In Situ Image Analysis of Batch Settling of Flocculated Minerals with Physical-Statiscal Feature Extraction Method

"An image analysis method based on chord length distributions is applied to the batch settling of polyacrylamide flocculated kaolin suspensions. Two features of the interaggregate void space, the void chord length distribution mean and the white-to-black distribution ratio are extracted from an image series obtained over the two hour settling experiments. This approach produces three signal values, from two different probability distributions that can be measured as an indication of underlying sediment bed density and aggregation state. Further work is required on the physical interpretation of the signal, the image acquisition conditions, and the parameter selection methods. It is suggested that this chord length based feature extraction technique may be a useful signal to monitor during batch settling or thickening.INTRODUCTIONFew in situ techniques are available that can be used to directly measure aggregate properties within sediment beds in thickeners or secondary settlers. Typical process variables monitored include rake torque and height, bed level and pressure, feed rate and density, underflow rate and density, settling rate, and overflow turbidity; yet these process variables are only weakly related to process dynamics (Concha et al. 2014). Additional process monitoring signals, based on the direct measurement of aggregate properties could improve solid-liquid separation control and allow the extraction of more detailed information from batch settling experiments.Several optical methods exist for measurement of moving suspensions of aggregated solids in the dilute concentration regime, such as focused beam reflectance measurement (Peng and Williams 1994), and the turbidity fluctuation method (Gregory 1985). In the more concentrated region of a thickener, acoustic backscatter penetration and attenuation measured with an acoustic probe and changes in the underlying solid concentration in the highly concentrated, near-bed region of a thickener as a function of height and time (Hunter et al. 2012). While machine vision for froth monitoring has been thoroughly studied with at least five commercialized technologies (Aldrich et al. 2010), its’ application to the study of thickening has been limited."