Using AI in predictive maintenance to forecast the future

As the cost of capital may continue to rise and macroeconomic trends could signal challenges ahead, enterprise budgets may be pressured on all sides. Existing infrastructure could be asked to do more while supply chain disruption might persist. To address these challenges, organisations could try to extract the most bottom-line value from existing assets and investments. An opportunity may exist in better managing physical assets to help minimise downtime, maximise productivity, and save as much capital as possible in labour, materials and expenditures.

Preventative maintenance, an approach to asset management, may no longer be enough in today’s globalised marketplace. Strategies such as scheduled maintenance checks and conditioned-based maintenance may not always be enough to confirm asset reliability in a fast-paced, complex environment for manufacturing, logistics and operations. Whether the concern is cascading damage to the wider system, the quality of products, the safety of the process and facility, or other consequences from ageing or failing assets, it may be important to build the capacity to help predict asset failure and help prevent it from occurring in the first place.

To better understand what is at stake and the options available, consider the many business-critical advantages that could flow out of predictive maintenance.

The multi-dimensional value in predictive maintenance

Industrial automation is growing quickly with the development of IoT technologies, reduced cost of data storage/computing, and advancements in AI/ML capabilities. Yet, based on our experience, maintenance organisations have not been able to harness the power of these technologies beyond pilots.

Maintaining assets at its base may be a task of limiting and avoiding downtime, and driving efficiency in maintenance could help to maximise asset usage and keep operations moving. This applies to assets within facilities, like manufacturing plants, as well as assets in the field. Yet, the business value in effective maintenance may be bigger than asset uptime.

Business value of effective maintenance

Predictive maintenance could support these outcomes. The challenge may come in knowing which technologies and process changes can support predictive maintenance operations across the business.

Using AI to help forecast the future

Historically, maintenance schedules may have been either based around estimates of a machine’s lifetime and projected time to failure or based on recommendations from the original equipment manufacturer. To help improve maintenance operations, the enterprise can replace educated guesses with data-based knowledge about how an asset is performing and when it will degrade.

Getting to this level of predictive maintenance begins with incorporating additional data sources. Sensors can be added to key components to capture data points about how the asset is working. Other data sources that can help unlock value include procurement and enterprise resource planning (ERP) data, historical maintenance and repair data, production data, and ongoing reports from employees in the field.

When the data is consolidated and interpreted with AI-enabled signal processing, the result can be a deeper and more nuanced understanding of not just individual machines but the larger network of interdependent assets. By harnessing the collective knowledge of people, sensors and systems, the business can use AI to analyse the information and output maintenance recommendations. These recommendations can be automatically prioritised, which may help to optimise how the human workforce allocates its time. In a way, the AI solution could serve as an omnipresent maintenance employee helping the human workforce make better decisions about when and where to target operations.

As an example, Deloitte worked with a major logistics provider that was struggling with conveyance equipment in its distribution centres. By adding sensors onto the assets and pulling data from every facility into a cloud environment, the company was able to use analytics to identify the lifespan of equipment across the facility network and target maintenance interventions before a failure. The result was faster and more efficient operations, which translated into greater competitiveness in the marketplace. 

Bridging the gap to predictive maintenance

The value may be evident and the enabling technologies known, but shifting the enterprise from reactive maintenance to proactive and predictive operations can be a complex endeavour. A transformational project requires specialised skills and knowledge to architect the systems, design the sensor strategies, and create data pipelines from the edge into the cloud. It also might require machine learning engineers and data scientists to prepare and train the predictive models, as well as a workforce that is trained to make sense of the signals and push those predictive insights into a case management system or workflow.

Each of these tasks takes specialised skillsets and leading practices rooted in experience, some of which may not be available in house. Working with an organisation like Deloitte can help expedite the process of orchestrating the data collection, inference engine and action engine, as well as provide the change management, documentation, and training needed to help the human workforce adopt and use predictive maintenance technologies.

To be sure, every business is different, just as every asset is different. A leading constellation of solutions and platforms could be somewhat different between enterprises, and a starting point is assessing maintenance maturity to identify where new data flows and AI analysis can begin improving operations. Transforming to predictive maintenance may not be an all-or-nothing proposition. Some organisations may test new capabilities in a pilot programme. With evidence of how predictive maintenance can help to improve quality, safety, workforce productivity, and asset uptime, the business could build momentum to seek other use cases, scale the programme, and promote enterprise-wide enthusiasm and adoption.

In this era of humans working with intelligent machines, the race may be on to find value-driving AI applications that can help to create competitive differentiators in the market. We may not need AI to predict the future of maintenance operations. The task may now be to make the targeted investments and decisions that can help to make it real.