Tue, 21 June, 2022
A knowledge-based AI platform is a computer system that uses artificial intelligence and a knowledge base to solve complex problems. Similar to the role of real-world knowledge in human intelligence and decision-making, a knowledge-based AI platform emulates human experts in decision-making and problem-solving. Throughout the years, knowledge-based AI platforms have been deployed to support many organisational processes, the most renowned of which is AI-powered customer support, including self-service systems.
The FORGE knowledge-based AI platform brings machine learning (ML) together with the codified knowledge of human experts to solve the complex and time-consuming problem of developing composite materials meeting specific performance criteria. The platform aims to reduce the number of experimentations, lowering the time and cost of new material development; this is done by aiding users in selecting material composition and manufacturing processes that achieve desired material performance characteristics. Users will utilise the platform to experiment with material compositions and examine the material performance space to optimise material properties, cost and environmental impact. To enhance and tune the platform’s performance and improve user experience, the platform will execute ML models beforehand to expand the knowledge base’s coverage. In addition to running existing ML models, the platform will allow users to retrain the models with user data and do hyperparameter tuning, resulting in user-defined models that best fit their needs.
Furthermore, the platform will employ knowledge codification to deduce new knowledge about materials from data and expert knowledge about material mixing performance. If accessible, the DSS will recommend material composition from the knowledge base. If not, the DSS will make material composition recommendations based on ML model predictions. Finally, users will screen and score material composition proposals using the multi-criteria decision analysis (MCDA) tool. Technical criteria such as material performance, manufacturability, ease of procurement; cost criteria such as initial cost, O&M cost, end-of-life cost; and environmental criteria such as CO2 emissions, particulate emissions, and energy savings, among others, will be considered by the MCDA.
The knowledge-based AI platform will comprise five integrated components in a single coherent system:
- Knowledge base to capture, formalise and process material data into knowledge and algorithms that support users in exploring the mixing performance of materials. The knowledgebase also can enhance material data based on the data and knowledge already available to the system.
- The cost module will predict the cost of new material compositions, which will involve, among other things, initial costs, operation and maintenance costs, disposal costs and end-of-life costs.
- The environmental impact module will predict the environmental impacts of new material compositions and will rely on factors including energy savings and CO2 and particulate emissions reduction.
- Machine learning models allow users to find material compositions predicted to possess desired performance criteria or expected performance of a material composition. Machine learning will also enable the knowledge-based AI platform to generate new knowledge continuously.
- A decision support system (DSS) will assist users in screening and ranking material compositions that suit their needs.
The FORGE project has received funding from the European Union's Horizon 2020 research and innovation programme under grant agreement No 958457
(This article was authored by TVS)