Improving in situ acoustic characterisation of suspensions with machine learning methods

Waste suspension flows are encountered across the nuclear sector, and their characterisation is of great importance to the safe transport of radioactive material. A waste treatment program that has received high priority in recent years, is the retrieval of legacy waste sludge from historical ponds to safe interim storage facilities. Suspension transfer via pipeline has encountered several problems resulting from a lack of relevant design data. As such, operations are often run with extreme caution and not necessarily at their optimum, which may cause further downstream problems. For example, to mitigate the potential for radioactive particles to block pipes, high flowrates are employed, leading to shear breakdown of aggregates that reduces waste consolidation rates downstream in the interim stores. This project seeks to understand these issues and overcome pipeline transportation problems in two ways – the development of an online acoustic backscatter technique for remote characterisation of particle size and concentration online, and the use of polymer additives to modify slurry characteristics enabling safe and efficient slurry transport.

Acoustic backscatter systems have been previously developed at Leeds as an in situ technique to characterise suspension wastes and they show great promise as a technique to measure slurries in pipe-flows non-invasively online. In particular, ultrasonic backscatter strength can help determine particle size or concentration, while simultaneously, speed of sound can also be used to correlate the aggregation state of slurries. There are still significant challenges however, in the use of ultrasonic measurements to characterise particle properties in pipes. While there is potential to extract size and aggregate data from acoustic backscatter parameters, information is highly convoluted, and currently quantification only extends to spherical dispersions that are unlike the wastes encountered. There is also a lack of theoretical models that may be used to predict size from backscatter and velocity parameters. As a further complication, the turbulent nature of pipe flows may alter aggregate size and structure over time, and the influence of particle aggregation states (in regards to backscatter and speed of sound measurements) also requires investigation. We will therefore initially aim to classify acoustic behaviour in a small well controlled variable shear cell (of similar dimensions to a full pipe) with various disperse and aggregated test materials of relevance to nuclear waste flows, before moving into full pipeline implementation. Other characterisation techniques (e.g. focused beam reflectance measurements and particle image velocimetry) will also be used to correlate key properties of the slurries. Focus on the project will be to deconvolute acoustic backscatter and speed of sound parameters by extending current acoustic theory to be able to extract size and aggregation information from backscatter measurements. The influence of large particle size distributions will be another important consideration. Artificial Neural Networks (ANN) and other machine learning techniques may also be applied to match acoustic response patterns to systems of different particle sizes and distributions.

A second major aim of the research project will be to understand the influence of polymer additives to modify slurry flow behaviour under various shear regimes, and the resulting effect on acoustic backscatter response. Polymer flocculation agents may be added inline to aggregate waste streams for enhanced sedimentation downstream. However, such additives will also affect slurry rheology and pipeline behaviour. Ultimately, the addition flocculation agents and other polymer structurants may result in highly aggregated paste type flows that will have very different deposition and flow behaviour to the non-flocculated wastes . Work will focus on firstly characterising the structure-property relationships of the polymer-particle slurries (e.g. through rotational rheometry and measurements in the shear cell) before their behaviour in full pipeline conditions will be examined.


Academic Lead: Tim Hunter
Researcher: Joe Hartley
Location: University of Leeds