Predicting mechanical failure = the “L” in PLM
John Hayes posted on June 07, 2013 |

Field failure due to friction is a common and costly problem.  Estimates of the cost of friction, driven by unexpected or frequent downtime, maintenance, and repair, range anywhere from 1% of GDP upwards

The worst place for such issues to occur is in heavy equipment installations like oil & gas plants, renewable energy fields, mining operations, and aircraft.  

Asset managers try to control these costs through preventative maintenance. However, unless they know when components are going to fail, they will either overspend on maintenance or overspend on repairs and downtime. 

Bringing Predictability to Field Failures

I spoke with Ed Wagner, VP of Sales and Marketing for Sentient Science, a company of tribologists and materials scientists.  So far Sentient has received $23M in government funding to study the problem.  Beginning late in 2012, they began to engage commercial customers with a software solution they call “Digital Clone”.

As the name implies, Digital Clone is simulation technology that can predict when a component, assembly, or system is likely to fail.  By simulating this failure in the design stage of a complex assembly, design engineers can change a product’s geometry or material properties to extend its useful life.  And for machinery that’s already in the field, Digital Clone aims to help engineers plan their maintenance programs.

Sentient is holding a news conference on Thursday June 13th to introduce their commercial technology. will live-stream the event.

Designing for Extended Life

Once assets are in the field, the opportunity to extend their life is limited.  Sentient is developing technology that allows virtual testing in the design stage.  The idea is to allow design engineers to set goals for field performance and work backwards through design iterations to make more informed trade-off decisions up front, before assets are built and deployed. 

This chart shows the fatigue life of a critical bearing used in a wind turbine gearbox. Each curve represents the failure rate of a specific manufacturer’s bearing, showing that not all components will perform equally. 

Graphical output like this gives visibility as to when the bearings will fail. 

This particular analysis shows that the bearings the manufacturer had intended to last for 20 years were going to predictably fail in as few as 6 years. This difference had a material financial impact to the turbine OEM and wind farm stakeholders.

Extending Product Lifecycles After Deployment

Once in the field, there are still options for optimizing the performance and lifecycle of assets. Operators can control how the systems are used and maintained, but need to understand the effect of each decision they make.

The following chart demonstrates two set-point scenarios for wind turbine performance, both of which meet a minimum energy output and a desired service date.  The left chart shows the reality of the current design, while the right chart shows the possible extended useful life. By understanding the impact of such factors, operators can accurately quantify the performance and financial consequences of any changes made to the system.

 Prognostics vs Diagnostics

Most fielded equipment is currently monitored via various sensors that deliver data on key condition indicators, such as heat and vibration.  This diagnostic approach has a high relative cost because it advises only when an event has already occurred, leaving no time to react if that event is critical (such as component damage or failure). This can trigger an emergency response, such as unplanned maintenance or shutdown. 

Knowing when such an event will happen ahead of time (via prognostics) provides for longer lead times and allows maintenance to be scheduled in batches, or at more cost effective times.  

The Science of Material Failure

Much of Sentient’s funding came from the DoD, where the price of field failure is measured in lives as well as dollars.  The specific challenge DoD engineers wanted to address was to model the effects of friction on material performance at the component, assembly, and system levels.  The goal was to provide insights into why and when damage starts, which in turn helps predict when failures will occur.  

A number of leading scientists at Sentient began addressing this problem 10 years ago.  The issues are complex for a number of reasons, including the wide range of possible materials involved.  For example, metal on metal friction behaves differently than metal on ceramic.

The software solution

The result of Sentient’s research is a library of material properties and a series of algorithms that relate these properties to motion, stress, and reliability.  This allows Sentient to calculate the future useful life of a component or assembly. 

To deploy this knowledge as a solution, Sentient has created a cloud-based service that can analyze designs and field implementations.  The field monitoring solution collects operating data from sensors (fewer sensors than diagnostic systems) and updates the useful life estimates based on actual operating conditions. 

Someday, the simulation service may be available within standard CAD packages.  Already Altair has partnered with the company to bring these features to their user community.

Ed made the case that simulation capabilities like Sentient’s will be ubiquitous in the future, stating, “The next high-value phase of engineering data management will be “Prognostics” that allow design engineers to accurately predict failure rates during product design.”

Live event on Simulation

If you are curious about how Sentient’s simulations work in real-world applications, check out the live-streaming of their announcement on June 13.  

Sentient Science has paid a fee to to live stream the announcement.

Recommended For You