- 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
|
发表于 2018-3-29 00:50:14
|
显示全部楼层
大多数人都会手动优化P,I和D控制器的增益。他们如何知道答案是最佳的?最佳的定义是什么?
大多数人在看结果时会改变一个控制器的增益。 20世纪90年代,我优化了这种收益。然后我意识到P,I和D控制器增益放置了闭环极点。闭环极点和零点决定了系统的响应。然后我开始考虑优化控制器的增益。我可以确定闭环电杆的最佳响应位置。然后我可以计算控制器的增益,将电杆放在理想的位置。
ITAE调试尝试找到放置闭环极点的最佳位置,以便在响应阶跃输入时使积分时间绝对误差最小化。还有一个是计算闭环极点的最佳位置。有IAE(综合绝对误差)和ISE或SSE(误差平方和)。每种方法都有其自身的优点和缺点。所有闭环极点位置都是最佳的,它们的版本略有不同。
我对人工智能持怀疑态度。 AI有一些很好的例子,但大多数应用程序不需要AI。一次又一次地制作相同的部分与扮演威麒(GO)或国际象棋的可能性无穷无异。
因为每块木头都是不同的,所以锯木厂里的木材也有无尽的可能性。优化如何切割木材的软件不在运动控制器上执行。有一个可以扫描木材的视觉系统。优化计算机并确定切割木材的最佳方式。优化程序可能会尝试比较数百种组合,并选择切割木材的最佳方法。切割木材的最佳方式是下载到运动控制器。我不会称这个AI。这只是一个奇特的算法,但除非出现问题,否则不需要人员
Most people optimze P, I and D controller gains manually. How do they know the response is optimal? What is the definition of optimal?
Most people change one controller gain at a time the look at the result. I optimized gains like this in the 1990s. Then I realized the P, I and D controller gains placed the closed loop poles. The closed loop poles and zeros determine the response of the system. Then I started to look at optimizing controller gains differently. I could determine where the closed loop poles should be for the best response. Then I could calculate the controller gains that put the poles in the desired position.
The ITAE tuning attemps to find the best place to put the closed loop poles so the integrated time absolute error is minimized when responding to a step input. There are other was of calculating the optimal position of the closed loop poles. There is the IAE ( integrated absolute error ) and the ISE or SSE ( sum of squared errors ). Each method has its own advantages and disadvantages. All have a slight different version of what closed loop pole locations are optimal.
I am skeptical about artificial intelligence. There are some good examples of AI but most applications do not need AI. Making the same part over and over again is not the same as playing weiqi ( GO ) or chess where the possibilities are endless.
Cuting wood in sawmills also have endless possibilities because each piece of wood is different. The software that optimizes how to cut the wood is not executed on the motion controller. There is a vision system that can scan the wood. The optimizing computers and determine the best way to cut the lumber. The optimizer program may try compare hundreds of combinations and chose the best way to cut the lumber. The best way to cut the wood is downloaded to the motion controller. I wouldn't call this AI. It is just a fancy algorithm but people are not required unless something goes wrong.
|
|