Skip to main content

Applying Artificial Intelligence to Big Data

A world-leading manufacturer with a highly complex transportation network, wanted to improve its demand strategy forecasting, due to significant increases in available data, in order to speed up its decision-making processes. The company’s engineers were finding limitations with commercial software, which could only deliver partial solutions.

As a result, they ran a competition for universities and in-house research teams to develop new technology, with the aim of optimising Caterpillar’s supply chain.

Brunel entered the competition and were successful in being selected to develop the technology, securing funding from Caterpillar to conduct extensive research and provide solutions. Working collaboratively with the organisation, we investigated fourteen different models, progressing two that performed well. Using and adapting these algorithms, the university developed solutions to provide optimal cost and efficiency savings in the supply chain. Data included many aspects of transportation such as which ports, how many ocean lanes, size of materials and containers required, and also factory production and assembly line processes. Brunel ran complex analysis, which also factored in resilience for situations beyond Caterpillar’s control e.g. weather.

Most of the previous solutions were based on linear programming: we designed bespoke artificial intelligence software to effectively manage data in a highly dynamic environment. The new technology is now fully deployed by Caterpillar. Since implementing the software, the company has made significant cost savings and minimised losses across the global supply chain. In addition, the software has improved managers’ visibility and understanding of their networks, allowing for more informed decisionmaking and substantial reductions in the time taken to make global supply chain decisions, to within 24 hours. Brunel is continuing its research into more complex programs, using intelligent algorithms and mapping against hardware structures. We are also looking at ways to develop and adapt the technology for other organisations.