In manufacturing, drilling, farming and other industries, equipment is instrumented with multiple sensors that collect extensive data and measure a variety of metrics including pressure, temperature, and vibration. Many of these sensors are used in a reactive manner. When a value gets outside of an expected range, or a leading predictor of failure is detected, the machine shuts down. However, the bigger opportunity lies in shifting to the broader lens of business optimization rather than just avoiding a particular failure. Business optimization opportunities include equipment maintenance, equipment and part sales, field service engineering deployments, global supply chain optimization, risk management and more.
The challenge in embracing this approach is that sensor data tends to live in its own silo, disconnected both physically and organizationally from other valuable data such as operations data, operating environment, cost information, employee log files, incident documentation etc. However, all this context is pertinent and is relevant in more accurately predicting failures and breakdowns as well as the opportunities and costs of those failures. These silos of data and the complexity of developing and linking various knowledge models make it difficult for the business to examine and anticipate how to foresee and address an equipment failure before it can have impact.
Operationalizing Predicting Equipment Failure the Knowledge Graph Way
Knowledge Models developed on Maana tackle this complexity in a systematic and repeatable way thereby accelerating the time to value. Maana’s approach is to develop several related and overlapping models that connect to many dozens of value decisions including prioritization of maintenance, prioritization of sales opportunities, optimization of field service engineering deployment and many more. Algorithms employed include feature extraction from time series data for failure events and operating context, and then random forests, gradient boosted trees, and generalized linear regression to predict failure events and their duration based on sources as wide as SAP, Salesforce, Operations Log files, and Incident Reports PDFs. These models are then linked to related business optimization models to compute opportunities and cost models of those failures prioritizing actions that accelerate profitability. Enterprises that hinge on the operation of equipment need a comprehensive view of their business to accelerate gaining insights to make reliable and informed decisions. Maana’s knowledge graph does just that — leading to better decision making and ultimately, action. Instead of planning responses based on just the value of a sensor at a single instance or the value of a sensor over a short period of time, business leaders now get a holistic view of their equipment gaining insight not only from sensors, but also relevant data related to the context of the asset operations and environment from other sources.
Maana’s Knowledge Platform accelerates the time it takes to get from data silos to operationalizing solutions based on predicted failures. How is this possible? At the core of the platform is Maana’s patented knowledge graph that captures the complex relationships between the business and physical concepts across organizations. Models can be constructed by domain experts, data scientists and data analysts collaboratively and interactively. These models are then operationalized by solution engineers in line of business or custom applications for the larger organization to use in the context of high value decisions. And the results of the actions based on these solutions lead to the iterative adaptations of the deployed models yielding the promise of cognitive computing.
General Electric is one company that knows well the difference this approach can offer. The company uses Maana’s knowledge platform in several divisions in order to optimize assets and processes. It has estimated that even 1% reduction in equipment failure in the power industry over 15 years equates to $66 billion dollars in additional revenue that would have been lost due to production downtime. And by using a solution like Maana to reduce the average 5% in a factory’s production downtime to 4%, that 1% improvement represents $6.5 billion dollars in annual savings for the company. With sensor data applied along with predictive analytics, we can start to see the promise of Internet of Things (IoT) for a wide variety of industries. It certainly indicates that the fourth industrial revolution (Industry 4.0) will lead to efficiently anticipating and prescribing actions that improve manufacturing automation and operational excellence.
Stay in the know with the latest information about Maana services, events, news and best practices by email.