Dissertations / Theses on the topic 'On-line tuning PID controller'

This article introduces a versatile control board design that can be used in many drying systems using direct heat transfer solutions in combination with static tray distribution based on development of a flexible self-tuning PID controller. The product is applied for a small oven drying solid waste samples or plant materials for further analysis of some parameters. The control board is built based on the Arduino embedded system using a flexible soft PID (Proportional–Integral–Derivative) controller that can automatically change its gains according to the required temperature thresholds to best meet the setpoint of temperatures. The system has a small steady-state error (SSE), fast response to the setpoints and keep stable with temperature deviation when reaching the required threshold around ± 0.5 0C. In addition, the controller board can operate in a variety of modes, including direct temperature operation, pre-set operation, and switching mode.<br>Bài báo này giới thiệu một thiết kế mạch điều khiển đa năng có thể áp dụng trong nhiều hệ thống sấy sử dụng các giải pháp truyền nhiệt trực tiếp kết hợp phân phối khí kiểu khay tĩnh trên cơ sở phát triển bộ điều khiển PID mềm tự chỉnh linh hoạt. Sản phẩm được ứng dụng cho một mô hình lò nhỏ sấy mẫu chất thải rắn hoặc mẫu thực vật phục vụ các nghiên cứu phân tích thành phần một số chất. Mạch điều khiển được chế tạo trên nền tảng hệ thống nhúng Arduino sử dụng bộ điều khiển PID mềm linh hoạt, có khả năng tự động thay đổi tham số theo ngưỡng nhiệt yêu cầu để đáp ứng nhiệt độ tốt nhất so với ngưỡng nhiệt độ đặt. Hệ thống có độ quá điều chỉnh nhỏ, nhanh đáp ứng tới các ngưỡng đặt và giữ ổn định với sai lệch nhiệt độ khi đạt ngưỡng yêu cầu trong khoảng ±10C. Ngoài ra, bộ điều khiển còn có thể hoạt động theo nhiều chế độ khác nhau, bao gồm hoạt động theo chế độ đặt nhiệt độ trực tiếp, hoạt động theo chu trình đặt trước, và chuyển chế độ hoàn toàn tự động.

An off-line method of tuning PID controllers used in process control applications is presented. The plant under consideration is of second order with delay and the loop is presumed to be operating at steady state under the influence of random disturbances. Tuning is performed by minimizing the cost function, namely the integral of squared error, with respect to the tuning parameters. To do this, a second order approximation of the cost function is obtained and minimized using the Newton-Raphson method. System identification is done without interrupting production by introducing test signals of low signal to noise ratio and using a Kalman filter to extract the response. The result is a model composed of Laguerre functions which emulates the frequency response of the system in question. The integral term is found to cause stability problems if left unconstrained because the lag it introduces is not beneficial in the case under consideration. On the other hand, the algorithm behaves well when tuning a PD controller if the Kalman filter yields accurate results because the model obtained is then very similar to the transfer function of interest. The Kalman filter presently used does not adequately model the noise, so its performance hampers proper tuning of the loop.

This thesis attempts to tune any controller without the mathematical model knowledge of the system it is controlling. For that purpose, the optimization algorithm of MATLAB&reg<br>6.5 / Nonlinear Control Design Blockset (NCD) is adapted for real-time executions and combined with a hardware-in-the-loop simulation provided by MATLAB&reg<br>6.5 / Real-Time Windows Target (RTWT). A noise-included model of a DC motor position control system is obtained in MATLAB&reg<br>/ SIMULINK first and simulated to test the modified algorithm in some aspects. Then the presented methodology is verified using the physical plant (DC motor position control system) where tuning algorithm is driven mainly by the real system data and the required performance parameters specified by a user defined constraint window are successfully satisfied. Resultant improvements on the step response behavior of DC motor position control system are shown for two case studies.

PID controllers are widely used in industrial control systems since it could significantly improve the performance of the systems. It is a basic and important tool for an engineering student.An educational PID controller model using LEGO Mindstorms NXT kit as a line following robot was presented in this thesis. This robot could be controlled by a PC via Bluetooth. MATLAB was used for analyzing data and controlling the robot. A MATLAB Toolbox named “RWTH—Mindstorms NXT Toolbox” was utilized for remote controlling and gathering the feedback data from the robot in real-time via Bluetooth.The algorithm and tuning methods of PID controllers were studied in this thesis. Moreover, the effectiveness of each parameter would be studied by comparing the result of P, PI and PID control type.

Control of Switched Mode Power Supplies (SMPS) has been traditionally achieved through analog means with dedicated integrated circuits (ICs). However, as power systems are becoming increasingly complex, the classical concept of control has gradually evolved into the more general problem of power management, demanding functionalities that are hardly achievable in analog controllers. The high flexibility offered by digital controllers and their capability to implement sophisticated control strategies, together with the programmability of controller parameters, make digital control very attractive as an option for improving the features of dcdc converters. On the other side, digital controllers find their major weak point in the achievable dynamic performances of the closed loop system. Indeed, analogto-digital conversion times, computational delays and sampling-related delays strongly limit the small signal closed loop bandwidth of a digitally controlled SMPS. Quantization effects set other severe constraints not known to analog solutions. For these reasons, intensive scientific research activity is addressing the problem of making digital compensator stronger competitors against their analog counterparts in terms of achievable performances. In a wide range of applications, dcdc converters with high efficiency over the whole range of their load values are required. Integrated digital controllers for Switching Mode Power Supplies are gaining growing interest, since it has been shown the feasibility of digital controller ICs specifically developed for high frequency switching converters. One very interesting potential benefit is the use of autotuning of controller parameters (on-line controllers), so that the dynamic response can be set at the software level, independently of output capacitor filters, component variations and ageing. These kind of algorithms are able to identify the output filter configuration (system identification) and then automatically compute the best compensator gains to adjust system margins and bandwidth. In order to be an interesting solution, however, the self-tuning should satisfy two important requirements: it should not heavily affect converter operation under nominal condition and it should be based on a simple and robust algorithm whose complexity does not require a significant increase of the silicon area of the IC controller. The first issue is avoided performing the system identification (SI) with the system open loop configuration, where perturbations can be induced in the system before the start up. Much more challenging is to satisfy this requirement during steady state operations, where perturbations on the output voltage are limited by the regular operations of the converter. The main advantage of steady state SI methods, is the detection of possible non-idealities occurring during the converter operations. In this way, the system dynamics can be consequently adjusted with the compensator parameters tuning. The resource saving issue, requires the development of äd-hocßelf-tuning techniques specifically tailored for integrated digitally controlled converters. Considering the flexibility of digital control, self-tuning algorithms can be studied and easily integrated at hardware level into closed loop SMPS reducing development time and R & D costs. The work of this dissertation finds its origin in this context. Smart power management is accomplished by tuning the controller parameters accordingly to the identified converter configuration. Themain difficult for self-tuning techniques is the identification of the converter output filter configuration. Two novel system identification techniques have been validated in this dissertation. The open loop SI method is based on the system step response, while dithering amplification effects are exploited for the steady state SI method. The open loop method can be used as autotunig approach during or before the system start up, a step evolving reference voltage has been used as system perturbation and to obtain the output filter information with the Power Spectral Density (PSD) computation of the system step response. The use of ¢§ modulator is largely increasing in digital control feedback. During the steady state, the finite resolution introduces quantization effects on the signal path causing low frequency contributes of the digital control word. Through oversampling-dithering capabilities of ¢§ modulators, resolution improvements are obtained. The presented steady state identification techniques demonstrates that, amplifying the dithering effects on the signal path, the output filter information can be obtained on the digital side by processing with the PSD computation the perturbed output voltage. The amount of noise added on the output voltage does not affect the converter operations, mathematical considerations have been addressed and then justified both with a Matlab/Simulink fixed-point and a FPGA-based closed loop system. The load output filter identification of both algorithms, refer to the frequency domain. When the respective perturbations occurs, the system response is observed on the digital side and processed with the PSD computation. The extracted parameters are the resonant frequency ans the possible ESR (Effective Series Resistance) contributes,which can be detected as maximumin the PSD output. The SI methods have been validated for different configurations of buck converters on a fixed-point closed loop model, however, they can be easily applied to further converter configurations. The steady state method has been successfully integrated into a FPGA-based prototype for digitally controlled buck converters, that integrates a PSD computer needed for the load parameters identification. At this purpose, a novel VHDL-coded full-scalable hybrid processor for Constant Geometry FFT (CG-FFT) computation has been designed and integrated into the PSD computation system. The processor is based on a variation of the conventional algorithm used for FFT, which is the Constant-Geometry FFT (CG-FFT).Hybrid CORDIC-LUT scalable architectures, has been introduced as alternative approach for the twiddle factors (phase factors) computation needed during the FFT algorithms execution. The shared core architecture uses a single phase rotator to satisfy all TF requests. It can achieve improved logic saving by trading off with computational speed. The pipelined architecture is composed of a number of stages equal to the number of PEs and achieves the highest possible throughput, at the expense of more hardware usage.

碩士<br>東海大學<br>化學工程研究所<br>81<br>Three kinds of closed-loop PID tuning techniques including time domain, frequency domain, and cascade control approaches are proposed in this study. Under proportional control, a single dynamic test ( such as step or pulse change ) in the set point is introduced to the system. An identification technique for the underdamped second-order-plus-dead-time model from few data points of the step response curve is developed. The ultimate frequency and gain can then be ob- tained based on the model from the literature. The ultimate frequency and gain can also be obtained from the pulse (or step) transient by the Fourier integral transform without any parametric model. Thus, PID controller settings can be found from the ultimate frequency and gain by the Ziegler- Nichols rules. On the other hand, a procedure for tuning simultaneously the primary and secondary controllers of cascade control systems based on a single dynamic testing using the Ziegler-Nchols rules is also proposed. Simulation results have demonstrated that the proposed techniques can provide good values of PID controller settings despite unanticipated process noise or load disturbances and is insensitivity to the used control- ler gains.

碩士<br>國立東華大學<br>電機工程學系<br>99<br>The main focus of this thesis is to design the intelligent controller, and discuss the system tracking control problem with disturbance influence. This thesis presents a Particle Swarm Optimization (PSO) approach for online tuning Proportional Integral Derivative (PID) controller. There is a great deal of general literature on PID controller with ignoring the disturbance interference. However, in practical applications, the ability of interference avoidance of PID controller should be considered in evaluating the performance of system response. In this research, using PSO online tuning mechanism, PID controller can track the input signal effectively even though the setpoint changes and load disturbance influences. The simulation results describe that the proposed method is able to obtain the good performance and stability.

碩士<br>中原大學<br>化學工程研究所<br>90<br>The inherent time-varying nonlinearity and complexity usually exist in chemical processes. The design process would be significantly deviated from the normal operating condition when only a slight disturbance occurs. Accordingly, the design of control structure should be properly adapted based on the instantaneous state. In this research, an improved conventional PID control scheme using linearization through a specified neural network is developed to control nonlinear processes. An input-driven output neural network ARX model trained by Levenberg-Marquardt algorithm is introduced in the model design. The linearization of the neural network model will be proposed to extract the linear model for updating the controller parameters. In the scheme of the optimal tuning PID controller, the concept of general minimum variance is presented and a constrained criterion is also considered. Like gain scheduling, the control system of the proposed method at each time interval is chosen from a set of predefined linearized dynamic model. Unlike gain scheduling control, the control parameters based on the neural network model have an infinite scheduling resolution when using a neural network for updating parameters. To apply the proposed method to most of the practical application problems, several variations of the proposed method, including the momentum filter, the updating criterion and the step size of the control action, are presented to provide significant improvement and make the proposed algorithm more practical. The proposed method has two advantages. First, less computation of linear adaptive control scheme is used. Second, the nonlinear characteristics of neural networks can be incorporated into the control design. To demonstrate the potential applications of the proposed strategies, several problems, including batch reactors and pH neutralization processes, are applied.

碩士<br>中原大學<br>電機工程研究所<br>95<br>It’s a very important and worthy to discuss how to use active yaw control to position the maximum power point with a horizontal axis wind turbine which usually needs wind direction sensor to position. The mountain climbing method based on measurement of the voltage and current from wind-turbine generator system (WTGS) to calculate the power for the maximum power point without any wind direction sensors is proposed. It’s verified by computer simulations and experiments. An on-line parameter tuning PID controller for DC motor to position the wind turbine is also presented. The on-line parameter tuning policy is based on the Simultaneous Perturbation Stochastic Approximation (SPSA) technique. Three PID parameters can tuned consecutively until the optimal values are attained. The proposed approach has been confirmed by computer simulations and experiments. The PID controller is also combined with the mountain climbing method to increase system’s stability and locate the maximum power point effectively. Finally, the FPGA is used to realize the proposed method.

碩士<br>國立高雄第一科技大學<br>電機工程研究所碩士在職專班<br>105<br>Recently, in the field of Virtual Reality, the development of visual and audio are very mature so that the tactile sensation becomes to the trend of recently studt. The technology is getting more and more developed. The research about network technology for tactile sensation interaction is improve every day, such as Da Vinci Si Surgical System and Robot. In this thesis, the“transient response specification formula”are employed to obtain the transfer function of Novint Falcon. These approaches based on the inertia weight, linear decreasing weight method and random inertia weight method to find the best tuning rate. In the simulation case, the linearly decreasing weights of PSO algorithm is the best method. The parameters of PID controller of haptics devices are turned on-line by the PSO algorithm. The purpose is to optimize control. Which makes the force feedback of haptics devices more realistic. Finally, We use the network to remote control between the Novint Falcon tactile sensing device and the Geomagic Touch touch-sensing device. The experimental results show that the trajectory of the Novint Falcon haptic device and the trajectory of the Geomagic Touch haptic device are almost fitted. By the proposed approach, we can get good performance for the remote control.

碩士<br>國立中央大學<br>電機工程學系<br>86<br>Abstract This paper presents some ideas related to the translation between the fuzzy controller and the quasi- classical controller. The key point is to obtain the quasi- classical PID for a given Fuzzy-PID type controller. From this idea, we focus the attention on the acquirement of a useful tool to get the approximately classical PID. This method can translate all the fuzzy controller to the quasi-classical controller. In order to improve further the performance of the transient state and the steady state of the PID type controller, we develop a method to tune the scaling factors of the PID type fuzzy controller on line. This method has two type tuners, one is the function tuner and another is the fuzzy tuner. The simulations of the PID type fuzzy controller with the self-tuning scaling factors shows the better performance in the transient and steady state response.

碩士<br>長庚大學<br>電機工程研究所<br>92<br>This thesis provides the design, analysis and investigation on the DSP-Based controller for the electric vehicles. This system consists of a digital control board, brushless DC motor, battery charger and the battery stack. Instead of using traditional PI control, a new Self-tuning PID control algorithm is proposed. The developed controller has rapid closed-loop response for load disturbance. The algorithm is easy to be implemented for relegation. This will improve the performance of the electric vehicle. According to some conferred reference, two-step charging method with C-V and C-C is widely used. The thesis is based on the method and we use DSP-2407 to be the main controller. In order to avoid low-voltage and over-current, it’s necessary to design the protective circuit for enhancing the reliability of the charging system. The function let the power system of electric vehicle more complete.

碩士<br>建國科技大學<br>電機工程所<br>94<br>The objective of this thesis is to carry out a Grey-prediction controller on a Programmable System-on-Chip (PSoC). To achieve this goal, two major issues will be discussed and put into practice. First, one stand-alone controller is designed and completed by using a PSoC system-on-chip. Second, the Grey-prediction controller is implanted inside Programmable System-on-Chip (PSoC). With the help of this novel embedded SoC technology, controller is compact in size, circuit layout is simplified, control scheme can be modified without hardware adjustment, and cost is also reduced. In order to establish two-way communication between computer and PSoC, a vivid control panel is written by LabVIEW graphical language. One temperature regulation control system as well as one motor speed control system are arranged and tested to confirm the validation of Grey-prediction controller scheme. From the data collected and studied, the proposed Grey-prediction controller outplays the traditional PID controller in all aspects including the transient response, steady state response and tracking control.

碩士<br>國立高雄第一科技大學<br>系統資訊與控制研究所<br>98<br>In this study, we used Particle Swarm Optimization (PSO) to tune the parameters of PID controller. Some error-criteria are the fitness function, the performance index, are used to judge the performances. The performance index have three functions: (1) The integral square error criterion (ISE) (2) The integral absolute error criterion (IAE) (3) The integral time absolute error criterion (ITAE). Otherwise, the “open loop testing method” and “trial and error method” are employed to obtain the transfer function of Stewart platform. In order to take optimum parameters of PID controller, the PSO algorithm be used and results are compared with traditional Z-N tuning rule. Finally, the experimental results verify the PID parameters based on PSO algorithm not only keep system stable, but also decrease overshoot.

碩士<br>建國科技大學<br>自動化工程系暨機電光系統研究所<br>104<br>The design of band-sawing machine acts as a front-end industry in the industry chain. Fuzzy self-adaptive should control the accurate requirement of PID control theory and make the cutting down material affect the following bench lathe and milling machine’s process. The designs include from the small machine to the large machine, the accuracy is the first requirement in machinery industry. In order to fit the requirement of modern industry, such as: fast sawing, accuracy standard…etc. Many adjustments depend on tools to adjust in order to achieve a standardized production. This study seeks to apply fuzzy self-adaptive to apply PID control theory in band-sawing design. The cutting-down material is more accurate. The experiment proves its feasibility. In addition, the industry competition of band-sawing machine can be promoted accurately. Keywords：band sawing machine, fuzzy adaptive tuning PID control theory, the material accuracy, precision standard

碩士<br>國立臺北科技大學<br>能源與冷凍空調工程系碩士班<br>99<br>PID controller is the widespread controller in industrial applications; however, the loop tuning needs experience of operator or factory settings for control and there is no optimum adjusting control parameter based on the in-situ operating environment on real time. There are many applications of Self-Tuning Regulator (STR) in engineering but is less in HVAC. In this study, utilizing STR to develop the controller with PID parameter regulation for the outlet temperature of chilled water. Accordingly, this research I proposed and implemented a forward-looking and advanced controller design ideas for the design of control, which is the essence and the original of controller that is how to integrate "wisdom, logic,and experience accumulated to form an intuitive thinking" into embedded controller. This thesis proposed a new regulation method of self-tuning PID parameter control in real time to create the parameters curves and rule through Recursive Least-Squares Estimation, Linear Regression Method, and Minimum Variance Controllers in order to self-tuning and tracing on system identification. In spite of the variation of operating environment, the Self-Tuning PID Strategy and Controller will adjust the optimum parameters on real time and optimization even though the system experiencing changing or failed in system identification.

This project tries to explore the potential of using soft computing methodologies in controllers and their advantages over conventional methods. PID controller, being the most widely used controller in industrial applications, needs efficient methods to control the different parameters of the plant. This thesis asserts that the conventional approach of PID tuning is not very efficient due to the presence of non-linearity in the system of the plant. The output of the conventional PID system has a quite high overshoot and settling time. The main focus of this project is to apply two specific soft-computing techniques viz. fuzzy logic and genetic algorithm to design and tuning of PID controller to get an output with better dynamic and static performance. The application of fuzzy logic to the PID controller imparts it the ability of tuning itself automatically in an on-line process while the application of genetic algorithm to the PID controller makes it give an optimum output by searching for the best set of solutions for the PID parameters. The project also discusses the benefits and the short-comings of both the methods. The simulation outputs are the MATLAB results obtained for a step input to a third-order plant.

The PID controller is regarded as a dependable and reliable controller for process industry systems. Many researchers have devoted time and attention to PID controller tuning and they all agree that PID controllers are very important for control systems. A PID equation is very sensitive; its parameters must always be varied following the specific application to increase performance, such as by increasing the system’s responsiveness. PID controllers still have many problems despite their importance for control systems in industries. The problem of big overshoot on the conventional gain tuning is one of the serious problems. Researchers use the PSO algorithm to try and overcome those problems. The tuning of the MIMO PID controller based on the PSO algorithm shows many disadvantages such as high-quality control with a short settle time, steady-state error, and periodical step response. The traditional PSO algorithm is very sensitive and it sometimes affects the quality of good PID controller tuning. This research has proposed a new equation for improving the PSO algorithm. The proposed algorithm is the combination of linearly decreasing inertia weight and chaotic inertia weight, after which a control factor was introduced as an exponential factor. This was very useful for simulations as it is adjustable. The Matlab simulation results of the experiments show that the simulations as it is adjustable. The Matlab simulation results of the experiments show that the new proposed equation converges faster and it gives the best fitness compared to linear inertia weight and oscillating inertia weight and other old equations. The MIMO PID controller system that consists of four plants was tuned based on the new proposed equation for the PSO algorithm (LCPSO). The optimized results show the best rise time, settling time, time delays, and steady-state compared to the systems that are tuned using the old equations. The exploration was directed at considering the impact of using the PSO calculation as an instrument for MIMO PID tuning. The results obtained in the examination reveal that the PSO tuning output improved reactions and can be applied to various system models in the measure control industry. The results for the MIMO PID controller tuned using PSO were assessed using integral square error (ISE), integral absolute error (IAE), and the integral of time expanded by absolute error (ITAE). The five well-known benchmark functions were also used to endorse the feasibility of the improved PSO and excellent results in terms of convergence and best fitness were attained.<br>Electrical and Mining Engineering<br>M. Tech. (Electrical Engineering)

碩士<br>中原大學<br>電機工程研究所<br>93<br>Abstract Conventional PI controllers were often in widespread use owing to their capability of efficiently improvement of steady-state responses of systems; however, transient responses are often their defect. Based on the advantages of PI controllers, the thesis proposed that in the case that a PI controller is served as a host controller, a on-line tuning fuzzy controller can be used as a servant controller to improve the transient response to get better performance of the whole system. In the past, the rule base of a fuzzy controller was mostly created by professional experience or trial-and-error. This was not only wasting a lot of time but also requiring redesign for different plants so that the fuzzy controller is not wide used in industry. Therefore, in the present study, an appropriate performance index was chosen first. Then a “Simplified Gradient Method” was used to decide integral correction of the active rules to get decreasing values of the performance index. The Simplified Gradient Method can properly tune the rule bases of the fuzzy controller for different control systems, and the system performances are improved effectively. At last, by computer simulation, we compared the responses of several systems controlled by PI controller along with those accompined by on-line tuning fuzzy controller. The simulation results reveal that the on-line tunung fuzzy controller can distinctly improve (decrease) both the rise time and maximum overshoot.

碩士<br>國立成功大學<br>化學工程研究所<br>81<br>A new autotuning method based on an identification procedure under P,PI or PID feedback has been developed to tune PID controllers automatically.The process is identified as a second- order plus dead-time model,based on the peak response data extracted from a single closed-loop test. Simple algorithms for tuning P,PI,and PID controllers are proposed to meet desired specifications on dominant pole placement. Several controller-setting techniques are compared for the PI and PID control of a second-order plus dead time model under set-point and load changes. The comparison is based on ISE,IAE, ITAE,and stability robustness. Our method yields satisfactory results. The method can be applied when initial PID settings are required or the settings need to be adapted to changing conditions . A simulation study demostrates the validity and self-tuning capacity of the method for a wide variety of process dynamics.

碩士<br>國立高雄應用科技大學<br>電機工程系碩士班<br>93<br>Nowadays, most process control in industries use P.I.D controller. The reasons that the PID controllers are widely used, are its relatively simple structure, easy to design and implement in practice. However, since tuning of any parameter in the PID control system will inevidently affect the other ones, which may result a poor performance, it is really a challenge work to find an effective method for tuning parameters; As to single input single output system, no matter it is stable, unstable, or any system with time-delay, conventional PID controller parameter designed method can get very good set point response, as well as having good rejection to the load disturbance. In different process, having these PID tuning formula, and knowing some information of the process, we can easily design the PID parameters for meeting the system’s performance criterions. The surveys of existing literatures for the design of PID controller and parameter tuning methods in this thesis, are focused on the typical single input single output systems, inner feedback loop, internal model control (IMC) and Smith predictor. We find all the proposed parameters design methods may guarantee the good response and the load disturbance rejection. However, in view of robust stability, each method has its limitation in applying to a specified system. Through exhaustive analysis and integration, this thesis makes a systematic comparison of parameter-tuning method for those existing PID control systems, and provides some suggestions which in turn will benefit those engineers who are working in the industries or in the laboratories.

碩士<br>國立交通大學<br>電機與控制工程系<br>88<br>In the thesis, we present a PID tuning method using the Fuzzy Neural Network (FNN) based on system parameters. PID tuning methods were widely used for stable processes, or some over-damped unstable processes. However, PID controller for under-damped unstable processes and higher order unstable processes is less common. An FNN approach is proposed to identify the relationship between system parameters and the PID controller parameters that meets the performance index. Then, the FNN is used to automatically tune the PID controller parameters for different system parameters so that neither numerical methods nor graphical methods need be used. Even though for some of the heavily oscillatory processes, the FNN still can find a suitable PID controller parameters. Simulation results show that the FNN can achieve the specified values efficiently.

碩士<br>國立臺灣科技大學<br>電機工程系<br>91<br>Abstract The purpose of this thesis is to investigate the application of the self-organizing fuzzy controller (SOFC) in the parameter tuning of speed and flux controller under indirect rotor-flux-oriented induction motor driver. In this paper, the control rule bank of SOFC will be established and updated by online learning algorithm from zero initial fuzzy rules. This approach has a learning ability for responding to the nonlinear time-varying characteristics of induction motor so that it can enhance the adaptability and robustness of the induction motor driver. In addition, the design of scaling factor addressed in this thesis for speed and flux controller can reduce time-consuming work for general fuzzy controllers. In the experimental setup, the control program is developed on the PC-Based motor controller development system, and tested on a TECO 1-hp induction motor supplied by a RHYMEBUS RM5 inverter. The performance of SOFC is compared with the conventional PI controller, the basic fuzzy controller, and the fuzzy model reference learning controller through many experiments. It shows that the proposed controller attain better performance than other controllers’. Also, the experimental results verify the feasibility of our control strategy.

碩士<br>國立臺灣科技大學<br>自動化及控制研究所<br>95<br>This research is based on the self-tuning PID control method for three PID control gains adjustment. In this study, the self-tuning PID control method is classified into two types as local tuning and global tuning. The first one is through simulation to discuss the problem of using the original self-tuning neural PID approach. Due to the use of sigmoid function, the modification range of parameters is only local. In view of this drawback, we proposed a novel self-tuning method for PID control gains adjustment which can have global modification for those parameters. This approach utilizes the gradient optimum concept property and sliding mode stable idea to design the adaptive laws. This approach is different from the general way of designing the adaptive laws which is usually obtained by using the concept of the Lyapunov analysis method. Furthermore, an inverted pendulum system and an uncertain chaotic system are employed to demonstrate the control performance of the proposed approach. Good efficiency is shown in our simulation results.

碩士<br>國立臺灣大學<br>化學工程學研究所<br>90<br>The ultimate goal of the research is to construct a portable PID auto-tuner. The auto-tuner is developed on the software platform named ScadACS. From the threshold, we hope to develop a methodology suitable for on-line auto-tuning. And we use the relay feedback test, which is sufficient to apply the dynamic information for our auto-tuning procedure. The auto-tuning procedure is divided into two parts: system identification and controller tuning. The purpose of model identification is to choose the PID controller tuning rules. Theoretically, the system dynamics can be treated roughly as two classes. One of them includes the FOPDT systems and SOPDT systems with damping factor bigger than 0.7, the other includes the SOPDT systems with damping factor less than 0.7. By a normalized system's relay feedback response curve, we treat the first class as FOPDT dynamics systems and the second class as underdamped SOPDT systems. Each system has its corresponding tuning rules based on parameters obtained from relay feedback test. Finally, by the method we proposed, we can develop a portable PID auto-tuning module on the software platform ScadACS.

碩士<br>國立中央大學<br>電機工程研究所<br>83<br>The most successful domains have been proved in the fuzzy control of various applications such as home appliances c- ontrol, automobiles safe control, industry control, and so on. Even considerable interest has been attracted in the weapon control systems. In order to make the fuzzy controller achieve the prospec- tive target, we have to adjust the parameter of the fuzzy controller. These parameters are membership functions, sc- aling facors and control rules. Among which only the scal- ing factor adjustment can achieve the requisite of the re- al time control. Therefore, if we want to apply the fuzzy control, the scaling factors of fuzzy controller which can be self-adjusted is a necessary requirement. In this thesis, we propose an on-line self-tuning scaling factor fuzzy controller to improve the tradtional PI-type and PD-type fuzzy controller. The result of experiments s- hows the advantage and necessity of the self-tuner. To apply the fuzzy control, many electron firms invest mu- ch money to study "Fuzzy Chip" . And it make the price of "Fuzzy Chip" expensive. There are also many problems in t- he design of "Fuzzy Chip" and they need to be overcome. If we can use "micro single chip computer" such as PIC or In- tel 8051 to be a fuzzy controller, and it can achieve the control target we demand. Then it will replace "Fuzzy chip "and become a cheap, useful and flexible fuzzy controller.

碩士<br>國立交通大學<br>控制工程系<br>84<br>In this thesis, we propose a new PID tuning method using fuzzy nrural network based on gain and phase margin specifications (FNGP).We use the fuzzy neural networks to determine the parameters of PID controllers. Because of the complexity of the basic definitons of gain and phase margin equations, an analytical design method to achieve the specified gain and phase margins is not available to date. First, a fuzzy neural modeling method to identify the relationship between the specifications and the PID controllers parameters is proposed. Then, the FNGP is used to automatically tune the PID controllers parameters with different gain and phase margin specifications so that neither numerical methods nor graphical methods have to be used. This makes iteasy to tune the controller parameters to have the specified robustness and performance. Simulation results show that the FNGP can achieve the specifications much better than other methods.

碩士<br>國立成功大學<br>航空太空工程學系碩博士班<br>97<br>For static friction identification, from Hsieh and Pan[1], there is a nonlinear spring in the static friction phase. So, use the equation that is represented by Hsieh and Pan[1] to fit the nonlinear spring curve via binary method. For on-line identification, Use estimated output and actual output to get error, and use gradient search method with weighting and gradient search method without weighting to improve the motor parameters. If the motor parameters change or the intial motor parameters are wrong, we will find and use on-line identification to decrease the error between the actual parameters and the estimated parameters. With simulation and experiments, verify the above method.

碩士<br>國立交通大學<br>電機與控制工程系<br>87<br>In the thesis, we present a PID tuning method for unstable processes using Fuzzy Neural Network based on gain and phase margin (FNGP) specifications. PID tuning methods were widely used to control stable processes. However, PID control for unstable processes is less common. A fuzzy neural network approach is proposed to identify the relationship between the gain-phase margin specifications and the PID controller parameters. Then, the FNN is used to automatically tune the PID controller parameters for different gain and phase margin specifications so that neither numerical methods nor graphical methods need be used. Even though for some of the unreasonable specifications, the FNN still can find a suitable PID controllers' parameters close to the specifications. Simulation results show that the FNN can achieve the specified values efficiently.

碩士<br>國立交通大學<br>電控工程研究所<br>104<br>In view of the widespread use of PID controllers in servo motors, this study first collects some widely used PID tuning methods and compares their performance. These methods use simple formulas to find the PID gains, but they cannot satisfy the requirements of robustness and performance at the same time. Some of the tuning methods tune the PID gains without explicit consideration of the robustness margin. To improve the tuning process, this study proposes a new tuning method which considers the performance of time domain and frequency domain at the same time. While minimizing the integral of squared errors, the proposed method also imposes frequency domain constraints. Using the frequency shaping techniques, the proposed method can satisfy the requirement of the user specified phase margin and maximize the bandwidth of the system. Therefore both robustness and performance are explicitly taken into account. Graphic user interfaces are designed to facilitate the whole tuning process of the proposed method. Then experiments are conducted in the position and velocity loops of a servo motor from Faulhaber. The results show that a better performance is achieved in comparison with the Faulhaber’s auto-tuning algorithm.

碩士<br>中原大學<br>化學工程研究所<br>91<br>Abstract Controlled processes in nearly all-chemical industries frequently encounter with inherently more than one variable to be controlled. They are known as multivariable or multi-input multi-output (MIMO) processes. The control of multivariable systems is not always an easy task due to its complex and interactive nature. The goal of this paper is to identify and control MIMO processes by means of the dynamic partial least squares (PLS) model, which consists of a memoryless PLS model connected in series with linear dynamic models. Unlike traditional decoupling MIMO processes, the dynamic PLS model can decompose the MIMO process into a multiloop control system in a reduced subspace. Without the decoupler design, the optimal tuning multiloop PID controller based on the concept of general minimum variance and the constrained criteria can be directly and separately applied to each control loop under the proposed PLS modeling structure. As for the nonlinear MIMO processes, the dynamic PLS (DynPLS) model, which are obtained by the instantaneous linearized neural network model at each sampling time, can still be used to decompose the MIMO process into a multiloop control system in a reduced subspace. The proposed algorithm has the following advantages: (i) It is easy to identify DynPLS since it is not necessary to identify the MIMO system by a sequence of relay identification. (ii) The coupling effect in the MIMO system is now overcome effectively. The PLS structure can be decomposed into each pair of input and output and it selects the number of control loops based on the variation captured by each pair. (iii) Unlike the sequential tuning of the multiple control loop for the iterative design in each control loop, the adaptive tuning PID controller strategy in the SISO system can be directly and simultaneously applied to each loop of the multiloop control design in the MIMO system under the decomposed structure of PLS. This simplicity and feasibility of the scheme can easily be extended to any multiloop control strategies. Simulation case studies are provided to demonstrate the effectiveness of the control design procedures of the MIMO process.