Process monitoring and data mining for quality control are insufficient in modern manufacturing as they lack the capability to anticipate defects before they occur. This research aims to develop a novel closed-loop quality control methodology which links defect identification with root cause analysis and corrective action for assembly systems.
It is based on the integration of in-process measurement and data mining with multi-physics variation simulation analysis through the development of simulation-driven surrogate models and closed-loop control strategy.
Duration: 5-6 years, Budget: £ 2,002,994
Industrial & Research Objectives
The overall goal is to develop, implement, demonstrate, and pilot in the industry, a systematic closed-loop quality improvement methodology with in-process adaptive monitoring capable of self-recovery from 6-sigma quality defects via:
(i) defects detection,
(ii) root cause analysis; and,
(iii) corrective actions and preventive actions (CAPA) synthesis.
The broad impact of the project is the integration of reconfigurable assembly systems with closed-loop in-process quality control. The results of the project can help to eliminate, reduce and correct defects before they occur. This will lead to increased productivity and product quality and shortened design phase.
The proposed approach comprises of three major thrusts:
(i) In-process adaptive measurement system
(ii) root cause analysis with corrective and preventive actions
(iii) multi-physics variation simulation
These three thrusts are integrated into three interlinked closed-loops:
(i) feedback-to-design in closed-loop (CL1)
(ii) real-time quality control in closed-loop (CL2)
(iii) real-time measurement control in closed-loop (CL3)
- Coverage path planning with targetted viewpoint sampling for robotic free-from surface inspection
- A novel hybrid shell element formulation (QUAD+ and TRIA+): a benchmarking and comparative study
- Spatio-temporal adaptive sampling for effective coverage measurement planning during quality inspection of free form surfaces using robotic 3D optical scanner
- 3D convolutional neural networks to estimate assembly process parameters using 3D point-clouds
- Quality and productivity driven trajectory optimisation for robotic handling of compliant sheet metal parts in multi-press stamping lines
To aid research, implementation across industry and university and reproducibility of results the research has been implemented as open-source toolboxes within Matlab and Python.
- Spatio -Temporal Adaptive Measurement (Matlab Toolbox)
- Deep Learning for Manufacturing (Python Toolbox with Tensorflow and Keras Backend)
Implementation in Measurement Stations
Figure 5: Adaptive In-Line Measurement System Using Blue Light Optical Gage
Figure 6: Adaptive In-Line Measurement System Using Blue Light Optical Gage
It is planned to develop both physical and virtual demonstrators. It is intended to demonstrate the technology in both (i) automotive industry and (ii) aerospace