Soft-Sensor Bioprocessing Patents: Who Leads, Where the Gaps Are 2026
- Concentrated at the top. The five leading assignees hold 52.5% of all 815 records in scope, and the leading filer alone accounts for 251 of them.
- Filing has cooled from its 2019 peak. Volume hit 144 records in 2019 and moved from 89 in 2021 to 73 in 2024, an 18% pullback over that span, though 2025-26 counts are still filling in as publications lag filings.
- AI classes dominate the claim text. G06N computing/AI models appear on 83.8% of the 815 records, well ahead of G05B control systems at 54.5%, showing the field is being claimed as a machine-learning problem more than a control-systems one.
Filing growth compares 2021 (89 records) with 2024 (73) — a three-year span. 2024 is the most recent year we treat as complete: publication lags filing by roughly 18 months, so 2025 onwards are still filling in and any growth rate that ends there would understate the field. Top-5 share is the combined record count of the five largest assignees divided by all 815 records in scope (CR5), not by the ranked leaders only.
What this landscape covers
This dataset tracks patent filings at the intersection of soft-sensor and virtual-sensor techniques — inferential estimators trained on process data rather than dedicated hardware — and the training mechanics that make them work: training data curation, model fitting, parameter tuning, cross-validation and regularization. The scope is set by IPC classes covering control and regulating systems, biochemical measurement, and AI/machine-learning models, which pulls in both classic process-control soft sensors and newer AI-native inferential models applied to bioprocessing and adjacent industrial monitoring.
The 815 records in scope span 2015 through mid-2026, with United States filings dominating the receiving-office mix and Europe, the WIPO PCT route, Australia, Canada and India making up the remainder. That geographic pattern points to a field still centred on US prosecution, with international filing used selectively rather than as a default strategy.
Let an AI agent run this analysis on your own technology
Pick a task. Every answer cites the patents behind it.
Filing trend and technology composition
Two views of the same 815 records: how filing volume has moved year over year, and which IPC subclasses the claims actually sit in.
Filing trend, 2017-2026
Annual filings rose to a peak of 144 records in 2019, then eased; the 2021-to-2024 comparison (89 to 73 records, -18%) is the most recent span not distorted by publication lag, since 2025 and 2026 counts will keep rising as later-filed applications publish.
IPC subclass composition
G06N (AI/computing models) touches 83.8% of the 815 records and G05B (control and regulating systems) 54.5%, with H04L, G06F, G06Q, G06K, H04B and G06V each present on roughly a fifth to two-fifths of filings — because records carry multiple classes, these figures sum to well over 100% and should be read against the 815-record total, not against each other.
Shares are the percentage of the 815 records in scope. A patent can carry several IPC classes, so the shares add up to more than 100%.
Go deeper on Soft-Sensor Bioprocessing Model Training with Eureka
This page is one run against one query. Ask Eureka your own question about soft-sensor bioprocessing model training and every answer comes back with the patent numbers behind it.
Try EurekaRepresentative and most-cited filings
US12645190B2 — Apparatus and method for switching a substrate processing apparatus to virtual sensor based control
An information processing apparatus acquires physical sensor output from a substrate processing apparatus, predicts a virtual sensor output for a designated prediction-target sensor using a statistical or physical model based on similarity to learned data, and compares the physical and virtual outputs to determine abnormality-related switching between physical and virtual sensor based control.Filed by Tokyo Electron and published 2026-06-02, this filing shows the pattern now common in the leading assignees' portfolios: pairing a trained inferential model with a runtime switch-over and validation check against the physical sensor it stands in for.
| # | Publication no. | Patent title | Citations |
|---|---|---|---|
| 1 | US20190339688A1 | Methods and systems for data collection, learning, and streaming of machine signals for analytics and mainten… | 973 |
| 2 | US20200348662A1 | Platform for facilitating development of intelligence in an industrial internet of things system | 732 |
| 3 | US20210157312A1 | Intelligent vibration digital twin systems and methods for industrial environments | 715 |
| 4 | US11620702B2 | Systems and methods for crowdsourcing information on a guarantor for a loan | 639 |
| 5 | US5386373A | Virtual continuous emission monitoring system with sensor validation | 613 |
| 6 | US6207936B1 | Model-based predictive control of thermal processing | 579 |
| 7 | US20200225655A1 | Methods, systems, kits and apparatuses for monitoring and managing industrial settings in an industrial inter… | 563 |
| 8 | US20200103894A1 | Methods and systems for data collection, learning, and streaming of machine signals for computerized maintena… | 513 |
| 9 | US20190171187A1 | Methods and systems for the industrial internet of things | 457 |
| 10 | US20190303759A1 | Training, testing, and verifying autonomous machines using simulated environments | 330 |
Citation counts accumulate over time, so older filings from the industrial-IoT and emissions-monitoring space lead the table; treat this as a signal of influence on later filings, not a ranking of current technical importance.
Each row carries its publication number; clicking a row searches Eureka by that number.
Put your own technology through the same analysis
Eureka on the web
When you want the answer in the next five minutes.
The agent works the prompt against patents and technical literature, citing every source.
Run your analysis now →MCP server & REST API
When it has to run inside your own pipeline.
Patent search, landscape analysis and assignee resolution as MCP tools. Drop them into any agent framework, or call REST directly.
Browse MCP servers →What the numbers mean for a filing decision
Three patterns worth weighing before drafting claims in this space.
The top of the field is crowded, the rest is thin
Five assignees hold 428 of the 815 records in scope, with the leader alone at 251. Below that top group the ranking thins quickly to single-digit and single-filing entrants, which means core claim territory around training-data curation and model-fitting workflows is likely already staked out by a small number of portfolios.
Volume has eased from its 2019 peak, not collapsed
Filings peaked at 144 records in 2019 and the last fully comparable span, 2021 to 2024, shows a -18% move. That is a cooling, not a退 field losing relevance — publication lag means 2025 and 2026 figures will keep revising upward, so it is too early to call a downward trend from the most recent years alone.
This is being claimed as an AI problem first
G06N (AI/computing models) appears on 83.8% of records versus 54.5% for G05B (control systems), and G06Q business-process classes touch nearly a third of filings. Drafting strategy in this space increasingly needs to address model-training and inference claim language, not just classical process-control claim structures.
Eureka can read the same corpus for gaps instead of for coverage: under-claimed branches adjacent to soft-sensor bioprocessing model training, with the prior art for and against each one.
Who holds the ground, and where it is open
The ranked leaders span industrial-IoT platform builders, aerospace and automotive OEMs, chip and AI-infrastructure firms, and process-instrumentation specialists — a mix that reflects how broadly the soft-sensor training approach has been adopted outside its original process-control home.
One portfolio dominates the count
The leading assignee's 251 records dwarf the rest of the ranked field, built on a broad industrial-IoT sensor-and-analytics platform rather than a single product line, which suggests its claims are structured around data-collection and model-training infrastructure rather than a narrow application.
A sharp drop after the top few
Fifth place sits at 17 records and tenth place at 10, a steep fall-off from the leader's 251. That gap between the top handful and everyone else is where most of the field's single-digit filers and recent entrants sit, competing for narrower, more specific claim territory.
Recent-year activity has slowed across named leaders
Several of the most active historical filers show sharp year-over-year pullbacks in the latest tracked year, including declines to zero for some. Given publication lag, this likely reflects filings still working through the pipeline rather than a genuine stop in R&D activity.
| Assignee | Recent year | YoY |
|---|---|---|
| STRONG FORCE TX PORTFOLIO 2018 LLC | 2 | -50% |
| Strong Force IoT Portfolio 2016, LLC | 0 | -100% |
| Robert Bosch GmbH | 0 | -100% |
| ISTARI DIGITAL INC | 0 | -100% |
| Pavilion Technologies, Inc. | 0 | — |
| NVIDIA Corporation | 0 | -100% |
| General Electric Company | 0 | — |
| Sartorius Stedim Data Analytics AB | 0 | -100% |
Where to take this analysis
The landscape numbers point to next steps rather than final answers — each one is worth a closer, claim-level pass.
Map claim scope against the leading assignees
With 52.5% of records held by five assignees, a freedom-to-operate check against their specific claim language is the natural first move before drafting in core training-workflow territory.
Run a claim-scope check in EurekaTrack the under-claimed branches
Uncertainty quantification, cross-scale model transfer and drift detection show comparatively thin filing activity against the core field — worth monitoring for a first-mover claim rather than assuming they are already occupied.
Explore white space in EurekaWatch the next publication wave
2025-26 filing counts will keep revising upward as the roughly 18-month publication lag closes; re-run the trend view in six to twelve months before concluding the field has cooled.
Set up trend monitoring in EurekaFrequently asked questions
A soft sensor, also called a virtual or inferential sensor, is a model trained on process data that estimates a value a physical sensor would otherwise measure directly, such as a concentration or quality parameter in a bioprocess. This dataset scopes to filings that combine that sensing concept with model-training mechanics like training data curation, cross-validation, parameter tuning or regularization. It spans both classical process-control soft sensors and newer AI-native inferential models, which is why the technology composition leans heavily toward AI/computing IPC classes alongside control-systems classes.
The ranking of 100 assignees is dominated by a small group at the top: the five leading assignees together hold 52.5% of the 815 records in scope, and the single largest holds 251 records on its own. Below the top handful the count drops quickly, with fifth place at 17 records and tenth place at 10, leaving a long tail of assignees with only a few filings each. The mix includes industrial-IoT platform companies, automotive and aerospace OEMs, chipmakers and process-instrumentation specialists, reflecting how widely the underlying training techniques have spread beyond their original process-control context.
Filing volume peaked at 144 records in 2019 and the most recent fully comparable period, 2021 to 2024, shows a decline from 89 to 73 records, an 18% drop. That is a real cooling from the 2019 peak rather than a collapse, and it should not be read as the field losing relevance. Publication typically lags filing by around 18 months, so the 2025 and 2026 figures in the raw trend are still incomplete and will rise as more applications from those filing years publish.
US12645190B2, filed by Tokyo Electron and published in 2026, covers an apparatus that predicts a virtual sensor's output for a designated physical sensor using a statistical or physical model trained on prior data, then compares that prediction against the real sensor reading to detect abnormalities and switch control between physical and virtual sensing. It is a representative recent example of the pattern seen across many records in this dataset: pairing a trained inferential model with a runtime validation and switch-over mechanism. Anyone building a similar switch-over or validation layer on top of a trained soft sensor should review its specific claim language closely.
The core territory around training-data curation and model-fitting workflows is well covered by the leading assignees, but several adjacent branches show comparatively thin filing activity, including uncertainty quantification on inferential outputs, cross-scale transfer of trained models between bioreactor sizes, and online drift detection for deployed virtual sensors. These are not proven-empty spaces, only areas where the filing density is lower than the crowded core, so they warrant a closer prior-art check rather than an assumption of full freedom to operate. A first claim in one of these branches would need to tie the specific technique tightly to a bioprocess application to distinguish it from the broader industrial-IoT filings that dominate the leader's portfolio.
Research Soft-Sensor Bioprocessing Model Training in depth with Eureka
Go past this page: query the whole soft-sensor bioprocessing model training corpus yourself, in your own scope.
Every answer comes back with patent numbers you can open.
Disclaimer. This page is generated from Patsnap Eureka data drawn from a limited snapshot of global patent and scientific-literature records, and is provided for general information and reference only.
Patent data carries inherent limitations: recent filings (typically the most recent 18–24 months) are under-counted due to standard publication lag; counts may be reported at either a patent-family or a patent-record basis and are not always directly comparable; classification, applicant-name, and citation data may contain errors, duplicates, or omissions; and the underlying search query defines and constrains the scope shown. As a result, the analysis may be incomplete or inaccurate and may not reflect the full technology landscape.
Nothing on this page constitutes an exhaustive prior-art, novelty, freedom-to-operate, or validity search, nor does it constitute legal, financial, investment, or professional advice, and it should not be relied upon as such. Any patent, commercial, or strategic decision should be verified independently and reviewed with qualified patent, legal, and domain professionals. Patsnap makes no warranties, express or implied, as to the accuracy, completeness, or fitness for any particular purpose of the information presented.
Machine translation. Assignee and organisation names originally recorded in Chinese, Japanese or Korean have been rendered into English by an AI translation step so that the tables stay readable. These renderings are best-effort and may not match a company’s registered English name; the original name is what the underlying patent record carries, and it is what any Eureka query launched from this page uses.