Skip to main content

REPAIR POLICY DETERMINATION BASED ON AI LEARNING METHODS - REPAY

Objective

The goal of REPAY is to create a system for the determination of the best repair policy of large and complex equipment.
A project was set up to create a system for the determination of the best repair of large and complex equipment, including the development of a decision support tool for policy determination in repairs, based on existing information technology (IT) solutions. 2 demonstrators were proposed: a smaller one for the repair of diesel engines, and a wider one for the repair of large and complex equipment. The software development method used for this project was rapid prototype development. This ensured that the industrial end users would be intimately involved at every stage and that a more useful product would result. The early prototypes were built and evaluated, resulting in a specification refinement. The final prototypes were then built, integrated and tested in continual cycle of development. A software system has been developed that assists estimators at repair yards to analyze enquiry documents and respond with tender documents. As part of this procedure, the software also assists with the construction of a budget for the repair job. The artificial intelligence components of the system can ensure that all steps of a repair process are taken into account in the budget, exclusion remarks are added to the tender as required, and advice is given to the estimator when necessary. Other modules include a workload planning module to assess the impact of repair jobs in terms of resource utilization at the yard, a financial currency hedging module to estimate the cost of heading strategies, and a sophisticated human computer interface. A list of companies in the United Kingdom that manufacture or repair large and complex equipment was drawn up and assessed in terms of the likelihood that they would be interested in the products. Companies who responded favourably were visited by a representative. Direct commercial benefit has already been experienced from the project by reduction in the administration and estimation time for producing budgets, tenders and invoices. Marketing and market research has been carried out in 3 areas: shipyards, shipping and railways. An exploitation plan includes academic papers, teaching and doctorate theses.
Systems used at present rely on an ad hoc combination of stored data, experience, heuristic rules and analytic assessments. They provide some guidance for the production of estimates but not enough to serve the needs of the repairer who has to be specific as to costs, timescales and workshop schedules and who, most times, works with a requirement from the owner which is not detailed nor can it be made more detailed until inspection and repair have actually started.

The present project aims at developing a comprehensive approach based on recent advances in A1 learning methods. The system will make the best use of existing historical data, of applicable analytical approaches, of heuristic rules and of A1, in assessing the implications of a requirement in terms of the possible resulting repair policies and in evaluating the impact of costs and schedules. At the same time the system will not be static and circumscribed by the rules and values input: it will have in itsf the capability to learn from experience and improve its policies.

2

Coordinator

British Maritime Technology Ltd
Address
Waldegrave Road 1
TW11 8LZ Teddington, Middlesex
United Kingdom

Participants (3)

Aabenraa Motorfabrik, Heinrich Callesan A/S
Denmark
Address
Naestmark 30
6200 Aabenraa
Codindustria - Cooperativa de Tecnicos de Desenvolvimento Economico e Industrial Crl
Portugal
Address
Av. Fontes Pereira De Melo 19
1000 Lisboa
Lisnave - Estaleiros Navais SA
Portugal
Address
82,Rua De S. Domingos Lapa 82
1103 Lisboa