Senseye (client registration required) deployed its equipment prognostics and maintenance management software platform at a major auto manufacturing plant in the U.K. in January 2016. After completing a pilot period, the customer has moved into a first paid commercial phase. Senseye is ramping up to analyze machine health for 12,000 machines across three production lines, including over a thousand robots. The customer contracted major industrial suppliers to instrument machines with sensors and to pull the data into an industrial cloud. Senseye then handles the analytics portion using its IoT-based solution. Senseye claims that the customer previously reviewed analytics offerings from major analytics players, including Microsoft and SAP, but that none offered the level of automation, return-on-investment (ROI), or prognostics that it required.
Senseye has developed generic failure and degradation models for machines used in manufacturing operations – including bearings, motors, pumps, actuators, conveyers, and gearboxes. The firm claims it can predict failures weeks or months in advance, compared with incumbent predictive maintenance solutions, many of which are costly and only offer basic predictive capabilities. The Senseye approach is based on prognostics, which is able to analyze more subtle patterns that have previously required years of machine observations to develop (which Senseye claims to bring to the table for any machine).
For example, a complex failure cycle for a bearing might take the following form: first, the system will pick up a small variation in the vibration harmonics. Second, the vibration will return to normal and then there will be a slight increase in power consumption. Third, power consumption will return to normal and then there will be a slight change in torque. The torque levels then will return to normal and the machine will appear to be functioning properly, although an important event has transpired that offers insight into the health of the bearing, which may indeed be sustaining damage. A major part of Senseye’s intellectual property (IP) is to apply custom machine learning to this pattern and identify it as an early failure mode. It combines this with an understanding of bearing’s kinetic and physical properties to then give a prediction of the remaining time until failure.
Senseye typically plugs into more advanced manufacturing operations that already have deployed sensors and data historians in their environments. Much of the data comes from real-time sensor feeds for measurements like vibrations, torque, temperature, and power consumption. It also gathers data from historians, which adds operational information, like which specific process the machine is running at a given time. Senseye is a cloud-based solution that normally leverages an on-site internet connection. Senseye hosts the solution in Amazon Web Services (AWS) and provides a user interface (UI) in one of two forms: the platform has a native UI that smaller organizations typically use; for larger organizations, it typically pushes insights into existing IT infrastructure, such as ERPs or maintenance systems.
Senseye explained that an ongoing challenge is determining a scalable pricing model suitable for different industry sectors. The firm typically charges monthly software-as-a-service (SaaS) fees based on the number of connected machines and their complexities: a simple machine with just a few sensors may cost as little as $150 per month, while a complex machine with a large array of real-time sensors may cost over $2,000 per month. This model works well for many large organizations, but for customers with tens of thousands of machines, the model requires volume discounting. Senseye also explored value-based pricing, but concluded that model would be difficult to implement.
Another challenge is validating performance of the solution at very large scale. Senseye is a relatively new solution (less than two years old) and the firm is still evolving the software. While the solution has proven robust and scalable in its projects thus far, taking on a customer of very large scale incurs certain risks.
Wait and See – while the project has moved into the first phase of scale commercial deployment, it is still early, and return on investment (ROI) figures have yet to be calculated. Senseye explained that since the project began, it has provided automatic insights about dozens of potential failures with an accuracy of over 85%. Without these alerts, several of these issues would have turned into machine failures stopping the whole line and leading to unplanned downtime. The standard industry figure is that every hour of unplanned downtime costs roughly $2 million.
The overall objective for this new solution is for the customer to reduce unplanned downtime by 50%, and Senseye believes this is possible. An additional goal is to increase overall equipment effectiveness (OEE), which is typically calculated as the product of availability, performance, and quality. Industry calculations indicate that every percentage point of OEE translates to a couple percentage points of overall profitability. Lux will keep tabs on this deployment and await more detailed ROI calculations as stakeholders amass more information on its value.
By: Isaac Brown