Web-based expert system for material selection of natural fiber- reinforced polymer composites

Conventional material selections are mostly based on the experience of product design engineers and the materials in common use. An inappropriate selection of materials for engineering component would result in entire product failure which ultimately has a negative impact on the society. Several algorithms, methods and spreadsheets are being proposed by various researchers in this field to improve materials selection. But, the computer oriented materials selection and knowledge-based expert systems are the robust approach in materials selection to handle huge amount of materials of choice. The decision of selecting optimised materials was complicated, as it involves diversified choice of materials, coupled with various influencing criteria for the selection. Usually more than one material satisfies the product constraints. In the exponentially growing material database, selection of optimal material for engineering design is Multi Criteria Decision Making (MCDM) problem as many properties of each material influence the selection process. In this research, first the implementation of Analytical Hierarchy Process (AHP) computational tool was explored for deciding optimum material for automotive components. The final judgement was performed with different scenarios of sensitivity analysis with prioritising the environmental factors and sustainability. The result showsM that the selected alternative materials for synthetic polymer was in compliance with the industrial Product Design Specification (PDS) and can be recommended to automotive component manufacturers to enforce green technology. Secondly, an expert system using Java programming technology with two tiers of search engine was developed to perform a fast selection of candidate materials in huge volume. The weighted-range method (WRM) was introduced to identify the range value and to scrutinize the candidate materials in the selection process. The expert system performance was tested with automotive component as a case study with high,medium and low precision criteria and the result sets generated by the expert system comply with industry benchmarks. In the third stage, hybrids of expert system with neural network technology was desired to narrow down the selection. So, the integration of Artificial Neural Network (ANN) with an Expert System for material classification was explored. The computational tool, Matlab was proposed for classification with Levenberg-Marquardt training algorithm, which provided faster rate of convergence for feed forward network. The system proved to be consistent with 93.3% classification accuracy with 15 neurons in the hidden layer. Finally, the developed expert system was deployed over the internet with central interactive interface from the server as a web-based application. As Java is platform independent and easy to be deployed in web based application and accessible through the World Wide Web (www), this expert system can be one stop application for materials selection.

Text ITMA 2015 8 D.pdf Download (1MB)

Actions (login required)

View Item