Industries / Industrial equipment and services / Industrial machinery / Industrial filtration
Monitoring and change-out
This work decides when a filter is spent, damaged or leaking, by measuring it in place or by testing it before and after use.
6
Milestones achieved, from 1970s to 2022.
2
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1970s
One dust calibrates counters
Automatic particle counters grade how clean hydraulic and lubricating oil is by counting the particles in set size bands. From the early 1970s they were calibrated with AC Fine Test Dust (ACFTD), made in batches by the AC Spark Plug Division of General Motors, against a size distribution worked out by microscope in the late 1960s on particular batches. ISO's later review says nobody measured how much the batches differed, and that these differences became a significant source of variability in particle counts.
Lessona reference dust whose batches were never compared carried their differences into every count made with it.
Credit
AC Spark Plug Division, General Motors; ISO 4402
1997
Broken bags found by charge
A baghouse is a bank of fabric filter bags. A triboelectric monitor is a metal probe in its clean exhaust that picks up a tiny current when particles strike it, and the current rises when a bag tears. In September 1997 the US Environmental Protection Agency (EPA) published guidance, prepared by Midwest Research Institute, on setting up such monitors as bag leak detectors. US rules now cite it: secondary lead smelters must run a bag leak detection system that its maker certifies can detect 1.0 milligram of dust per actual cubic meter of exhaust, with an alarm.
Lessonthe guidance warns that the signal grows with the square of gas velocity and changes with particle size, charge and humidity, so the rule asks only for relative dust readings, not a measured mass.
Credit
US EPA Emission Measurement Center; Midwest Research Institute
1999
Counts tied to NIST
In 1992 the AC Rochester Division of General Motors stopped making ACFTD. In 1999 ISO replaced the old calibration with ISO 11171, a calibration traceable to the US National Institute of Standards and Technology (NIST), and rewrote the ISO 4406 cleanliness code and the multi-pass filter test (ISO 16889) to match. Sizes measured the new way are written µm(c), and ISO's technical report says sizes once called 2, 5, 10 and 15 µm become about 4.6, 6.4, 9.8 and 13.6 µm(c).
Lessonthe industry rebuilt its references only when the one company making its test dust stopped.
Credit
ISO; NIST
1999
Nuclear filters tested twice
A HEPA filter (high-efficiency particulate air filter) for US nuclear sites must let through no more than 0.03% of particles 0.3 µm across. In its Technical Report 23 of 1999, the Defense Nuclear Facilities Safety Board said that although makers routinely test their filters, the Department of Energy's Filter Test Facility still rejected 3 to 6% of those supplied, and the Department kept independent testing for safety-related filters. Reviewing the facility's results from 2013 to August 2020, Southwest Research Institute found about 9% of filters rejected for labels and paperwork, about 7% for defects seen on inspection and about 1% for failing the performance test.
Lessona maker's pass did not mean the filter would pass an independent check.
Credit
Defense Nuclear Facilities Safety Board; US Department of Energy; Osvaldo Pensado, George Adams and Imad Khalek of Southwest Research Institute
2017
A ten-year limit tested
The Department of Energy's Nuclear Air Cleaning Handbook says HEPA filters may not be used beyond ten years from manufacture, or five where they may get wet, while in normal use they are changed when the pressure drop across them reaches 4 to 6 inches of water. The ASME AG-1 code committee, which writes the nuclear air-cleaning code, rejected adding a service life appendix. Mississippi State University's Institute for Clean Energy Technology tested 11 aged filters from commercial power plants, supplied by EPRI, in 2016 and reported in March 2017 that the study gave only limited insight: too few filters, and no loading to failure.
Lessonthe report says the calendar limit is expensive and has left very few filters held in reserve, yet ageing is unlikely to show up in the yearly in-place leak test.
Credit
Institute for Clean Energy Technology, Mississippi State University; DOE Nuclear Safety Research and Development program
2022
Filters tested before use
A sterilising-grade filter is the last barrier against microbes in an injectable drug. The European Commission's revised Annex 1 of the EU rules for making medicines, dated 22 August 2022 and in force from 25 August 2023, says the filter assembly should pass an integrity test after it is sterilised and before use (pre-use post-sterilisation integrity test, or PUPSIT), as well as after use. Named tests include the bubble point, the gas pressure that pushes the wetting liquid out of the largest pores, and diffusive flow. Where a test before use is not possible, a documented risk assessment may replace it.
Lessona test after use shows the filter was whole at the end of the run, and the engineering firm CRB notes that product or process conditions can mask a flaw, so a test after sterilisation catches damage done in preparation.
Credit
European Commission
Watch for
Longer sterile filter runs
Donaldson says its Alpha-Pore membrane cartridges run up to 2.5 times longer than conventional polyethersulfone (PES) cartridges under controlled test conditions, so plants change filters less often.
Achieved whena user or independent laboratory publishes run times for Alpha-Pore and PES cartridges on the same fluid, with integrity test and microbial retention results.
Claimed by
Donaldson, November 2026
Pulsing only when needed
Pulse-jet collectors clean their bags with blasts of compressed air, often on fixed timers. Emerson says that pairing real-time dust readings with adaptive control means filters are cleaned only when needed, cutting compressed-air use and extending the life of filters and pulse valves, but gives no figures.
Achieved whena plant publishes compressed-air use, bag life and emissions before and after the change, measured by someone other than the supplier.
Claimed by
Emerson, in Processing Magazine, July 2025
Open
01
A public bag leak test
A bench dust collector with fabric bags into which holes of known size are cut, run with a harmless, non-combustible test powder such as calcium carbonate. On the clean side, log a commercial triboelectric bag leak detector beside two low-cost optical particle sensors, and weigh dust collected on sampling filters as the reference. Report the smallest leak each device detects and how fast it alarms, in milligrams per cubic meter, against the 1.0 mg per actual cubic meter that US rules ask leak detectors to reach.
The briefCredit · unclaimed02
An open dust collector log
A public, labeled set of operating logs from at least 10 working dust collectors over a year: pressure drop across the filters, pulse valve firing, compressed-air pressure, airflow and dust trend, with every bag or cartridge change and every leak marked. Add a baseline model that predicts the next change-out and publish its error.
The briefCredit · unclaimed
Open 01 · the brief
A public bag leak test
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A bench dust collector with fabric bags into which holes of known size are cut, run with a harmless, non-combustible test powder such as calcium carbonate. On the clean side, log a commercial triboelectric bag leak detector beside two low-cost optical particle sensors, and weigh dust collected on sampling filters as the reference. Report the smallest leak each device detects and how fast it alarms, in milligrams per cubic meter, against the 1.0 mg per actual cubic meter that US rules ask leak detectors to reach.
Why it is still open
Peter Bächler's 2024 doctoral work at the Karlsruhe Institute of Technology used low-cost sensors (Sensirion SPS30 and Alphasense OPC-N3) to locate leaking bags in test rigs and baghouses, but found that leak peaks pushed the sensors past their rated limits, such as 3,000 particles per cubic centimeter for the Sensirion sensor. The federal figure is certified by detector makers, and no independent side-by-side test of cheap sensors and triboelectric detectors was found.
Built with what exists
A small fabric-bag dust collector; calcium carbonate powder; Sensirion SPS30 and Alphasense OPC-N3 sensors; a commercial triboelectric probe; a Raspberry Pi logger; a sensor board that a model such as Claude Fable 5.2 can help design (see the Note "Three working devices").
First step
Build the rig and log a week of readings with intact bags to set each device's baseline.
Done looks like
Five hole sizes, 20 runs each for every device, with detection threshold and time to alarm per device, raw logs and rig drawings published in an open repository.
Open 02 · the brief
An open dust collector log
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A public, labeled set of operating logs from at least 10 working dust collectors over a year: pressure drop across the filters, pulse valve firing, compressed-air pressure, airflow and dust trend, with every bag or cartridge change and every leak marked. Add a baseline model that predicts the next change-out and publish its error.
Why it is still open
Makers' connected services, such as Donaldson's iCue, measure pressure drop and alert users as filters approach the end of their usable life, but show the data only in each customer's dashboard. Searches of research papers, Zenodo, Kaggle and GitHub found vendor pages and patents but no public collector data set.
Built with what exists
Exports from plant historians or IO-Link pressure sensors; Python with pandas; LightGBM for the baseline model; a script that strips plant names and locations.
First step
Write the data schema and ask three plants that already log their collectors for one month of export.
Done looks like
Ten or more collectors, 12 months each, at least 30 labeled change-outs or failures, and a model scored on collectors it never saw, with its error stated in days.