Particle size is a critical indicator of product quality, significantly affecting product stability, solubility, and flowability. With the advancement of science and technology and the improvement of industrial standards, particle size measurement has become increasingly important in many fields, such as chemical engineering, pharmaceuticals, and materials science. Among the numerous particle size distribution measurement techniques, ultrasonic attenuation spectroscopy has attracted the attention of many researchers due to its strong penetration ability, wide frequency range, fast response speed, and non-contact advantages. The most classic theoretical model in ultrasonic attenuation is the ECAH model, which is widely applicable because it covers most comprehensively the attenuation mechanisms. However, a major limitation of the ECAH model is that it requires many material properties parameters, many of which are unknown or inaccurate. Given some test run results, it is possible to use a retrofitting process to determine what those unknown/inaccurate values should be to minimise the error between the measured and ECAH predicted results. The aim of this project is to compare error minimisation algorithms and evaluate how they perform in different scenarios. The main novelty of the research is that both unknown/inaccurate material properties and particle size distribution (PSD) parameters can be determined simultaneously through an optimisation process. The tested optimisation algorithms include Genetic Algorithm (GA) optimisation, Particle Swarm (PS) optimisation and Parallel Traversal (PT) algorithm.
This research will have a significant impact on the field of ultrasonic attenuation spectroscopy for PSD measurement. Firstly, for the first time, a systematic sensitivity analysis has performed for all the optimisable parameters. This is useful in narrowing the range of values a parameter can have when it is being optimised, thus helping to speed up the optimisation process. Secondly, the simultaneous optimisation of both material properties and PSD parameters has been shown to give more accurate results than optimising parameters and PSD separately. Finally, test result have indicated that if the number of parameters to be optimised is small (e.g., <=3), PT is the quickest among the three for comparable setups; for more parameters, GA runtime is more predictable than PS.
