Predictive Maintenance Patents: Top Filers & Filing Trends 2026
- Filing is accelerating fast, not maturing. volume moved from 6 records in 2017 to 80 in the latest year, with the midpoint year 2022 at just 2 filings and 2025 the peak so far at 106.
- Control systems and AI models dominate the claim mix. G05B (247 records) and G06N (229) dwarf the testing and structural-balance class G01M at just 17, despite vibration feature extraction being a named search criterion.
- India leads filing volume by a wide margin. 226 of the receiving-office records in this dataset originate in India, more than four times the United States at 51.
What this patent landscape covers
This dataset tracks 310 patent families published between 2015 and mid-2026 that combine predictive maintenance or remaining-useful-life prediction with technical triggers named directly in the search criteria: vibration feature extraction, labeled failure scarcity, false alarm cost, sensor drift and maintenance scheduling. The classification mix spans G05B (control and regulating systems), G06N (AI-based computing) and G01M (testing and structural balance), which means the corpus captures both the control-loop software layer and the underlying sensing hardware.
Filing activity has moved from a small handful of records in 2017 to a rapidly climbing curve through the most recent complete year, concentrated heavily in India's receiving office relative to the United States and other jurisdictions. The most-cited records anchor the data-ingestion layer of the stack rather than the prediction algorithm itself, a pattern worth keeping in mind when assessing where freedom to operate is tightest.
How the filing pattern has moved since 2017
Filing volume in this dataset has moved from a handful of filings in 2017 to a broad, still-climbing curve, with the technology composition skewed heavily toward control systems and AI-model computing rather than pure sensor hardware.
Filings are still accelerating, not plateauing
Filings moved from 6 in 2017 to 80 in the most recent year, with 2025 the peak year so far at 106 records. The midpoint year of 2022 sat at just 2 filings, which means almost the entire filing volume in this dataset has arrived in the last few years — publication lag means the most recent year understates true filing activity further.
Control systems and AI models dominate the classification mix
G05B (control and regulating systems) leads at 247 records and G06N (AI-based computing) follows at 229, with G06Q (business and admin data processing) at 144 showing that a large share of this field is being filed as scheduling and decision logic, not raw sensing. Testing and structural-balance claims (G01M) sit far behind at 17, alongside data recognition (G06K) and monitoring/checking devices (G07C), also at 17 each.
Shares are the percentage of the 310 records in scope. A patent can carry several IPC classes, so the shares add up to more than 100%.
Go deeper on Predictive Maintenance for Industrial Equipment with Eureka
This page is one run against one query. Ask Eureka your own question about predictive maintenance for industrial equipment and every answer comes back with the patent numbers behind it.
Try EurekaRepresentative filing and the most-cited records
Multi task learning with incomplete labels for predictive maintenance
The filing describes training multiple predictive maintenance models through a multi-task learning architecture with shared generic layers and task-specific layers, addressing the problem of incomplete labelling — a direct response to the labeled failure scarcity issue that limits how much real-world failure data is available to train these models.Filed by Hitachi, Ltd. on 2021-02-18.
| # | Publication no. | Patent title | Citations |
|---|---|---|---|
| 1 | US20200103894A1 | Methods and systems for data collection, learning, and streaming of machine signals for computerized maintena… | 511 |
| 2 | US20200301408A1 | Model predictive maintenance system with degradation impact model | 190 |
| 3 | US20030216888A1 | Predictive maintenance display system | 172 |
| 4 | WO2019216975A1 | Methods and systems for data collection, learning, and streaming of machine signals for analytics and mainten… | 169 |
| 5 | US20190339687A1 | Methods and systems for data collection, learning, and streaming of machine signals for analytics and mainten… | 154 |
| 6 | US20200150643A1 | Methods and systems for data collection, learning, and streaming of machine signals for analytics and mainten… | 145 |
| 7 | US20200133254A1 | Methods and systems for data collection, learning, and streaming of machine signals for part identification a… | 139 |
| 8 | US6735549B2 | Predictive maintenance display system | 138 |
| 9 | US20200150644A1 | Methods and systems for determining a normalized severity measure of an impact of vibration of a component of… | 133 |
| 10 | US20200150645A1 | Methods and systems for data collection, learning, and streaming of machine signals for analytics and mainten… | 129 |
Citation counts are drawn from the searched corpus only and favour older filings; treat them as a signal of influence rather than current commercial 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 citation and filing data signals
Reading citation counts and filing density together shows where influence concentrates and where the field is still open, though citation counts inside any searched corpus favour older records simply because they have had more time to be cited.
Data-pipeline patents anchor the field
The most-cited record in this dataset covers collecting, learning from and streaming machine signals via the industrial internet of things — a data-infrastructure claim, not a prediction-algorithm claim. Two related continuation and PCT filings from the same lineage carry 169 and 154 citations, showing the influence sits at the ingestion layer.
Growth has not levelled off
Filing volume has grown roughly thirteen-fold since 2017, with the midpoint year 2022 recording only 2 filings — nearly all activity in this dataset is recent. 2025 is the peak year so far at 106 records.
Control logic is claimed far more than sensing hardware
Control and regulating systems (G05B) and AI-based computing (G06N) together dominate the classification mix, while testing and structural-balance claims (G01M) sit at only 17 records — a sign that the sensing and calibration layer is comparatively open ground.
Filing activity is concentrated in one jurisdiction
India accounts for the largest share of receiving-office records at 226, more than four times the United States at 51, with Europe, WIPO, Australia and China each in single or low double digits.
Eureka can read the same corpus for gaps instead of for coverage: under-claimed branches adjacent to predictive maintenance for industrial equipment, with the prior art for and against each one.
Who is filing, and where activity is shifting
Recent-year momentum in this dataset skews toward academic and institutional filers rather than established industrial equipment makers, with the largest historical citation counts held by earlier commercial filers.
Muthayammal Engineering College leads recent volume
Muthayammal Engineering College (Autonomous) recorded 6 filings in the latest year, the highest recent-year count in this dataset, ahead of a cluster of other academic filers.
Aditya University doubled its filing pace
Aditya University moved to 4 filings in the latest year, a 100% year-on-year increase, placing it among the fastest-growing filers in this corpus even though its absolute volume remains modest.
Vellore Institute of Technology's filing pace has dropped sharply
Vellore Institute of Technology filed just 1 record in the latest year, a 75% decline year-on-year, a notable pullback after earlier activity.
Established filers hold the highest-cited records
The highest citation counts in this dataset belong to older commercial filings covering industrial-IoT data streaming and predictive maintenance display systems, rather than the academically-led filers driving recent-year volume.
| Assignee | Recent year | YoY |
|---|---|---|
| MUTHAYAMMAL ENG COLLEGE (AUTONOMOUS) | 6 | — |
| ADITYA UNIV | 4 | +100% |
| VELAMMAL ENG COLLEGE | 3 | +50% |
| VELLORE INSITUTE OF TECH | 1 | -75% |
| Johnson Controls Technology Company | 0 | — |
| Strong Force IoT Portfolio 2016, LLC | 0 | — |
| Siemens Industry, Inc. | 0 | — |
| International Business Machines Corporation (IBM) | 0 | — |
Where to go from this landscape
This page summarises filing trends, classification composition and the most-cited records in the dataset; deeper claim-level review is needed before any filing or freedom-to-operate decision.
Check claim scope on the data-ingestion layer
The most-cited records in this dataset cover machine-signal collection and streaming architectures, which sit underneath most prediction models built today. Review these claims before finalising a data pipeline design.
Search these filings in EurekaScreen the under-claimed sensing and calibration branch
G01M (testing and structural balance) shows far lower filing density than the control-systems and AI-model core, suggesting open ground in sensor drift correction and vibration feature extraction hardware.
Explore this branch in EurekaTrack academic filer momentum in India
Recent-year filing activity is concentrated among Indian academic institutions rather than established industrial players, a pattern worth monitoring for early signals of new technical approaches.
Monitor assignee activity in EurekaFrequently asked questions
The most-cited records in this dataset are industrial-IoT data-collection-and-streaming architectures, led by a family cited 511 times that covers collecting and streaming machine signals into a computerized maintenance management system. Closely related continuation and PCT filings from the same lineage carry citation counts of 169 and 154, which shows sustained influence over how later filers structure their own data pipelines. A separate high-citation family covers degradation-impact modelling layered on top of prediction output, and an older filing from 2003 on predictive maintenance display systems still carries 172 citations, showing that some foundational interface concepts remain referenced decades later.
Yes, and the growth is accelerating rather than levelling off. Filings moved from 6 in 2017 to 80 in the most recent year, with the midpoint year of 2022 sitting at only 2 filings — meaning nearly all the volume in this dataset has arrived very recently. The peak year so far is 2025 at 106 records; because publication typically lags filing by around 18 months, the 2026 figure will understate actual filing activity once later applications publish.
Testing and structural-balance claims (IPC class G01M) sit at only 17 records against 247 for control systems, despite vibration feature extraction being a named technical trigger in this field. Data recognition and presentation (G06K) and monitoring/checking devices (G07C) are similarly thin at 17 records each. This gap suggests the hardware-side calibration, drift-correction and sensor-validation layer is less contested than the software prediction and scheduling layer built on top of it, which is where most filing activity concentrates.
India leads receiving offices in this dataset with 226 records, well ahead of the United States at 51. Europe (EPO) and WIPO (PCT) filings sit at 11 and 10 respectively, with Australia and China trailing at 5 and 4. The concentration in India alongside a long tail of smaller academic and institutional filers suggests a large share of recent activity originates from academic and research institutions rather than established industrial equipment makers.
Start by mapping any planned architecture against the two claim clusters anchoring this corpus: industrial-IoT data-collection-and-streaming methods at the ingestion layer, and shared-backbone multi-task learning methods at the model-training layer. Because these occupy different layers of the stack, avoiding one does not automatically clear the other, so both need independent claim-scope review. Look specifically at whether your model architecture shares generic feature-extraction layers across multiple failure-mode outputs, and whether your data pipeline replicates the specific collection-and-streaming method in the most-cited family, since those are the structural elements most heavily claimed in this dataset.
Research Predictive Maintenance for Industrial Equipment in depth with Eureka
Go past this page: query the whole predictive maintenance for industrial equipment 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.