Lead Institution: University of Portsmouth
Industry Partners: KCC Ltd, Turbulent Designs Ltd, and Advanced Manufacturing Research Centre
Project Team: Dr Hongjie Ma (Portsmouth), Ann Swift (Portsmouth), Dr Hui Yu (Portsmouth) and Dr Ruby Hughes (AMRC)
Project Duration: 12 months (April 2018 – April 2019)
This project will seek to develop an advanced abnormal perception algorithm for cutting down the cost of the maintenance and improving the efficiency of the production process of a flexible manufacturing SME factory (KCC Ltd). KCC has worked with the UK’s largest and smallest food producers and retailers for over 35 years. They have passed the first round of migration testing, meaning they now comply with US FDA (Food and Drug Administration) Regulations and passed the UK certification for oily and alcoholic content foods. Their claim to fame is that they are leading the world with a carbon friendly and sustainable ready-meal tray, which has the potential of vast sales because of the consumer demand. With their products customers can be amongst the first to start replacing cPet and aluminium foil trays with a 100% compostable alternative.
The application of autonomous in manufacturing presents the opportunity to increase productivity, reliability, add value in a competitive arena and compensate for an ageing skilled workforce. It should be emphasised that the higher flexibility often means higher risks to reliability. Autonomy makes it a fully man-labour free factory, but it could lead to the loss of traditional advantages, such as the system failures and risks identified based on the human senses and experience of workers.
In response to these challenges, many individual Advanced Abnormal Perception(AAP) systems have been developed to control the quality of the production or ensure reliable operation of the system. Such as singular value decomposition of digital image has been used to detect the surface defects on steel strops , current signals were used to detect the broken rotor bars of induction motors and advanced self – diagnosis algorithms of the production line were used by the control system.
The shortcomings of these AAP technologies are that their versatility is not enough, which could limit the use in autonomous manufacturing. The ideal general AAP technology should be application independent, which means that It should be as follows:
 A self-learning system with the abilities of on-site unsupervised learning or semi-supervised learning. This means that the anomalies of different objects (mechanical system, hydraulic system, electrical system etc.) in the autonomous manufacturing with different physical signals (vibration, pressure, current, temperature, etc.) can be perceived with no difference.
 A measure independent system, which doesn’t interfere with production line. This can reduce the deployment cost of sensing technology, especially for legacy systems. This means that general AAP technology should take non-invasive sensors as inputs, such as camera, vibration sensor and inductive current sensor etc.