Embedded virtual sensors
We have been selected to create an algorithm and a methodology allowing the deployment of «virtual sensors» - algorithms able to estimate the value of a measure based on other sensors.
Using specific sensors only present on flight test aircrafts (e.g gauges on aircraft structure), we trained Machine Learning algorithms to infer their values based on standard sensors – sensors embedded on any production aircraft. We validated it on a large amount of data including various flight conditions and transitory changes. We have been able to create a matrix of 50 virtual sensors, which provides valuable information on the long-term fatigue and usage.
2% of relative uncertainty
compared to real data on physical sensors