A soft sensor, sometimes called a virtual or inferential sensor, is a model-based estimate of a process variable that is not measured directly by a physical instrument — for example inferring a metabolite concentration from temperature, pH and flow readings. In this dataset the term is used broadly, covering thermal processing, semiconductor fabrication and petroleum refining as well as bioprocessing, because the underlying validation methods are shared across industries. Patent claims typically combine the estimation model itself with a validation step, such as an external validation set or a residual-analysis check, that proves the estimate is reliable enough to act on.

Filing is concentrated: the ranked leader alone holds 16 of the 84 records in scope, and the top 5 assignees together account for 54.8% of all records. The leaders are drawn mostly from industrial process-control and instrumentation backgrounds rather than biomanufacturing specifically, reflecting the field’s roots in thermal and chemical process control. Beyond the top 10 filers, who together cover 73.8% of records, the remaining ranked assignees each hold only a handful of filings.

Filings grew from 6 in 2021 to a peak of 10 in 2024, a 67% increase over that three-year span, which is genuine growth rather than noise. The apparent drop in 2025 and 2026 should not be read as a slowdown: patent publication typically lags filing by around 18 months, so the most recent one to two years are still incomplete in any dataset. The 2024 figure is the most recent year that can be treated as a complete picture of filing activity.