| Summary: | 碩士 === 長庚大學 === 化工與材料工程學系 === 100 === In this study, we performed an optimized parameter search of the eflucimibe polymorph nucleation mechanism. This mechanism uses gradual disruption theory as a theoretical background to calculate the induction period. A comparison of the average relative errors was then conducted with the eflucimibe polymorph induction period experimental data obtained from literature. The combination of parameters with the lowest average relative error was the optimal parameter.
Eflucimibe forms A-form and B-form polymorph crystals. Their relative nuclear rates determine the final crystal form obtained. Therefore, the nucleation mechanism’s influence on eflucimibe is extremely significant. The aggregation and disruption mechanisms in gradual disruption theory use Smoluchowski’s aggregation theory. Then, because clusters smaller than critical nucleus size have unstable thermodynamics, they gradually dissolve and are disrupted. Although this theory comprises only aggregation and disruption parameters, the nucleation rates of dozens of clusters of varying sizes are interrelated. Thus, appropriate nucleation parameters are difficult to speculate. We employed optimization algorithms to identify the optimal nucleation parameters and investigate whether gradual disruption theory is consistent with eflucimibe nucleation mechanisms.
The optimization system used in this study had virtually no restrictions, its search range was extremely broad. To increase the search efficiency, we divided the system into several search ranges. Genetic algorithms and the random jump method were used to seek lower average relative errors. Subsequently, the parameter combinations with lower average relative errors were used as Nelder-Mead Simplex Method starting points to obtain the minimum average relative error values. Following analysis and comparison of the parameter combinations in each range with lower average relative errors, ranges where optimal parameters may exist were suggested. Using this method, the optimal parameter values were identified.
Numerous divergence points are typically encountered when seeking optimal parameters due to a lack of understanding of the objective function surface characteristics. This leads to poor genetic algorithm search efficiency. The Nelder-Mead Simplex Method follows function surface decreases and can efficiently avoid divergence points. Additionally, it does not require prior understanding of the search space and is suitable for use in this study’s optimization system. Therefore, the most appropriate algorithm combination for this study is the random jump method matched with the Nelder-Mead Simplex Method. Finally, we successfully obtained the optimal nucleation parameters kA: 3.62 10-21(cm3/s), kB: 5.92 10-21(cm3/s), kd,A: 8.61 10-2(1/s), and kd,B: 17.15 10-2(1/s), and the minimum average relative error of 8.15%.
|
|---|
