Homogeneity analysis of particle mixing in a fluidized bed by using colour histogram and artificial neural network analysis / Syaidatul Akma Mohd Zuki

In pharmaceutical industry, mixing is a critical process that plays a significant role towards the quality of the final product. The mixing process, therefore, must be scrutinized to ensure homogeneity of the mixtures. Homogeneity of the process can be analyzed by either invasive or non-invasive method. Because of disadvantages of invasive methods, non-invasive methods such as Digital Image Processing (DIP) have shown great potential in the field of particle mixing and in this study. However, the DIP methods that have been used by many researchers used expensive image acquisition devices and complicated techniques. Hence, low cost web camera as image acquisition device and a simple technique called colour histogram was implemented in this research to analyzed particle mixing in a fluidized bed. There are three objectives in this study which are; to develop an image acquisition system and analyze on mixing images in fluidized bed for determination of homogeneity time, to study the effect of air velocity and particle size on mixing performance of particles in fluidized bed by colour histogram analysis and to develop an Artificial Neural Network (ANN) based system. This study begins with image acquisition of mixing image in a fluidized bed using web camera. Then, the mixing images have been analyzed by colour histogram analysis to determine mixing time, segregation time and homogeneous time. ANN analysis has been done to support the finding. From the results, the homogeneous times for random position were in the range of 44 to 82 second and 48 to 86 second for layered position. The homogeneous time was decreasing as the air velocity increased for all experiment. For set 2 and 3 that consist of different particle size, the homogeneous time was slightly higher compared to set 1 and 4 that had the same particle size. It was found that layered position gave higher homogeneous times compared to random position. ANN analysis has showed that most of the data set produced acceptable performance accuracy which is above 80 % which means that the data was considered a good data. The proposed method serves as a proof of concept that particle mixing alters the colour distribution in the image, thus can be used to analyze the quality of the mixing process.