No need to look into a crystal ball: Predictive maintenance
How and in what time period does the position of a camshaft change? Under what conditions and how quickly do valves coke? When does a component need to be replaced before it gives up the ghost? Such forecasts do not require a glimpse into the crystal ball, but rather predictive maintenance. IAV is one of the technology leaders in predictive maintenance in the automotive sector.
Acting instead of reacting
Predictive maintenance means acting instead of reacting: Sensors record parameters such as vibrations or temperatures. Algorithms analyze the measured data in real time and provide information on whether and when maintenance or repair is required. “With AI, we expand the possibilities of predictive maintenance. For example, self-learning algorithms detect hidden patterns in the data streams and recognize complex cause-and-effect relationships,” says Nabert. The advantages: Even before damage impairs the function of a machine, it can be repaired in a targeted manner, thereby minimizing downtime. Preventive maintenance at fixed intervals is also no longer necessary, which reduces costs.
These methods enable more than just maintenance: “They also enable us to determine, predict and optimize the current state of a system – whether vehicles, engines, test benches or other machines. We call this predictive health monitoring,” says Nabert. That means: With the help of calculations during operation, the systems can work optimally.
Neural networks and algorithms
IAV has been gathering experience in predictive maintenance and predictive health monitoring for years. In one application, artificial neural networks (auto-encoders) and counterfactual analysis were used to detect errors in the environment of engine control units and to precisely estimate the size of the error, for example in the trimming of the camshaft. IAV used the methods for robust monitoring of component wear and its prediction for valve coking, for example.
It is not only the automotive industry that relies on Predictive Health Monitoring: Christian Nabert’s IAV team is working on machine monitoring with IoT for an electrical wholesaler in southern Germany. “The dealer wants to know when to offer their customers which spare parts, for example for a cable cutting machine,” says the specialist, outlining the task. To do this, sensors were attached to the machines and corresponding models were developed. This enables the wholesaler to recommend spare parts to their customers in good time and score points with additional service. They also optimize their warehousing and save costs.
The article was published in automotion 03/2020, the automotive engineering magazine of IAV. Here you can order the automotion free of charge.
«We develop functions to ensure that a vehicle is always in optimum condition so that, for example, the requirements for emissions, durability and performance are met with maximum reliability at the same time»
— Specialist in Predictive Health Monitoring at IAV
Predictive status models are used to predict errors and system failures as well as to improve system performance. System monitoring makes it possible to maintain vehicles, test benches and other machines in line with requirements by forecasting changes in individual components – for example due to wear and tear. In addition, the methods and information from the engine control units can be used to monitor essential vehicle parameters. “We develop functions to ensure that a vehicle is always in optimum condition so that, for example, the requirements for emissions, durability and performance are met with maximum reliability at the same time,” reports Nabert. IAV uses these methods extensively from combustion engines to electrified drives, as well as for data plausibility checks and monitoring of test benches.
