(83i) Particle Size Distribution Estimation through Image-Based Sensors: Sequential Simplex Optimization for Selection of Optimal Parameters

Particle size distribution (PSD) and shape are critical to product quality in crystallization processes. A narrow particle size distribution with specific particle shape is indicative of good product quality; however, the absence of robust online sensors for size distribution prevents real-time monitoring and control. With the developments in sensor technology, instruments such as the Particle Vision and Measurement (PVM) from Lasentec / Mettler Toledo are widely used in manufacturing units to monitor PSD and shape variation. These have triggered research interest in image-analysis techniques that convert inline images to information about the process state ? especially particle size distribution. In this work, we study the robustness of image-analysis algorithms (IAA) and use design of experiments based techniques for tuning the parameters associated with the IAA.

Several approaches to estimate the process state or product quality (e.g. size, shape, and growth rate of particles) in real-time and in situ (Sarkar et al., 2009, Zhou et al., 2006, Wilkinson et al., 2000, Wang et al., 2008, Patience, 2002, Li et al., 2006, Larsen et al., 2007) have been reported recently. Crystal size estimation from video images, especially for high aspect ratio systems (e.g. L-glutamicacid), has received much attention since laser-diffraction based methods that are an alternate way to estimate size do not perform adequately for such systems. Any image-analysis algorithm involves several steps such as image selection, image enhancement, edge detection and morphology operation. A large number of user-defined parameters are involved in each of the above steps. Our studies (Zhou et al., 2009) show that with proper settings of these IA parameters, the estimated PSD can be obtained within ~10% error (when compared to results obtained by the human-eye). However, when these parameters are specified unsuitably, very large errors (~50%) could occur. Given the large variability in the quality of the results, a systematic way to select the user-defined parameters for IAA is critical. In this paper, we propose to use a systematic design of experiments methodology for selecting optimal IA parameters.

In this work, uniform design (UD) (Liang et al., 2001) and sequential simplex optimization (SSO) (Kamoun et al., 2009) are applied to find optimal IA parameters. In UD, it is important for the designed parameters to uniformly fill the whole space of parameter range. Hence, the range and the number of levels of each parameter are predefined. The results from these designed experimental runs are used to build a mathematical model relating the input parameters and output results. The the optimal IA parameters are obtained by solving this mathematical model for the optima. In SSO, experimental runs are performed with specific number of random guesses (within the allowable limits) and parameter settings for subsequent experiments are calculated iteratively based on the results from prior runs.

Mono-sodium Glutamate (MSG) seeded cooling crystallization process is selected as the case study. We focus on 7 IA parameters that pertain to the key image analysis steps of edge detection and morphology operations. Results from the two approaches will be reported and compared based on efficiency and accuracy.

References

Kamoun, A., Jaziri, M. & Chaaouni, M., (2009), The use of the simplex method and its derivatives to the on-line optimization of the parameters of an injection moulding process. Chemometrics and Intelligent Laboratory Systems 96, 117-122.

Larsen, P.A., Rawlings, J.B. & Ferrier, N.J., (2007), Model-based object recognition to measure crystal size and shape distributions from in situ video images. Chemical Engineering Science 62, 1430-1441.

Li, R. F., Thomson, G. B., White, G., Wang, X. Z., Calderon De Anda, J. & Roberts, K. J., (2006), Integration of crystal morphology modeling and on-line shape measurement. AIChE Journal, Vol. 52, No. 6, 2297-2305.

Liang, Y., Fang, K. & Xu, Q., (2001), Uniform design and its applications in chemistry and chemical engineering. Chemometrics and Intelligent Laboratory Systems 58, 43-57.

Patience, D. B., (2002), Crystal engineering through particle size and shape monitoring, modeling and control. University of Wisconsin-Madison, PhD Thesis.

Sarkar, D., Doan, X. T., Zhou, Y. & Srinivasan, R., (2009) ?In-situ particle size estimation for crystallization processes by multivariate image analysis?, Chemical Engineering Science 64(1), 9-19.

Wang, X. Z., Roberts, K. J. & Ma, C., (2008), Crystal growth measurement using 2D and 3D imaging and the perspectives for shape control. Chemical Engineering Science 63 (2008) 1173-1184.

Wilkinson, M. J., Jennings, K. H. & Hardy, M., (2000), Non-invasive video imaging for interrogating pharmaceutical crystallization. Microscopy and Microanalysis, Vol (6) (Supplement), 996-997.

Zhou, Y., Doan, X. T. & Srinivasan, R., (2006), Real-time imaging and product quality characterization for control of particulate processes. Joint 16th ESCAPE and 9th PSE: Computer-Aided Chemical Engineering 21A, W. Marquardt and C. Pantelides (Eds), 775-780. Elsevier, Garmish-Partenkirchen.

Zhou, Y., Srinivasan, R. & Samavedham, L. (2009) ?Critical evaluation of imaging based techniques for real-time crystal size measurements?, Computers & Chemical Engineering, volume 33, issue 5, 1022-1035.