According to retrieved records, the core BO architecture comprises: (1) a surrogate model — predominantly Gaussian Process regression — that approximates the unknown objective function using prior observations; (2) an acquisition function such as Expected Improvement, Lower Confidence Bound, or entropy search that determines the next evaluation point by balancing exploration and exploitation; and (3) an iterative update loop that incorporates new experimental or simulation results to refine the surrogate.

In this dataset, Robert Bosch GmbH holds the most manufacturing-specific BO patents, with 5 active or pending US patents filed between 2021 and 2025. These cover manufacturing machine parameter setting, evaluation point selection with safety-constrained posterior variance limits, hybrid experiment-simulation BO, and a 2025 active patent on laser material processing machine parameter optimization.

IBM’s batch BO patents (2022–2023, US and GB) focus on introducing real-time stopping criteria applied to BBO acquisition scores during industrial process runs to avoid wasting experimental budget across parallel trials. Robert Bosch’s constrained BO patents (2021–2024, US) restrict next evaluation point selection to regions where the predictive variance of the posterior model is below a specified limit, enabling safe exploration on physical manufacturing equipment with hard operating constraints.

Geminus.ai’s patent family (US 2024, WO 2024, US 2025, US 2026 pending) co-mingles data across multiple concurrent acquisition functions — including Expected Improvement and model variance — running against physics-based digital twin models. This moves beyond single-acquisition sequential BO toward parallel, multi-strategy optimization, achieving more globally optimal solutions than any single acquisition function alone.

Literature records in this dataset cover: atmospheric plasma spraying and fused deposition modeling (2021–2022); CNC machining center MPC parametrization and autonomous turning process setup (2020); powder film drying with 32,768 parameter combination search (2022); biopharmaceutical seed train design and upstream bioreactor optimization (2021–2022); battery electrode material design using multi-fidelity BO (2023); MIMO PID controller tuning, cascaded axis drive control, and HVAC system energy optimization (2020–2023); and particle accelerator and free-electron laser tuning (2020–2021).

The dataset identifies two primary open IP opportunities. First, transfer learning and warm-starting BO methods — covered by literature records on warm-starting BO (2016) and transfer learning-based search space design (2022) — are underrepresented in the patent record, suggesting opportunity for manufacturers seeking to reuse optimization knowledge across product variants or production lines. Second, biopharmaceutical and battery manufacturing are identified as fastest-growing application verticals based on recent literature density, with the materials synthesis hyperparameter tuning cluster (2022–2023 records) flagged as containing patentable implementations.