Modeling of a roller-compaction process using neural networks and genetic algorithms

Abstract

In this study, roller-compaction of acetaminophene was studied to model the effect of binder type (hydroxypropyl methyl cellulose (HPMC), polyethylene glycol (PEG), Carbopol), binder concentration (5, 10 and 20%), number of roller-compaction passes (one or two), and extragranular microcrystalline cellulose addition on the properties of compressed tablets. Forty-two batches resulted from the experimental design. The artificial neural network methodology (ANN) along with genetic algorithms were used for data analysis and optimization. ANN and genetic models provided R-2 values between 0.3593 and 0.9991 for measured responses. When a set of validation experiments was analyzed, genetic algorithm predictions of tablet characteristics were much better than the ANN. Optimization based on genetic algorithm showed that using HPMC at 20%, with two roller-compaction passes would produce mechanically acceptable acetaminophene tablets. PEG and carbopol would also produce acceptable tablets perhaps more suitable for sustained release applications. Using PEG as a binder had the additional advantage of not requiring an external lubricant during tablet manufacturing. (C) 1999 Elsevier Science B.V. All rights reserved.