Project summary: Nonlinear data streams
The REMIND project has demonstrated the potential of two distinct condition monitoring methods for improving the operational management of industrial systems.

We have developed a model of the National Grid System for which the model parameters represent physical quantities. In order to make operational decisions, grid system managers require estimates of these parameters and their uncertainties. With this information, system managers are better able to run the grid system more efficiently. In addition, we have surveyed the potential of nonlinear time series modelling techniques for use in novelty detection in vibrating machinery. This work is of direct relevance to InterTec (UK), as a source of new techniques to enhance their condition monitoring products and to enable users to schedule their maintenance more effectively.

Of the two methods investigated, the first provides insight into the scope of application of Bayesian parameter estimation to simulation models and the challenges faced in fusing sound theoretical methods with real-world constraints. The second method uses an industrially relevant application of nonlinear time series techniques for fault or novelty detection in multiple data streams.

Hai-Bin Wan of NGT confirms, “Preliminary results have shown that the developed model can capture the basic dynamics of the grid frequency of the National Grid system. It is expected that the model can provide useful operational information when real time data are used, which is essential for control engineers to operate the system economically and securely.”

For InterTec (UK), Dave Mellor says, “The research suggests potential for improving on current industrial practice by supplementing conventional threshold alarms and spectral analyses with techniques able to recognise subtle but significant changes in condition without extensive pre-knowledge of the machine. If further work could demonstrate the practicality of locally computable algorithms, this might enable initial assessments to be undertaken at machine level by distributed monitoring systems and reduce the quantity of data that would otherwise have to be routinely collected and transmitted onwards for remote processing.”

[Other case studies]

Download 'ProjectSummary_REMIND.pdf' (75 Kb).