Active control of the RTM process under uncertainty using fast algorithms

Home > Research > Feasibility Studies

Active control of the RTM process under uncertainty using fast algorithms

Host Institution: The University of Nottingham

Lead Investigator: Andrew Long, Michael Tretyakov

Co-Investigator: Marco Iglesias, Mikhail Matveev

Aims

Consistent and repeatable manufacturing is seen by the composite industry to be the key to achieving repeatable mechanical properties of composite components. For the resin transfer moulding (RTM) process, repeatability is achieved not only by the absence of incomplete mouldings, but also by repeatable cycle time or mould filling time.

The possibility of a dry spot can prevent the RTM process from being used for high-value components. Difficulties during the RTM process are often results of defects and variabilities in a dry preform. While the uncertainty in the RTM process has already been modelled to some extent, it is still impossible to employ existing predictive techniques for online monitoring and control. This project aims to improve the repeatability of the RTM process by developing novel algorithms which are fast enough to work as part of an active control system (ACS) to ensure that the mould filling process has minimum deviations from the design. The novelty of the project will include fast online estimation of the local permeability using a Bayesian inversion technique and hence more accurate process control. Via estimation of permeability, a non-destructive, cheap and quick assessment for the quality of components will be developed. The project will develop an active control system and supporting algorithms, which will aid more robust RTM processing. This will be achieved through several steps starting with developing an algorithm for an online estimation of the permeability distribution using a Bayesian inversion. The estimated permeability will be used to predict the local resin arrival times and the total mould filling time. Finally, the ACS will be integrated with these algorithms to provide a control of the mould filling.

Progress

The project has already implemented an algorithm for Bayesian inversion for estimating permeability and this is currently being integrated with commercial resin infusion software.

Data from sensors

Data from sensors

Estimated permeability

Estimated permeability

Variance of the estimated permeability

Variance of the estimated permeability

Figure 1 :Estimation of the local permeability will be performed using Bayesian inversion

Copyright © 2017 CIMComp