Discover How Vehicle Maintenance is Transformed Through Machine Learning
According to the American Transportation Research Institute (ATRI), maintenance costs for heavy-haul vehicles in North America have increased more than 50 percent over the past seven years, and maintenance costs now represent nearly 10 percent of the operational cost per mile. ¹
In a previous blog post, we noted that the data elements collected from the Trimble customer base generate more than 10 billion data points a day. These data points include a wide range of vehicle-centric information, including weather and road conditions as well as driving behaviors like vehicle speeds and acceleration rates. As we noted, each of these data elements can be transformed through machine learning, which is a technology that uses computer algorithms to detect patterns in large data sets and predict outcomes.
Another crucial aspect of this data collection is related directly to engine performance. This wealth of engine data can be transformed by machine learning, detecting anomalies in engine and vehicle health. Trimble has built an algorithm that uses this data to detect when a semi-truck will derate over the next one to three days. This insight can allow fleet managers to route a truck to maintenance before it fails, getting the load transferred to a different truck and preventing a driver from being stranded on the side of the road waiting for an expensive repair.
How machine learning is applied to fleet maintenance
The data utilized to make predictive maintenance models are captured from the onboard device connected to the engine’s electronic control module (ECM). This integration provides a wealth of data, including more than 50 signal variables, such as engine temperature, turbo speed, oil pressure, velocity and coolant levels.
Applying machine learning to this data, Trimble is able to identify trucks not hitting the “performance” mark compared to the normal operating conditions for its peers. Trimble specifically built a model to identify vehicles that are likely to breakdown due to a fault. Prior to feeding this data into the machine learning algorithms, data scientists “clean” and prepare it to ensure the predictions and outcomes are accurate.
Machine learning analyzes the massive amount of data collected from the truck’s ECM, and feeds it into TMT’s Predict.Fault Code application to show the truck’s health score along with the indicators identifying a potential fault. Fleets can act on this information to bring a truck in for maintenance, prior to an unexpected, costly and inefficient roadside breakdown.
Benefits of proactive maintenance
Prior to predictive modeling, vehicle maintenance typically fell into three distinct categories: do nothing, run a vehicle until a component fails and then repair it or plan and schedule routine maintenance to prevent equipment failure. With predictive maintenance, you can reliably predict faults before they occur, based on data anomalies instead of a more arbitrary maintenance schedule. By relying on data-based decisions, you can identify and prioritize trucks that require immediate service.
Predictive maintenance gives fleets information to fix vehicle components before they break.
The benefits of predictive maintenance don’t stop there. Preventing 35 percent of unplanned repairs can save a fleet an estimated $490 per vehicle every year in towing, labor, parts and lost margin. Forecasting when a breakdown is looming not only keeps vehicles up and running but also reduces the likelihood of road breakdown events, helping ensure your drivers stay safe and off the side of the road.
Curious to learn how Trimble is applying machine learning to revolutionize the maintenance process? Check out our free machine learning whitepaper or contact us today to learn how we can help your fleet reach new levels of vehicle performance and efficiency.