A multi-input multi-output system to implement on-line process monitoring and intelligent control of complex curvature friction stir welding was proposed. An extra rotation axis was added to the existing three translation axes to perform friction stir welding of complex curvature other than straight welding line. A clamping system was designed for locating and holding the workpieces to bear the large force involved in the process between the welding tool and workpieces. Process parameters (feed rate, spindle speed, tilt angle and plunge depth), and process conditions (parent material and curvature), were used as factors for the orthogonal array experiments to collect sensor data of force, torque and tool temperature using multiple sensors and telemetry system. Using statistic analysis of the experimental data, sensitive signal features were selected to train the feed-forward neural networks, which were used for mapping the relationships between process parameters, process conditions and sensor data. A fuzzy controller with initial input/output membership functions and fuzzy rules generated on-line from the trained neural network was applied to perceive process condition changes and make adjustment of process parameters to maintain tool/workpiece contact and energy input. Input/output scaling factors of the fuzzy controller were tuned on-line to improve output response to the amount and trend of control variable deviation from the reference value. Simulation results showed that the presented neuro-fuzzy control scheme has adaptability to process conditions such as parent material and curvature changes, and that the control variables were well regulated. The presented neuro-fuzzy control scheme can be also expected to be applied in other multi-input multi-output machining processes.
| Identifer | oai:union.ndltd.org:netd.ac.za/oai:union.ndltd.org:nmmu/vital:9612 |
| Date | January 2006 |
| Creators | Hua, Tao |
| Publisher | Nelson Mandela Metropolitan University, Faculty of Engineering, the Built Environment and Information Technology |
| Source Sets | South African National ETD Portal |
| Language | English |
| Detected Language | English |
| Type | Thesis, Doctoral, DTech |
| Format | xxi, 222 leaves, pdf |
| Rights | Nelson Mandela Metropolitan University |
Page generated in 0.0019 seconds