33 research outputs found
Model Reference Input Shaping Using Quantitative Feedforward-Feedback Controller
Get PDFInput shaping convolutes the reference signal with a sequence of impulses, whose amplitudes and timings are designed to produce a shaped reference that avoids exciting lightly-damped modes to reduce residual vibration from a quick movement. The input shaper can be made robust to uncertain mode parameters by adding more impulses, which delays the reference signal, resulting in longer move time. Instead of using more impulses, in this paper, a feedforward-feedback control system, based on the quantitative feedback theory, is placed in the loop to match the closed-loop system, with uncertain plant, to a known reference model. The feedforward-feedback system handles the uncertainty, so the input shaper, placed outside the loop, needs not be robust. The closed-loop system emphasizes on selected frequencies and reduces the cost of feedback. It is shown that the proposed feedforward-feedback system is less conservative than the pure-feedback system. Other sources of vibration such as external disturbances and noise can be handled by the feedforward-feedback system as well. Simulation shows that the proposed technique can withstand large plant uncertainty with fast move time when compared to traditional robust input shaper
Backstepping intelligent control applied to a flexible-joint robot manipulator
No full textControl designers usually require a plant model to design a controller. The problem is the controller\u27s performance heavily depends on the accuracy of the plant model. Using adaptive control, the plant model is now allowed to have unknown parts, but the unknown parts must only appear linearly as multiplications to the known parts. Intelligent systems such as neural networks and fuzzy systems are gaining more popularity because they can be used to approximate continuous functions without knowing their mathematical structures. However, their typical offline learning is slow and limits their usefulness. Our work extends the work of other researchers who incorporate some intelligent systems into the backstepping structure to create a control system that does not require the plant model and allows online adjustment of the parameters of the intelligent systems. We extend their work in a number of interesting ways. First, we solve their problem of having to know some parts of the plant by completely using the intelligent system to learn every part of the plant. Second, we extend from using intelligent systems whose adjustable parameters appear linearly to using intelligent systems whose adjustable parameters appear nonlinearly. This leads to using more sophisticated intelligent systems such as neuro-fuzzy systems and differential neural networks. Third, we incorporate a nonlinear observer to allow the control system to be designed only from the plant\u27s output signal. Our focus is on the control system consisting of a plant in strict-feedback form, the three-layer neural network as plant estimator, a nonlinear Luenberger-type observer, and a backstepping and variable-structure controller. The control law is designed from the Lyapunov\u27s second method. We apply our control system to a two-link flexible joint robot manipulator. The flexible-joint robot manipulator, despite being a complicated and difficult-to-control system, has many practical uses. Our control system enables us to control the manipulator from only link angular position signals whereas typical control designs require link and motor angular positions, velocities, accelerations, or joint torque. We perform simulation and experimentation on an actualrobot. These results confirm the practicality and effectiveness of the control system
Track-Following Controller for Hard Disk Drive Actuator using Quantitative Feedback Theory
No full textOutput-feedback model predictive control with constraints - improve robustness using past information
No full textEnergy consumption data collection: case study on data center in a Thai University
No full textAbstract Objective Energy usage in has been increased due to the rising demand of cloud infrastructure. The government policy has been focused on building the green IT data center. The energy data need to be collected in order to monitor the energy usage. However, in an old typical data center, the building has been built with no support of such data collection. In this research, we aim to design the energy data collection system for our existing data center, a case study of data center in Thailand at the university. Based on the collected data, an energy usage monitoring system and prediction can be developed. Methods In the case study of Kasetsart University data center, the building and electric layouts were predetermined. The building layout and existing IT hardware were investigated. We designed the meter types and the number of meters to be installed for the building the energy data collection system. The corresponding database system was also designed for data logging, data visualization and analysis purpose. Results As a result, 25 installed meters along with the add-on network system were installed for logging data. A data usage example was demonstrated by building the data visualization and analysis. The presented 1 year dataset collected showed the changes of energy usages which can be used to compare with real activities happening in the campus. This encourages the integration of other related environment data such as outside temperature which may affect the electric billing cost. The dataset can be used for prediction of the electric usage; thus, the policy for reducing the electric billing cost could be established. In this paper, as a data note, we focus on the methodology of data collection required for data center
Improved Backstepping Controller for Rigid-Flexible System using Input Shaping Reference Model Matching and Neural Network
No full text
Intelligent Backstepping System to Increase Input Shaping Performance in Suppressing Residual Vibration of a Flexible-Joint Robot Manipulator
No full textMotion control of two-link flexible-joint robot with actuator nonlinearities, using neural networks and direct method
No full text
