OBJECTIVE To optimize the molding process of Xiebai Granules (XG) using the Box-Behnken design-response surface method combined with the BP neural network method, and to evaluate the consistency of particle quality between different batches by establishing physical fingerprint.

METHODS Dry paste powder was used as the main drug, dry granulation was adopted, and the forming rate, dissolution rate, moisture absorption rate and angle of repose of the granules were used as evaluation indexes, the excipients dextrin, maltodextrin and lactose of the particles, were screened by single factor test combined with simplex-lattice design and entropy weight method, and the optimal excipient ratio was selected. The entropy weight method combined with the Box-Behnken design-response surface method and the BP neural network algorithm were used to optimize the process parameters, and the process verification was carried out. The physical fingerprint was used to comprehensively characterize the bulk density (Da), hygroscopicity (H), moisture (HR), tap density (Dc), angle of repose (α), Hausner ratio (IH), relative uniformity index (Iθ), Carr index (IC), and interparticle pore number (Ie), and the consistency of particle quality in different batches was evaluated.

RESULTS The optimal ratio of excipients was dextrin 15%, maltodextrin 48%, and lactose 37%. The optimal process parameters were conveying speed 95 r ·min^-1, pressure wheel speed 4 r ·min^-1 and hydraulic pressure 7 MPa. The similarity of the physical fingerprints of the five batches of XG was greater than 0.98.

CONCLUSION The optimized molding process of XG is stable and feasible, and the quality of different batches of XG is stable, which can provide a reference for the development and industrial scale-up production of XG.