Simultaneous modeling prediction of three key quality attributes of tablets by powder physical properties

Tensile strength (TS), solid fraction (SF), and tablet weight variability (TWV) are all key quality factors to be considered during the tablet manufacturing process. For predicting them, a novel powder texture measurement methodology was proposed to study the texture attributes of 32 types of powders, and 10 types of unknown powders were used to validate the model. Principal component analysis (PCA) was used to classify the powders. Standard least-squares models were constructed to predict the TS and SF of tablets using texture properties as independent variables, with model R² values of 0.9188 and 0.8672, respectively. Moreover, due to the advantages of decision trees in classification performance and computation time, multi-node decision trees were constructed, and the approximate range of the TS and SF can be quickly predicted by using only 2-4 key texture attributes. Partial least squares-discriminant method (PLS-DA) algorithm was used to determine whether TWV was qualified by using texture and other physical attributes of powders. In conclusion, the constructed models predicted TS, SF, and TWV well and provide an effective reference for facilitating the development of tablet formulation for direct compression (DC).