While no one can see the future, the right data can signal what lies ahead well in advance. That is increasingly true in maintenance: With enough data on your equipment, you can plan servicing and repairs before there are costly shutdowns or failures. Embed enough sensors into your equipment, the thinking goes, and you can create a "prediction possibility zone"1 that can anticipate failures in your most critical systems, saving you time--and money.
According to ReliabilityWeb,2 an effective predictive maintenance program can produce a savings of eight to 12 percent over preventive maintenance. Writing in 2016, Manufacturing Business Technology predicted3 that predictive maintenance could help companies save $630 billion over the next 15 years.
In predictive maintenance, you analyze data gathered from your equipment for patterns that can help extend its life. Machines, after all, don't just stop working for no reason. As one problem leads to another, data is being generated. Diving into that data can help you to understand what is going on within a machine or device, and see a potential problem before it slows down work. With the appropriate data analysis, you can determine when a maintenance team should work on a certain part or machinery. Remember, downtime means no production, so the more you can see into the future, the better.
Here are the key components in working with predictive analytics:
- Collect the data
- Analyze the data
- Act on or learn from the data analysis
- Review the progress and gains
- Repeat the data collection
The first key component is collecting the data. Gathering real-time data from equipment can help forecast potential issues and bring the Internet of Things (IoT) to the factory floor. Just as in a home IoT environment, where a thermostat or security system can be controlled remotely by computer or smartphone, the industrial IoT gives equipment the ability to connect and exchange data with other equipment--and share that data with operators and maintenance staff. Workers conducting maintenance know the machinery well, but more diagnostics can help them to truly get into what may fail systemically within it. For example, sensors installed in heavy machinery can notice certain peaks of heat, implying a future crack in a part, or can gather data on the speed of ball bearings. Collecting the data by sensors is much more accurate and timely than by manually tracking. With sensors, you can have the data captured and uploaded in real-time for immediate analysis. Just take care to not collect more data than you might actually need: The business research firm Gartner has said that 70% of the manufacturing data4 collected today is never used.
Keep in mind, you'll need an organized approach to sift through all the data you do collect. Have an expert available who can look through, analyze and determine how to use the data findings. It can be someone on your team, or through your software partners or vendors. They can look at past failures based on previous data, if available, analyze current data, and develop a model for the future.
Establish regular analysis of incoming data, to ensure the data is being recorded and analyzed. The term "dashboard" comes to mind, though it can be in whatever format you recommend or prefer in reporting out to the different levels of the company, from on the floor to leadership and management team members. It all comes down to identifying actionable information and then acting on this data to make a difference, such as reduced lag times and higher productivity.
Here's how things played out at a polyester manufacturer. Its fiber-spinning machinery runs very hot, which means that if it must be serviced, the machines must first be shut down for 24 hours. Failing to catch problems is costly: It is 10 times more expensive to rebuild spinning equipment after a failure than before. The business solutions firm that worked with the manufacturer 5 helped it to identify patterns of usage and when to best conduct maintenance. The result was fewer failures and improved run times, which led to a 2.7 percent increase in gross production, for a $7 million annualized return.
The company's reaction to all that was probably not hard to predict.
1. "Predictive Maintenance: Big Data on Rails." DataConomy. Bhoopathi Rapolu. April 2015.
2. "Unleash the Power of Predictive Analytics! Can Your Machine Tell You When it Will Fail?" ReliabilityWeb. Mario Montag.
3. "Revolutionizing Manufacturing with Predictive Maintenance Analytics." Manufacturing Business Technology. Sundeep Sanghavi. August 2016.
4. "My New Year Wish – Less Hype For Big Data Analytics, More Buzz For Smart Manufacturing." Manufacturing Operations Management.
5. "Predictive Analytics Eases Manufacturing Equipment Maintenance." BlueGranite. Brian Carlson. April 2016.
"Five Ways To Minimize Manufacturing Downtime." Manufacturing.net. Tom Bonine. January 2013.
"How Predictive Maintenance Fits Into Industry 4.0." Engineering.com. Jennifer Roubaud. October 2017.
"My Maintenance Program Cost Me!!" IBM. Dan Barrett. July 2013.
"What Everyone Must Know About Industry 4.0." Forbes. Bernard Marr. June 2016.