By Ivan Romanow
By Ivan Romanow
Manufacturers need to adjust their processes and understand analytics. They need to be empowered to better understand what’s going on in real time by measuring the right things and understanding the data and how to link cause and effect.
According to the Association for Manufacturing Technology, approximately 95 per cent of manufacturing machines are not monitored. Given that manufacturers purchase machine tools over time to match contracts, it’s no wonder that a plant may contain machines with up to 30 generations of eight different controls, all running closed proprietary protocols. This is a manufacturing automation pre-condition that severely limits the ability of these machines to communicate at all.
Putting accurate information into the right hands at a manufacturing facility in a timely manner can result in:
• Increased visibility and information from the plant floor;
• Reduced product recall/warranty cost and repair times;
• The discovery of hidden capacity in your equipment;
• Improved uptimes;
• Reduced costs;
• Improved profitability;
• Improved quality;
• Reduced time to market;
• Paperless production reporting; and
• The ability to make profitable decisions.
Often, data collection focuses too much on the collection and parsing of data, and too little on how it can be used to make better and more effective decisions in specific areas of a business. The process is simply to collect large amounts of data with the hope that something useful will emerge from analysing it at some point in the future.
Start with the end in mind
Going forward, manufacturers seeking to extract maximum value from investing in big data projects should work backwards and ask some fundamental questions:
1. What business processes or decisions do you want to improve? (Make sure to get management buy-in at this level and alignment across the organization.)
2. How will these decisions improve the business? Customer profitability? Product rationalization? Capacity planning?
3. What are you trying to maximize? Profits? Asset utilization? ROI? Revenues?
4. What are the most meaningful metrics to measure progress towards those goals? Unit margins? Profit-per-hour of machine time? OEE (efficiency of the machines, utilization, part counts, performance)?
5. What types of analysis do you need to perform to expose the data, explore “what if” scenarios and iterate through alternatives to maximize profitability?
6. And then, finally, what types of data do you need to collect to feed the above analysis and decision-making?
Some additional advice includes:
• Start small. Pick a machine to run a pilot test on.
• Choose downtime reason codes that affect your process.
• Collect data for several days.
• Select and report data for baseline benchmark.
• Dissect the data so it is meaningful to you, and propose a solution using lean techniques.
• Implement proposed changes to your process from your analysis.
• Re-run the machine and check before and after data.
• Extrapolate and estimate the effort involved to apply the same techniques to all of the machines on the plant floor.
Data projects will only be successful when teams first focus on the desired end result. Using that as a filter for the data collection and aggregation will lead to success, while only focusing on massive data collection will lead to confusion, wasted time and resources, and missed deadlines.
Using the processes above will enable the utilization of the right data in a useful and meaningful way. Without such a solution, data collection is nothing more than meaningless, empty noise.
Ivan Romanow is CET Director of Sales with Gescan Ontario, a division of Sonepar Canada.