Modeling Industrial Crystallization with LLM Support: Strengths, Pitfalls, and Practical Insights

Contribution to the 31st International Workshop on Industrial Crystallization (BIWIC 2026), 2-4 September 2026, Technorama, Winterthur, Switzerland. Advanced process models are indispensable in industrial crystallization for process design, optimization, monitoring, and control. Yet their development remains challenging, as it requires both deep process understanding and substantial programming effort. Physically meaningful models must capture not only mass and energy balances, but also crystallization-specific phenomena such as nucleation, growth, and the development of particle size and shape distributions. Building such models and implementing them efficiently in code is often labor-intensive and demands broad expert knowledge. This contribution explores how large language models (LLMs) can support this task. Focusing on modeling and programming, it shows how LLMs can assist in formulating model equations, structuring simulation approaches, generating and revising code, debugging implementations, and documenting model development. Using evaporative crystallization as an example, it illustrates both the practical potential and the current limitations of these tools. In addition to highlighting strengths, typical pitfalls are discussed and practical recommendations for their effective use are provided. Emphasis is placed on the need for critical review and physical validation of LLM-generated content. Overall, this contribution offers a snapshot of the current state of readily accessible LLM tools for advanced modeling and programming in industrial crystallization.