Industrial Artificial Intelligence
Data in industrial artificial intelligence (AI) can be of poor quality, broken, sparse or involve highly transient patterns, which makes interpretation difficult. Meanwhile, industrial applications have zero tolerance for false positives or negatives because they are dealing with critical issues relating to safety and reliability of operations. These must be viewed in the context of fast productions times and high equipment costs.
Our research is centred on deriving efficiencies from large volume data via the application of AI based machine learning algorithms including Bayesian and GAN-based methodologies. Non-parametric models that integrate new knowledge obtained from hybrid datasets, such as synthetic and actual data, are built to enhance the quality of prediction.
Finally, we attempt to go beyond AI prediction and offer root cause analysis for anomalies, using domain expertise and know-how of modelling processes to integrate failure modes and degradation mechanisms into models which explain AI.