Every car engine, medical implant, and smartphone component rolling off a production line today must be measured for surface quality—but current inspection systems are too slow to keep up with modern manufacturing speeds. This project tackles that bottleneck by combining physics models with machine learning to create measurement systems that work in real time, without slowing production. The problem is that existing surface measurement technologies require either stopping the line or taking parts off for inspection, compromising throughput. The researcher has already shown that using prior information about what needs measuring can dramatically speed up the process. Now they will embed that approach into manufacturing by training machine learning algorithms on physics-based models of the measurement process itself. If successful, the result would be quality control that runs continuously, adapts to changing conditions, and improves over time—without sacrificing speed or accuracy. This is not a small tweak to existing tools; it is a fundamental rethinking of how measurement integrates with production, potentially transforming digital quality control across industries that manufacture precision components.
View original technical description
The world is experiencing the first stages of a digital industrial revolution: Industry 4.0. However, current digital quality control solutions are not delivering in terms of speed, capability, efficiency or futureproofing. An essential part of manufacturing is quality control, which is achieved through measurement. One of the most important measurands for quality control is the surface of the part; both shape and fine-scale topography are critical when considering tolerances, assembly and ultimately functionality. But current integrated surface measurement technologies are too slow and have little flexibility under variable processing conditions. Measurements are taken after manufacture or by slowing down the process - compromising the all-important throughput. To take surface measurement from lab to application can require speed increases of several orders of magnitude, and this is often beyond the capability of current technology. However, I have demonstrated that these challenges can be tackled using an emerging approach: information-rich metrology - the use of a priori information to enhance the measurement process by optimising what needs to be measured, so increasing the spatial bandwidth but decreasing the measurement time. Such optimisation generally requires complex physics models of the measurement; this is where a recent revolution comes to the rescue: machine learning, which I will use to combine newly developed physics models with a priori information to produce enhanced measurement systems that are an integral, real-time, and constantly learning part of the manufacturing process. This is not a proposal to make incremental developments; rather I seek to transform the field by combining the advances of three fields (basic physics, machine learning and metrology) - a binding energy approach that will be more than the sum of the parts. The proposed project will revolutionise digital quality, making measurement a seamless, yet constantly evolving part of manufacturing.
Plain English summaries and category classifications on this site are generated by AI and may not perfectly reflect the original research.
Is something wrong? Let us know