By Bryan Christiansen
Predictive Analytics are being used all over today, from business intelligence to fraud detection, to even law enforcement. As manufacturing has begun to embrace the ethos of IoT and Industry 4.0, predictive analytics have crept into all areas of the shop floor. This may include areas such as process optimization, supply chain, and even the maintenance department.
Maintenance may seem like an odd place for analytics at first glance, but it is an area ripe for advanced uses of data. Breakdowns cause downtime, affecting production and logistics. They also cost money – in fact, downtime is typically one of the biggest costs in manufacturing. Many report the cost of downtime at $100,000 to $300,000 per hour, with a smaller percentage of businesses reporting the cost in excess of $1 million per hour.
So as manufacturing becomes more connected and automated, it is moving from a traditional reactive maintenance approach to become predictive and even prescriptive in practice.
From Reactive to Predictive
A traditional maintenance approach goes something like this:
- Equipment breaks down
- Operations notifies maintenance
- Maintenance mobilizes and repairs the equipment
- Production resumes
The repair portion can take varied amount of time, especially if replacement parts are needed and not available. This adds to downtime and/or idle time. Traditional maintenance is reactive, disruptive, and inefficient. But with modern data collection and analytics tools, businesses can move to a Predictive Maintenance (PdM) strategy.
With a PdM approach, real-time machine data is used to accurately predict when a breakdown may occur. Data is collected via sensors (temperature, vibration, pressure, etc.) and fed into a software package – typically summarized in a dashboard. The maintenance staff can monitor the health of the machines this way and respond to real-time issues before they become catastrophic breakdowns.
When you can predict a problem a week out, you can plan appropriately with minimal effect to production and very little cost to the operation. In this way, maintenance efficiency is maximized. The numbers support this – 10%-40% reduction in maintenance costs, 10-20% reduction in waste and 10-50% new improvement opportunities uncovered.
Starting a Predictive Maintenance Program
The general framework for starting a PdM program might contain the following steps:
- Select Critical Assets
- Review Asset History & Root Cause Failure Modes
- Develop a Predictive Model
- Start Pilot on Asset
- Adjust & Replicate
Most likely, you already know what your most important assets are. Determine an asset that you want to start a pilot program on. This is the machine that causes much pain when it has problems, and/or the one that you need to be running most often.
Review Asset History & Root Cause Failure Modes
Next, review the maintenance history of the asset and understand the failure modes for equipment breakdown. For example, is it a bearing failure? Is it a problem of motor overheating?
Develop a Predictive Model
With the failure modes understood, you can start to develop a model for prediction. This could be a theoretical model (using engineering equations) or empirical model using real data. You may need to install sensors for vibration or temperature to track this data and find the line where action is needed.
Deploy by tracking data on the asset, using the model. Ensure you have clear ownership of the program so that someone will act when needed. You will have to determine levels at which actions are taken. Once the pilot shows benefit, start to replicate across the plant.
For a more detailed breakdown, you can refer to this guide on how to start a predictive maintenance program.
The next step with predictive analytics in maintenance is Prescriptive Maintenance (RxM). Where PdM may predict when a machine will break down, RxM will prescribe specific actions relevant to a machine based on its history and current situation. For example, instead of PdM informing you to make a repair soon, RxM will prescribe the actual steps needed to make the repair.
RxM accomplishes this through more advanced digital techniques – such as Machine Learning and Artificial Intelligence.
Prescriptive Maintenance also goes further than simply suggesting when to make a repair – it can help determine the most efficient way of running a machine or help to extend the time between repair events. As maintenance becomes more data driven, RxM is the natural progression after PdM.
Maintenance has not been left out of the digital transformation. Using improved data collection and analysis tools, new strategies have emerged, changing maintenance forever.
By enacting a PdM program you will start to maximize efficiency of your maintenance team. This will lead to lower maintenance costs, along with other costs that you incur during a schedule disruption. Through modern technology you can start to reap the benefits of a predictive maintenance program today.
About the Author
Bryan Christiansen is the founder and CEO at Limble CMMS. Limble is a modern, easy to use mobile CMMS software that takes the stress and chaos out of maintenance by helping managers organize, automate, and streamline their maintenance operations.