陶氏化学公司

The closed-loop performance of Model Predictive Control (MPC) depends on the nontrivial selection of several tuning parameters. Recently, data-efficient methods such as Bayesian Optimization (BO) have been proposed for automatic MPC tuning. In practice, it is often challenging to specify a single objective function to balance multiple criteria, especially when some of these are qualitative in nature. In these cases, Preferential Bayesian Optimization (PBO) can be used as a human-in-the-loop alternative to BO-based automatic tuning methods. By incorporating expressed preferences between pairwise comparisons of different closed-loop responses, PBO searches for the optimum of an underlying utility function that reflects the user’s preferences towards the closed-loop response from various controller parameters. However, standard PBO does not leverage comparison data from previous tasks, resulting in the need to learn preferences from scratch for each new task. In this paper, we introduce Multi-Task PBO (MTPBO) for MPC tuning, which leverages data from previous preference-based controller tuning tasks to accelerate the search of optimal parameters for either a new human user or a similar closed-loop process. Additionally, we introduce a multi-task initialization strategy that enables a more effective warm start by proposing a batch of promising initial experiments for new tasks. The advantages of MTPBO with initialization are shown on benchmark optimization functions and an offset-free MPC tuning problem with feedback from multiple actual human users on different simulated processes. Overall, the proposed MTPBO framework leads to more preferred responses with a lower experimental budget and can be applied to general controller tuning problems.

Artificial intelligence (AI) is rapidly transforming the chemical industry, offering solutions to longstanding challenges in optimization, process monitoring and control, and product development. This article provides new insights by explicitly connecting recent technical advances in AI with organizational strategies, offering an integrated perspective on how these elements collectively drive transformation in the chemical industry. While the article discusses the potential of large language models, it places greater emphasis on the critical role of data availability, policies, and broader AI adoption challenges. The article concludes by listing possible improvements achievable through AI and emphasizing the importance of leadership and collaborative initiatives between industry, academia, and government.

A homologous series of dendrimer tetrasiloxane (tris(trimethylsiloxy)silylpropyl) ammonium surfactants are synthesized and investigated for their aqueous surface tension, bulk phase behavior, and wetting behavior. Each surfactant homologue comprises two strongly surface-active components, a tetrasiloxane monoquat and a tetrasiloxane diquat, and the different homologues have different mono-to-diquat ratios. The two components appear to mix nonideally in the bulk solution as well as at the liquid-air interface, as evidenced by the formation at above a critical aggregation concentration of two coexisting association structures in the bulk liquid and two condensed surface states. While the lower surface tension state allows superspreading of water over a solid paraffin surface for all the homologue surfactant in the series, the maximum spreading area differs greatly for the different homologues. Surface tension measurements from this work also confirm a benchmark for the lower bound of aqueous surface tension enabled by the permethylated siloxane surfactant to be near 19.5 mN/m, which is still a significant gap from the lowest possible surface tension exhibited by neat permethylated siloxane oil (15.7 mN/m) and perfluorinated surfactant in aqueous solution.

Flexibility is a crucial characteristic of industrial systems that face increasing volatilities and is therefore essential to ensure feasible operation under uncertainty. Flexibility is often closely tied to the design of a system, and careful consideration must be taken to understand the trade-off between design cost and operational flexibility. In this work, we introduce a design optimization approach that we call design for flexibility , which incorporates a rigorous measure of flexibility directly into the objective function. We employ adjustable robust optimization to model uncertainty and allow for recourse in operational decisions. Compared to traditional flexibility analysis, the proposed approach can accommodate complex uncertainty sets beyond hyperrectangles as well as multiple flexibility indicators, allowing for a more comprehensive representation of uncertainty. We apply the proposed approach to three case studies, where the results demonstrate its versatility and effectiveness in rigorously evaluating the trade-offs between cost and flexibility when designing industrial systems.

Next generation risk assessment (NGRA) aims to enable transparent, reproducible chemical safety assessments based on human-relevant, animal-free new approach methodologies (NAMs). The Alternative Safety Profiling Algorithm (ASPA) was developed within the ASPIS cluster to provide an algorithmic workflow that structures problem formulation, evidence integration, and decision-making across three main pillars – hazard, ADME (toxicokinetics), and exposure. A stakeholder workshop was organized to refine ASPA. Four breakout groups systematically reviewed corresponding workflow sections, identifying strengths, conceptual gaps, and opportunities for harmonization. Across groups, participants endorsed ASPA’s modular, technology-neutral nature and its focus on standardizing processes rather than prescribing specific test batteries. The hazard pillar discussions emphasized a sensitive, hypothesis-generating Tier 1, complemented by a specific, mechanistic Tier 2, capable of deriving points of departure (PoDs). ADME experts supported a physiologically based kinetic (PBK) modelling strategy, advancing from generic towards more complex models, using mechanistic information and experimental data. The exposure group proposed refinements for transparent, tiered exposure modelling, with emphasis on realistic worst-case scenarios and explicit uncertainty communication. Cross-pillar discussions highlighted the importance of feedback loops among all pillars, and the documentation of decision points to achieve consistency and defensi­bility. The workshop outcomes informed three parallel developments: (i) algorithmic refinement and re-design toward the next ASPA version, (ii) the creation of detailed guidance for each building block, and (iii) the establishment of practical case studies to demonstrate workflow implementation. This report already contains a first case study (developmental neurotoxicity assessment of desnitro-imidacloprid). These advances increase the operability, transparency, and regulatory readiness of ASPA.