Industries / Industrial equipment and services / Industrial services / Vibration analysis
Reporting
Reporting is judging whether a reading is acceptable, writing down what is wrong and how urgent it is, and showing the people who decide on repairs what the work is worth.
4
Milestones achieved, from 1939 to 2020.
1
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1939
A chart for rough running
T. C. Rathbone, chief engineer of the turbine and machinery division of the Fidelity and Casualty Company of New York, published "Vibration Tolerance" in the magazine Power Plant Engineering in 1939. It gave curves of vibration amplitude against frequency that link a level to a machine's condition and risk of failure, for machines running below 5,000 rpm. Downham and Woods wrote in 1971 that Rathbone's limits came from the "subjective opinions of several practical engineers and inspectors" checked with "relatively crude instruments". They add that the German engineers' association guideline VDI 2056 took over his curves for power and process machines on heavy, rigid foundations, drawn as straight lines and extended to 12,000 rpm. Istec, a company that sells vibration monitoring systems, calls the paper the foundation of modern industrial vibration measurement.
Credit
T. C. Rathbone
1949
Ships get their own limits
H. G. Yates set vibration limits for marine geared turbines from many tests on ships, published in 1949 by the North East Coast Institution of Engineers and Shipbuilders. Comparing the two in 1971, Downham and Woods found that a level at 22 Hz that Rathbone's curve called too rough was only slightly rough on Yates's. They put the difference down partly to ship machinery sitting in flexible steel shells rather than on massive foundations. They also note that Shell, like many other companies, wrote its own limits by adjusting published ones to its machines.
LessonA limit fits only machines built and mounted like the ones it was drawn from.
Credit
H. G. Yates
2010
Analyst exams accredited
The Vibration Institute, a nonprofit in Naperville, Illinois, certifies vibration analysts in four categories, I to IV, against a body of knowledge that follows ISO 18436, the international standard series on qualifying people who do machine condition monitoring. Its program was first accredited by the American National Standards Institute (ANSI) to ISO/IEC 17024, the standard for bodies that certify people, in March 2010. The accreditation now comes from ANAB, the ANSI National Accreditation Board.
Credit
Vibration Institute; ANSI
2020
What maintenance pays, surveyed
Douglas S. Thomas and Brian A. Weiss of the US National Institute of Standards and Technology (NIST) surveyed managers of machinery maintenance at US manufacturers and published the results in June 2020. Eighty-five responses came back, and some were dropped. On average, 17.3% of maintenance was predictive, 31.8% preventive and 45.7% reactive, meaning repairs after a breakdown. Among plants doing less than half reactive work, the half doing more predictive maintenance had 15% less downtime and an 87% lower defect rate. The authors note that such savings "may not be completely caused" by the maintenance method, since a plant may take several measures at once.
Credit
Douglas S. Thomas and Brian A. Weiss, NIST
Watch for
AI reports on computed evidence
Luigi Gianpio Di Maggio of the Politecnico di Torino published an open-source server built on the Model Context Protocol (MCP), a standard way for AI models to call outside tools. Through it a large language model runs FFT spectra, envelope analysis (which picks out the repeated knocks of a damaged bearing) and severity checks against ISO 20816-3, then writes a diagnostic report from the computed results, with a record of every step. The paper evaluates traceability, not diagnostic accuracy, and says its demonstration package "does not represent typical plant heterogeneity" and that its workflows are not "field-validated procedures".
Achieved whenan outside group runs it on recorded cases from working machines with confirmed faults and publishes how often its reports name the right fault and urgency.
Claimed by
Luigi Gianpio Di Maggio, March 2026
Open
01
Test reports on their readers
A reader test of vibration report layouts. Thirty or more maintenance planners and supervisors, not analysts, read the same twelve machine cases written three ways: a severity code with a short note, a plain sentence naming the action and the deadline, and a full page with spectra. Each reader says what they would do and by when, and the answers are scored against what the analyst meant.
The briefCredit · unclaimed02
Savings claims, recounted
An open ledger of 50 published savings claims from vibration programs, such as conference case studies, award entries and vendor papers. Each is recomputed with one stated method (repair and downtime cost avoided, less the program's cost) and graded by the evidence it gives, so that the gap between claimed and recomputed figures can be seen.
The briefCredit · unclaimed
Open 01 · the brief
Test reports on their readers
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A reader test of vibration report layouts. Thirty or more maintenance planners and supervisors, not analysts, read the same twelve machine cases written three ways: a severity code with a short note, a plain sentence naming the action and the deadline, and a full page with spectra. Each reader says what they would do and by when, and the answers are scored against what the analyst meant.
Why it is still open
Alan Friedman writes in Reliabilityweb's series on vibration programs that reports should carry severity levels whose meaning everyone agrees on, and that some planners who get a low-priority report schedule the repair at once "because they have not been trained to understand the meaning of the severity levels". A search in November 2026 found practitioner advice and vendor pages on report design, but no study measuring how readers act on different layouts.
Built with what exists
Cases written from published case studies and checked by a certified analyst; LimeSurvey, an open-source survey tool, to show the layouts in rotated order; Python or R for scoring.
First step
Write the twelve cases with one certified analyst and try them on five planners.
Done looks like
At least 30 readers each deciding all 12 cases, with the share of decisions that match the analyst's intended action and deadline given for each layout, and the cases and forms published.
Open 02 · the brief
Savings claims, recounted
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
An open ledger of 50 published savings claims from vibration programs, such as conference case studies, award entries and vendor papers. Each is recomputed with one stated method (repair and downtime cost avoided, less the program's cost) and graded by the evidence it gives, so that the gap between claimed and recomputed figures can be seen.
Why it is still open
NIST's 2020 report says that many economic assessments of predictive maintenance are "individual case studies, personal insights, and other anecdotal observations", so that its benefits "are largely unknown or are based on anecdotal observations". Reviews since, such as Khalid Lefrouni and Saoudi Taibi's 2025 review of 29 articles on AI for predictive maintenance, deal with methods rather than money. Sibeko's 2026 review in the Dutch Journal of Finance and Management sorts the financial measures used and the obstacles to measuring returns, from articles found through Google searches, without recomputing any claim. Searches in November 2026 found no review that recomputes published savings claims on one method.
Built with what exists
Crossref and OpenAlex to find case studies; Zotero to store them; a spreadsheet or Python with pandas; an AI model to pull figures from PDFs, with every number checked by hand against the page.
First step
Write the method on one page and apply it to ten case studies from one conference's proceedings.
Done looks like
50 claims, each with its link, the claimed saving, the recomputed saving, the costs the claim left out and a grade from 0 to 3, published with the method and the median ratio of claimed to recomputed savings.