(438a) A Method to Reduce Dimensionality of Powder Flow Characterization
- Conference: AIChE Annual Meeting
- Year: 2017
- Proceeding: 2017 AIChE Annual Meeting
- Group: Pharmaceutical Discovery, Development and Manufacturing Forum
Tuesday, October 31, 2017 - 3:15pm-3:37pm
Background: Material flow properties often have substantial impact on the performance of the manufacturing processes. For solids materials, the flow behavior is a multi-dimensional characteristic. Currently, a practical way is to measure powder properties using several standardized lab scale tests so that the flow behavior of a material can be described using a large group of indices. However, this approach is often time-consuming, and some of the indices obtained may be redundant. Purpose: The purpose of this study is to examine the intrinsic relations between different flow indices in a multivariate approach, and to develop a methodology that identifies an optimal set of measurements depending on the availability of time, materials, and instrument. Method: The proposed methodology includes techniques to characterize material flow properties and multivariate analysis. Pharmaceutical materials with varying flow properties were firstly characterized. The flow properties of each material was represented by multiple flow indices. To reduce the dimensionality of the database, the mathematical correlations between flow indices within each test were firstly examined. Computational iterations of principal component analysis was then performed. Depending on availability of instruments, the amount of materials, and time needed for each test, each flow index can be associated with an assigned weight. The criteria of selecting the optimal set was based on correlation coefficient of the Weighted Euclidean distance. Results: Results suggested that by assigning different weights to each flow index depending on the availability of time, materials, and instrument, different optimal sets of characterization measurements can be identified. The optimal set significantly reduced the cost of measurements. In addition, compared to the original dataset, the optimal set was able to capture maximal information of the material flow behavior. The optimal set of measurements can also be used prior to establishing the predictive correlations between material flow properties and process performance. Conclusion: The work presented here has shown an efficient approach to reduce dimensionality of material flow testing. This approach is especially powerful when the amount of time, instrument, or a material is limited, or when latent variables representing material flow properties are needed to develop process control strategy.