Manufacturing Analytics: Intelligent Supply Chains with AI - DIAL Newsletter Summer 2019

AI is at the very heart of the Manufacturing Analytics group, whose mission is to design, develop and use state of the art AI approaches for improving factory, supply chain and logistics operations. Recent examples include:

Uncertainty quantification in supply chain disruptions: Deep neural networks are a powerful technique for learning complex functions from data. However, their appeal in real-world applications can be hindered by an inability to quantify the uncertainty of predictions. DIAL’s Tim Pearce has recently developed a technique for the generation of prediction intervals (PI) for quantifying uncertainty [1]. High-quality PI should be as narrow as possible, whilst capturing a specified portion of data, with no distributional assumption. Tim’s algorithm is currently being applied to predict supplier delivery performance in Boeing. Whilst previous classification based efforts focussed on predicting whether a delivery would be late or not, Tim’s approach enables us to predict a late delivery with an associated time window of uncertainty, which is more informative to the manufacturer.

Autonomous Supply Chains: In the Pitch-in project, we are developing a platform to demonstrate how IoT, machine learning, and agent-based technology can work together to deliver an autonomous supply chain. Drawing upon the inputs from sources such as IoT data and machine learning models, agents in this platform can act on behalf of the stakeholders in the supply chain to autonomously select suppliers, procure goods and respond to unplanned events. For example, a response to a traffic disturbance or a fridge break down during the cold chain could involve automated rerouting. Automated pricing algorithms can be plugged into the platform to adjust prices based on delivery delays, and resulting quality degradations. Stay tuned to find out more!

Link prediction: In a previous study we had found out that in large-scale complex supply chains, first tier manufacturers had a high probability of sharing links. Visibility into these linkages is vital to plan for disruptions. In a project sponsored by Boeing, we are creating graph mining based classification approaches to predict the existence of a dependency relationship between suppliers [2].

  1. Pearce T., Zaki M., Neely A., Brintrup A., High quality prediction intervals for deep learning: A distribution free objective function, 35thInt. Conf. on Machine Learning, July 2018.
  2. Brintrup, A., Wichmann, P., Woodall, P., McFarlane, D., Nicks, E. and Krechel, W., 2018. Predicting hidden links in Supply Networks.Complexity,2018.

Date published

25 July 2019

For further information please contact:

Dr Alexandra Brintrup

E: ab702@cam.ac.uk

 
Share This