度为0。 在实践和模拟中,情况并非如此,因为在相同的斜坡时间使用七阶运动曲线时,峰值加速度较高。
Back2049,尝试使用具有速度和加速度前馈的PID优化增益。 我敢打赌,即使执行器参数为K = 10,DF = 0.3333和NF = 10,以下误差也会远远大于0.2mm
|
|