| 摘要: | PID控制器係工業界最普通的控制器，但若遇到模式參數不確定性(Uncertainty)、 非線性(Nonlinearity)系統，固定控制器增益(gain)值，系統控制性能常無法滿足需求，為能滿足複雜的功能需求(Complex objectives)，例如強健控制除了對性能需求外，對外來之干擾及內部系統參數的不確定要有一定強健功能，控制器增益(gain)須能可調。因此本論文採用智慧型PID控制器在不同非線性系統調整增益值的探討。 首先，由於基因演算法須使用大量電腦運算，所以使用電腦模擬先找出PID控制器的參數值，再把這些參數值帶入實際系統，去執行遮蔽式金屬弧焊接(SMAW)。但是在實際焊接時焊條長度、電弧長會改變。由於本實驗採用事先模擬出的PID控制器的參數值 離線作業(off-line)，所以無法隨著焊接過程中弧長的改變，即時調整控制器參數值，即時改變弧電流。因此焊接品質有些瑕疵不夠均勻，所以把使用基因演算調整PID增益值應用於Shielded metal arc welding (SMAW)焊接過程探討及結果放於附錄處。 其次，在實際控制系統幾乎含有飽和致動器，它可能產生積分終結現象，這種積分終結現象是由於積分器造成輸入飽和所造成。如果PID控制器使用固定增益值，假設一個控制系統的工作範圍很大，在設定點改變或有大的負載干擾時，致動器有可能被驅動到極限，此時致動器只能保持在它的極限，無法隨著系統的輸出做更進一部的調整，此時實際操作量小於預期，所以誤差較大。如果控制器包含積分器，則致動器飽和時的大誤差會一直被積分，初始時輸出為零，此時誤差值最大，積分器很快地輸出足夠大的值，而使得致動器(Actuator)達到飽和狀態，但由於致動器的飽和現象，使得誤差的修正需要更長時的時間，且產生較大的超越量，無法達到積分器的控制目標。這種積分控制系統因飽和非線性的因素而影響控制系統性能的現象，稱之為積分終結。(Controller windup) 。造成較長的安定時間(Settling time)與較大的超震量(Overshoot) 。但縱使解決積分終結問題，若是增益值太大，造成反應太快，控制器輸出值大於飽和致動器上下極限值，也會造成程序(Plant)飽和，產生程序終結(Plant windup)。本論文針對含有界扇形非線性(Sector-bounded nonlinearity)飽和元件探討，提出三種抗積分終結(Anti-windup)方法，使用Circle Criterion、Describing function (描述函數)method、Popov’s(波波夫) Criterion作不同的穩定度探討，且使用模糊控制系統(Fuzzy PD system)即時(on-line)調整PID控制器的增益值( )，修正誤差值，產生抗積分終結功能，避免產生 windup，改善控制性能(Control performance) Fixed-gain PID (Proportional-Integral-Derivate) controller is the most commonly used controller in the industry. However, it is obviously not able to fulfill the needs of a system with gain uncertainty and nonlinearity. In order to contain the case of complex objectives applied on the system, such as the external noises and the uncertain parameters of system, an adjustable gain controller is necessary for the system robustness requirement. This thesis aims on the study of intelligent PID controller with adjustable gains applied on different cases of nonlinear system. Intelligence defines as “the capacity to acquire and apply knowledge.” The use of such a broad definition could imply that the simplest microprocessor implementing a Proportional-Integral-Derivate (PID) controller is, in fact, intelligent because it continuously acquires knowledge (plant output data, reference inputs, the error relationship between them, etc.) and applies it (by generating control inputs to the plant). Intelligent control methodologies include the use of (biologically motivated) genetic algorithms to solve control problem (see, e.g., the work for the use of genetic algorithms for adaptive control). Genetic algorithms do not have much mathematical requirements about the optimization problem. Other intelligent control methodologies include the fuzzy control methodology. Firstly, a based on genetic algorithms (GA’s) tuning method for PID controller parameters control design is presented using shield metal arc welding (SMAW) process. Due to a large number of computer operations must carry out on GAs computing; computer system simulation on was performed to search for the optimum controller parameters. After simulation, the parameters were then brought into the actual system to carry out the welding tasks. However, because of the continuous changing of welding rod length and electric arc length, the simulated controller parameters are not fully satisfied to the changing of welding condition. Consequently the experiment showed a not so uniform welding quality which was caused by the fixed controller gain applied on the system. This part of study was arranged on the appendix A. Secondly, actuator saturation exists in almost every real control system and may give rise to the windup phenomenon. The term “windup” stems from the tendency of integral controllers to “wind up” during input saturation. If input saturation is active, the open loop dynamics become effective; these can give rise to enormous controller output signal during saturation, causing big overshoots and often badly decaying transients. Since these undesired effects of plant input saturation can be attributed to the controller, they are called “controller windup”. But also after controller windup prevention, the loop remains nonlinear, and input saturation can have a destabilize effect. If the closed loop dynamics assigned by this control are “too fast”, nonlinear overshoots, limit cycles or closed loop instability can occur. Since this problem is not related to the compensator but to the compensator but to the controlled plant dynamics, it is called “plant windup”. This effect is caused by inappropriate plant states, and is therefore referred to as “plant windup”. However, this “windup phenomenon” leads to degradation in the performance of controlled systems and results in large overshoots, long settling times, severe transient oscillations, and system instability. Therefore, when the controlled process is nonlinear, a fixed gain PID controller cannot usually give satisfactory control performance at some operating points, since the controller parameters must be adjusted following a change in operating condition. Hence, some anti-windup schemes are proposed for the stabilizing nominal control system, which with sector nonlinearity input saturation. This linear transfer function has no effect when the actuators are operating linearly, but modifies the system’s behavior during and following a saturation event to ensure stability and eventual escape from saturation, and so that the intended linear behavior is restored reasonable quickly after saturation has occurred. |
