Maintenance records you already have: what they show before you buy sensors

Sensors, dashboards and "predictive" platforms are easy to pitch. Before you spend on any of them, it is worth asking a plainer question: what do the maintenance records we already keep tell us?

This article is for a maintenance or operations lead who has a work-order system, a spreadsheet or even a folder of paper forms, and wants to know whether that is enough for a first useful conclusion. In many plants it is more than people expect.

What you probably already have

Most sites keep some mix of the following:

• Work orders from a maintenance system or a shared spreadsheet.
• Repair notes written by technicians, often in free text.
• Spare-parts issues from the stores: which part went to which machine and when.
• Shift logs where operators note stoppages and short fixes.
• Service reports from outside contractors.

None of these was designed for analysis. They were written to get a job done and to leave a trace. That is fine, as long as you know what you are looking for.

Five columns that are enough to start

You do not need a clean database. For each record, try to get these five pieces of information into one table:

1. Asset - which machine, line or unit. Use one name for one asset.
2. Date - when the work was done or when the stoppage began.
3. Type of work - planned service, breakdown, adjustment, inspection.
4. Time lost - how long the asset was out of use, even as a rough estimate.
5. What was done - a short description or a part replaced.

If your system holds more, good. If it holds less, the first piece of work is to find where the missing information lives, for example in the stores log or the shift book, and add it by hand for a sample period.

What the records can show

With a table like this you can answer several practical questions without any new equipment.

Which assets take most of the lost time

Sort assets by total time lost over the last six to twelve months. In most operations a small number of assets account for a large share of stoppages. This tells you where any further effort, sensor or not, should go first.

Which repairs repeat

Look for the same asset with the same type of repair more than once in a short period. A bearing replaced three times in a year is a different problem from three unrelated faults. Repeat repairs point to a root cause that the repair itself does not remove: wrong part, wrong setting, a neighbouring component, or a process condition.

How long the gaps between failures are

For the assets at the top of your list, put the breakdown dates on a line. If the gaps are fairly regular, a planned replacement or inspection at a fixed interval may solve the problem without any measurement. If the gaps are irregular, you have a harder problem, and that is where a sensor might help.

Whether planned work is being done

Compare planned services that were scheduled with those recorded as done. A gap here is not a sensor question at all; it is a scheduling or capacity question, and no data feed will fix it.

Which spare parts move with which failures

Stores records often show a pattern that the work orders miss: the same part leaving the shelf for the same machine, month after month. It is a cheap way to confirm a suspicion.

Making use of free-text notes

Technicians' notes are the richest source and the hardest to use. You do not need software to start. Take one or two months of notes, read them, and add a simple tag to each: "wear", "leak", "electrical", "operator setting", "contamination", "unknown". Keep the list short.

After one or two months of tagging you will see whether most problems have a handful of causes. If "unknown" is the largest tag, that is itself a finding: the records are not good enough yet, and the first improvement is to ask technicians for one more line in each note.

Common problems with records

Expect these, and decide how you will handle them rather than trying to fix everything:

• Different names for the same asset. Agree one name and map the old ones.
• Time lost left blank. Estimate from the shift log for the sample, and make the field required from now on.
• Closed work orders with no description. Count them. If they are a large share, say so in your summary.
• Emergency fixes never recorded. Ask operators; their logs often show stoppages the system does not.

Do not wait for perfect data. A rough table with known gaps is better than no table, as long as the gaps are written down.

When sensors start to make sense

Sensors are worth considering when three things are true at once:

1. The failure repeats and costs real time or money. Your table shows it, not a supplier presentation.
2. Something measurable changes before the failure. Vibration, temperature, pressure, current or flow - ask your technicians and, where needed, the equipment maker whether there is a known warning sign.
3. You can act on a warning. Someone has to be able to plan the work, have the part available and get access to the machine. A warning nobody can act on has no value.

If one of these is missing, a sensor is premature. If all three are present, start with one asset and a small trial, and decide in advance what result would make you extend or stop it. Our article on a first data pilot for condition-based maintenance describes how to size that kind of trial.

A two-week plan

You can do the first pass in about two weeks with one person part-time:

• Days 1-3: collect the sources and list the assets that appear most often.
• Days 4-8: build the five-column table for the last six to twelve months, noting gaps.
• Days 9-11: tag a sample of free-text notes.
• Days 12-14: write a one-page summary: top assets by lost time, repeat repairs, gaps in the records and one recommended next step.

The next step may well be a small change in how records are written, a change in service interval, or a root-cause check on a repeat failure. Only if the table points to a repeating failure with a measurable warning sign does the sensor question become real. For how to present the outcome to management, see a maintenance report with three monthly numbers.

What to avoid

• Buying first and looking at records later. The records tell you which asset and which failure to measure.
• Promising savings from a first look. A first pass gives direction, not a business case. Treat any savings figure as a hypothesis until it is tested on a real asset.
• Cleaning everything. Clean only what the first questions need.
• Starting with a platform. A spreadsheet is enough for the first table. For the wider picture, see how predictive maintenance reduces downtime.

Summary

The maintenance records you already have can show where your equipment loses time, which repairs repeat and how regular your failures are. Five columns and two weeks are enough for a first look. If the table shows a repeating failure with a warning sign you can measure and act on, sensors have a clear job. If not, the money is better spent on the records, the service interval or the root cause.

Sources

• Our related articles linked above.
• The five-column table, the two-week plan and the three conditions for sensors are a working practice, not a standard; adapt them to your site and team. Examples are illustrative. No client results or market figures are claimed in this article.