Probabilistic Transient Propagation

Probabilisitic Transient Propagation Research Project funded by EPSRC (EP/N027507/1), http://gow.epsrc.ac.uk/NGBOViewGrant.aspx?GrantRef=EP/N027507/1

Further info can be obtained by contacting Dr Richard Collins r.p.collins@sheffield.ac.uk

Department of Civil and Structural Engineering, University of Sheffield, UK

Vision

The vision for this proposal is to develop a novel and completely revolutionary Probabilistic Transient Propagation (PTP) model to transform how our water supply infrastructure is modelled and understood. The model will provide the ability, for the first time, to assess where and when extreme, and/or repetitive, transient pressures occur and the risk of damage they pose to our vital infrastructure assets. In this project a modelling tool will be created that intrinsically accounts for the inherent uncertainty in these systems. By considering uncertainty from the start, we will not only be able to predict the transient pressure at any given point in the network, but also be able to give the probability of occurrence, placing vital additional knowledge in the hands of network operators.

Hydraulic transients are modified by every feature of the system through which they pass; they then capture a huge amount of system information as they propagate. This information, if suitably decoded, can give access to vital knowledge about the condition and operation of the networks. By predicting the likelihood of how transients will propagate, transform, and interact with pipe infrastructure, we can ensure the long term sustainable operation of these indispensable yet ageing networks by creating the ability to determine, in high resolution, the system state and condition throughinference from real world measured data. The first stage towards this long-term goal is the focus of this proposal; the development of the PTP model. This model is the critical foundation for the full probabilistic inference framework, and, by itself, will give water utilities the ability to properly assess transients and their impact on the infrastructure.

Large scale transient activity captured as part of the PI’s previous study in live Water Distribution Systems

Approach

The current state of the art in transient modelling involves the parameters of the model being assigned as single values based on assumed best guesses. A single deterministic solution is generated. It is then left up to the modelling operator as to the level of trust that should be placed in the results. Using the new modelling tools proposed here the various uncertainties in model parameters will be collated and included in the model from the outset. The dynamics and interactions of the uncertainties are then propagated naturally through the simulation, producing a result that predicts not only the outcome but also the probability of that value. This produces two transformative advantages: the first ensures that all system uncertainties are accurately represented in the output, removing the need for modelling operators to assign their own beliefs to the results; the second is that it gives the results as magnitudes/probabilities, allowing outputs to be naturally put into a risk frame, directly giving network operators the tools needed to assess the risk of negative impacts on their assets. The approach taken in this proposal will be to treat data uncertainty as random variables or random processes.

Report abuse