Predictive maintenance could reduce maintenance costs 10-40%, reduce downtime by 50%, and lower equipment and capital investment by 3-5% by extending machine life – McKinsey[1]
In business, few people like surprises. Wall Street can punish a company’s stock price if the numbers are not in line with estimates. Shocks in supplier or customer behavior can derail the best strategic plans. In manufacturing and other industries that depend on reliable machine performance, there may not be anything more disruptive than unexpected breakdowns. They can create expensive, emergency actions like rush delivery, overtime, or acting under dangerous conditions to restore normal operations.
What would it mean to your business if you could predict the probability of failures before they happen, with enough notice to prevent them? Predictive maintenance solutions offer the ability to assess the performance of in-service equipment, identify patterns preceding failure, and automatically deliver alerts to responsible parties. This allows organizations to save time and money with advanced warning signs about maintenance needs, and preemptive equipment repair.
Predictive maintenance modeling techniques also offer a buffet of interesting predictions to choose from, customized to your needs. For example, if it usually takes 2 days to schedule a repair crew once a failure occurs, you might configure your solution to alert you of a pending breakdown 3 or 4 days in advance to achieve zero downtime. Even better, you can predict not only the probability of failure, but also which root cause (i.e. particular component) is most likely within a future time period. This helps manage both inventory and crew costs. Pretty cool!
Here’s a glimpse of some of the expected business value of predictive maintenance:
- Proactively manage maintenance needs, rather than waiting to react to the problem until equipment fails. Reduce unscheduled downtime, waste, and rework.
- Decrease supply chain costs by optimizing inventory. Reduce holding and shipping costs for spare parts and rush orders.
- Better utilize and optimize your existing assets and resources – people, technology and equipment – to avoid costly disruptions in operations.
As the old saying goes, “If this was easy, everyone would be doing it.” Predictive maintenance is a modern solution that performs best with modern data and analytics tools. A key aspect of effective predictive maintenance solutions is combining connected/IoT-enabled devices with advanced analytics to translate data from physical assets to insights from predictive models. And the more data, the better: failure and error data, general machine and operator information, and repair history all contribute to more accurate predictions. Given these considerations, here are some of the key challenges to a useful predictive maintenance solution:
- Business-focused objectives and data: predicting an outcome for which you can take action; if you only have data that captures which day a failure occurred and it’s important to predict which hour it will occur, you have a problem
- Complexity and diversity of data: combining at-rest (e.g. maintenance history) and in-motion (e.g. failures and errors from telemetry) data into a single source connected to an analysis system
- Data and analytic pipelines: storing, scheduling, moving, and creating useful data sets for analysis
- Data science: considering multiple operational factors simultaneously for predictions, not just static thresholds of single measures
- Delivering insights: bringing the right information to the right person at the right time, especially with real-time streams
 |
 |
 |
What’s the best way to deal with these challenges? I discussed some solution options in a recent webinar, but here’s a summary. Assuming you have clearly defined business objectives, use a modern data platform that’s equipped to handle streaming and Big Data sets and is agile when it comes to moving that data around, combining it, and using it in machine learning models. With the Microsoft Azure IoT Suite and Cortana Analytics Suite, you can connect and monitor your devices and analyze their data in real time. These analytic technologies, along with data platform tools like Azure Data Lake, work seamlessly together in both development and production to construct a predictive maintenance workflow, regardless of data size or complexity. You’ll also want a data scientist who is familiar with the particulars of predictive maintenance modeling. Microsoft has provided an excellent primer here. Finally, a dashboard tool like Power BI provides interactive reporting of predictive results to the people who can take action.
As a Microsoft partner, BlueGranite is well-positioned to deliver predictive maintenance solutions using the Azure platform. Get started quickly with a predictive maintenance preconfigured solution during a 3- to 4-week proof-of-concept engagement. For more details and to learn more, check out our Predictive Maintenance offer page.
References:
- McKinsey, “The Internet of Things: Mapping the Value Beyond the Hype”
- Microsoft, “Creating Business Value with Predictive Maintenance.”
