bmnhy 18 Tháng một 2007 - #1 Performing Nonlinear Feedback Control (Lượt xem: 1,508) Performing Nonlinear Feedback Control of a Single-Link Flexible Manipulator with LabVIEW Real-Time and PXI http://sine.ni.com/csol/cds/item/vw/p/id/285/nid/124400 Author(s): Pawel Majecki, Industrial Control Centre, University of Strathclyde, Glasgow; Alaa Shawkym, Industrial Control Centre, University of Strathclyde, Glasgow Product: Data Acquisition, LabVIEW, LabVIEW Real-Time, PXI/CompactPCI The Challenge: Creating a control system for flexible manipulators in a difficult and fast-moving environment. The Solution: Creating a control system for flexible manipulators in a difficult and fast-moving environment. A Photographic View of the Experimental Setup of the Flexible System Modelling and Controlling Flexible Manipulators for Lightweight Robotic Systems The design of flexible link robot arms is primarily motivated by the need for lightweight robot systems. Because there are numerous potential applications demanding these lightweight, robot systems, including placement of equipment in space, using standard rigid-link robot manipulators is impractical. Therefore, we are developing flexible structures to achieve increased payload capacity, reduced energy consumption, cheaper construction, faster movements, longer reach, and safe operation. The modelling and control of flexible manipulators has been a subject of a PhD research study in the Industrial Control Centre. We performed experimental validation of the flexible-manipulator model and designed control algorithms using a laboratory model of a single-link flexible manipulator. A DC servomotor, which uses LabVIEW Real-Time and a PXI controller, drives this manipulator. We then compared the results with the computer simulations. Our main objective was to validate the proposed dynamic model of a single-link flexible manipulator for use in future design and control applications, as well as to control the beam. Experimental Setup The flexible manipulator used in our experiments is an aluminium beam, measuring 1 m by 3 cm by 3 mm. We attached this beam to the shaft of an Aerotech permanent magnet DC servomotor, which applies the hub-control torque. We use the motor with the BA20 servo amplifier, which is a reliable brushless servo amplifier that we can easily adapt to drive both brush and brushless servomotors. The optical incremental encoder provided with the motor gives 4000 pulses per revolution of the motor shaft <u>on each of two quadrature (A and signals.</u> We used a quadrature clock converter to convert these signals into upclock and downclock signals. We can connect these signals directly to the counter on the PXI 6070E 12-bit 16 analog input multifunction DAQ board and process them with LabVIEW to obtain the position measurement. We measure the vibratory response with a strain gauge mounted on the beam. To measure small changes in resistance and compensate for the temperature sensitivity, we use strain gauges in a wheatstone-bridge configuration with a voltage excitation source. To improve the readings, we also have used a dedicated strain gauge amplifier specifically configured for resistive bridge measurement and in particular the strain gauges. We use the PXI 6070E to acquire the amplified signal, and we convert it to the actual tip deflection with LabVIEW. Using LabVIEW in the Modelling Phase The control system design for flexible structures is complicated because the dynamics of such systems are highly nonlinear and complex, which makes accurate modelling essential for a successful design. One challenging problem was to cause the output of the system to track a desired trajectory while maintaining internal stability of a nominally unstable system. We developed the theoretical model of the flexible link using Lagrange’s equations of motion, and we experimentally verified the model on the test bed. We used the LabVIEW transfer function analyser to obtain the frequency response of the system, which clearly indicated the resonant modes. We performed this task by applying a swept sine signal to the motor and observing the standing waves on the beam. The LabVIEW software automatically generated the magnitude and phase plots. Applying PID Control with LabVIEW Real-Time and PXI To implement the control algorithms, we originally tried to use a general-purpose PLC controller. Once we discovered the challenges of programming a complex control algorithm in ladder logic, we decided to use LabVIEW. We primarily chose LabVIEW for Windows and the Lab-PC-1200 data acquisition board because it was the only such board readily available to us. However, we soon discovered that the Lab-PC-1200 board presented significant performance limitations in terms of processing speed and real-time requirements. The theoretical minimum sampling rate of Lab PC-1200 is 10 ms, but in practice, the real limitation lies in the graphs, loops, and blocks that execute every sampling time in the VIs. Depending on their number, the practical achievable sampling rate was of order 10-50 ms, which was too low for this kind of application. Currently, we are using the NI PXI real-time controller with PXI-6070E DAQ board. This made it possible to use sampling periods of only a few milliseconds. The main advantage of this setup is real-time operating system, which guarantees the response occurs in the precisely specified time – this property is crucial to high-speed applications. Thus far, we have been successful in applying PID control of the manipulator using a built-in LabVIEW control block set, and we plan to apply a more complex nonlinear control algorithm – the state-dependent Riccati equation technique. Achieving Desired Results with LabVIEW Real-Time and PXI We tested the validity of the model by simulating the model response to a bang-bang input torque signal profile. After we applied the same torque to the actual test bed, we observed that the experimental results closely matched the predicted results of the hub position and tip displacement obtained from the model based on the assumed modes method. With the use of LabVIEW Real-Time and PXI, we achieved the performance specifications required for successful control of a difficult highly-nonlinear, fast-moving plant. Moreover, the ease of programming in LabVIEW made it possible for us to focus on the practical problems rather than wasting time on programming issues.