Robot Dogs, Drones and Smart Sensors: Meet the New Maintenance Team

In Part 3 of our predictive maintenance series, we looked at how machines provide early signs of deterioration. Changes in vibration, temperature, acoustics and other condition data can help Maintenance identify a developing problem before it becomes a failure.

Traditionally, collecting this information has meant either installing sensors directly on equipment or sending technicians to inspect assets and take measurements.

But there is now another option: bring the sensor to the machine.

Across industrial sites, mobile robots and drones are increasingly being used to perform inspections, capture condition data and reach equipment that is difficult or potentially hazardous for people to access.

Some fly. Some walk.

And yes, some look remarkably like dogs.

Why Put Sensors on a Robot?

The idea is relatively simple.

A factory may contain thousands of motors, pumps, bearings, gearboxes and other components. Installing permanent monitoring equipment on every asset is not always practical or economically justified.

Manual inspection rounds provide another option, but they require skilled employees to move through the facility, take measurements and document what they observe.

A mobile inspection robot sits somewhere between these two approaches.

Rather than permanently installing a thermal camera, acoustic sensor or visual inspection system on every individual asset, the same sensors can move through the facility and collect information from multiple inspection points.

Boston Dynamics, for example, positions its four-legged Spot robot as a mobile inspection platform that can perform visual, thermal and acoustic inspections along predefined routes. The robot can repeatedly return to the same equipment and collect comparable measurements over time.

That repeatability is important. The objective is not simply to take an interesting picture of a machine. It is to create useful condition data that Maintenance can compare, analyse and act upon.

Robot Dogs Have Already Entered the Factory

The image of a four-legged robot walking through a production facility may still look futuristic, but there are already practical industrial deployments.

At Nestlé Purina, Spot has been used to automate thermal and acoustic inspections on packaging lines. According to the company's maintenance team, the aim is to identify issues between production runs so that repairs can be planned earlier rather than waiting for a reactive intervention.

AB InBev is using the same technology at its brewery in Leuven, Belgium. Spot performs thermal inspections to identify hotspots and acoustic inspections to detect compressed-air leaks. In one documented example, the robot identified a faulty transport motor that could otherwise have caused at least six hours of packaging-line downtime.

Other manufacturers are taking a similar approach with different robotic platforms. At Grace's chemical plant in Düren, Germany, ANYbotics' four-legged ANYmal robot has been used for autonomous thermal, visual and acoustic inspections. The robot performs more than 120 inspection points per mission, helping the site identify anomalies and plan maintenance interventions.

These examples illustrate an important shift. Robots are not replacing the maintenance decision. They are increasingly automating part of the data collection that comes before it.

Sometimes Walking Isn't Enough

Factories do not only contain equipment that is difficult to monitor frequently. They also contain assets that are difficult to reach at all.

Think of silos, ducts, stacks, cooling towers or confined spaces. Inspecting them may require scaffolding, rope access, specialised safety procedures or a shutdown.

This is where drones can change the inspection process.

At cement plants, for example, specialised drones are being used to inspect kilns, silos, chimneys and ductwork. They can collect visual information, LiDAR data and, with specialised payloads, ultrasonic thickness measurements without requiring inspectors to access the same areas physically.

In one cement-plant case, an inspection drone was used inside a chimney after a structural problem was suspected. The resulting 3D model helped the maintenance team locate the damage and plan the repair before the work began.

The value here is not simply that drones are faster or more impressive than traditional inspections. It is that previously difficult-to-access condition information can become easier to collect.

And once information is easier to collect, it may also become practical to collect it more frequently.

More Data Is Not the Goal

There is an obvious temptation with all of this technology: more sensors, more robots and more inspection data must mean better maintenance.

But as we saw in Part 3, data collection is only the beginning.

A robot can take a thermal image of a motor. A drone can identify a crack. An acoustic sensor can detect an unusual sound. What matters is whether that information helps the organisation decide what to do next.

The interesting development is therefore not the robot itself, but the increasingly automated chain around it.

A robot can follow the same inspection route at regular intervals, collect measurements from the same assets and flag anomalies for further investigation. Maintenance teams can then focus their attention on equipment where the data indicates that something has changed.

At Swiss cement producer Vigier Ciment, for example, an ANYmal robot performs nightly inspection rounds across three mills and six floors. According to ANYbotics, it completed more than 33,000 inspections across 450 inspection points during its first 16 months in operation. The resulting inspection report is available to the maintenance team in the morning and is used to help plan repairs, spare-parts orders and site visits.

That is perhaps a better way to understand the role of autonomous inspection.

The objective is not to remove Maintenance from the process. It is to reduce the amount of routine data collection required before Maintenance can make a decision.

From Autonomous Inspection to Autonomous Action?

This brings us to a much bigger question.

Today, a robot can walk through a factory, inspect a motor, detect an unusual temperature and flag the anomaly to Maintenance.

What happens if the systems around it become more connected?

Imagine that condition monitoring indicates that a bearing is deteriorating. The asset is automatically identified. The expected intervention window is calculated. The corresponding spare part is known. Current inventory is checked across the production network, and the organisation can determine whether the component is already available or needs to be sourced.

We are not suggesting that every one of these decisions should happen without human involvement. Maintenance interventions involve engineering judgement, operational priorities, safety considerations and business decisions that cannot simply be reduced to an automated workflow.

But the direction is interesting.

The technologies required to detect deterioration, collect condition data and automate inspections are advancing quickly. At the same time, manufacturers are improving the systems that manage maintenance, inventory and procurement information.

The more these systems can exchange reliable information, the shorter the distance becomes between detecting a problem and being ready to solve it.

The New Maintenance Team Is Still a Team

Robot dogs make good photographs. Drones flying through industrial facilities make impressive videos.

But the most interesting change is less visible.

Condition monitoring no longer has to depend entirely on permanently installed sensors or people manually moving from machine to machine. Mobile inspection technologies can extend the reach and frequency of data collection while keeping Maintenance focused on interpreting that information and deciding what action is required.

The maintenance team of the future may therefore include technicians, reliability engineers, sensors, AI models, drones and autonomous robots.

They do not all perform the same job.

They contribute different pieces of information to the same objective: understanding what is happening to the equipment early enough to do something about it.

And once manufacturers become better at predicting what will need attention and when, another question becomes increasingly important:

Can spare parts planning become predictive too?

Next in the series: From Predictive Maintenance to Predictive Spare Parts Planning.

Missed the other parts of the series?

• Part 1 – From Maintenance Schedules to AI: We looked at how predictive maintenance evolved from historical reliability methods and condition monitoring to AI-supported prediction.
• Part 2 – Predictive Maintenance Gives You 30 Days. Now What?: We explored what needs to happen after an early warning, from identifying the correct spare part to checking availability and current supplier lead times.
• Part 3 – How Does a Machine Know It’s About to Fail?: We examined how sensors, condition data and AI can detect changes in machine behaviour and turn them into useful maintenance information.
• Part 5 – From Predictive Maintenance to Predictive Spare Parts Planning: We explore how condition data and expected maintenance demand could help make spare parts planning more forward-looking. (to be published soon)
• Part 6 – The Autonomous Factory: Can a Machine Order Its Own Spare Part? We bring the series together and ask how far the process from detecting deterioration to identifying, locating and procuring the right spare part could eventually be automated. (to be published soon).