HomeVirtual Sensor Technology for Aero-Jet Engine

Virtual Sensor Technology For Aero-Jet Engine


Implementing Virtual Sensor Technology For Aero-Jet Engine Test Environments

As industrial systems have become progressively more complex, risk analysis and environmental safety issues have become paramount. Advancements in sensors coupled with powerful cloud-based AI platforms, have facilitated the automation of mission critical industrial environments. Aero-engines are complex mechanical systems that provide the power component of aircraft propulsion systems. The development, and optimization of aero-engines require significant investment in developmental and testing costs. Sensor based monitoring of engine components and data analytics can be deployed in tracking component degradation and facilitate the automation of diagnostics of mechanical subsystems.

Deployment of these automation technologies in Engine Test Environments can significantly reduce the costs of engine development, while at the same time improve the performance and reliability of the developed systems. However, the deployment of sensors in these sub-systems is challenging since the aero-engines work at harsh operational conditions, including high temperatures, pressures, and speeds and are composed of complex mechanical (sub)systems that multiple causes for failure. Locating physical sensors at some key locations might not be feasible due to spatial or environmental conditions.

Physical sensors are subject to a number of limitations including noisy signals , degradation over time, cost or the impracticality of collecting the physical parameters across the entire system complexity. Software-based virtual sensors are an abstraction layer composed as a proxy of sensor hardware, aggregating input from the physical sensors. Virtual sensors are used to overcome the problems of physical sensors, offering lower operating cost and increased reliability, and even providing indirect measurement of non-measurable properties. IVA has developed virtual sensors for Aero-Engine Test Environments that handle the reliability of different component subsystems.

These virtual sensors are being deployed by our industrial partner as critical analytical components to their test environment. This paper describes the strategy for this deployment, employing IVA’s Graph-Machine Learning platform, DataCivet. The use of DataCivet for this application has shortened the development time of the Virtual Sensor by several orders of magnitude, while ensuring that the Machine Learning based system performs within the exacting operational parameters set by our industrial partner. The next section describes the literature studies, approach, and Section III, the results and future work planned.


The world is now in the juncture of automation and machine learning. With the right data, artificial intelligence can create intelligent expert systems that lead to enterprise success. So we utilize our product Datacivet where the data is pre-processed with minimal or no human intervention which can be used for further predictions and deployment of outcomes. Data acquisition is the primary step in developing a machine learning model. A customization to our proprietary tool, Datacivet helps the data scientists to make it faster to save the cost and time. With a provision of a pool of algorithms, numerous algorithms can be selected.

An automation of complete data science life cycle from data preparation stage with noise detection and removal, high pass filtration of signals and feature extraction, model creation to the final deployment stage, Datacivet enables the organization to focus on the results of AI/ML applications rather than the problem of process. This cloud-based ML Platform allow users to input sensor data to train an ML model. It makes it easier to build and use ML models in the real world by running systematic processes on raw sensor data and selecting models that pull the relevant information from sensor data.


It is clear that human resource and time can be saved for knowledge exploration in future if a proper, consistent data repository is maintained. The sequence of operations involved in this approach will allow us to make predictions about the performance of different parts of the jet engine, using patterns extracted from the data. We can arrive on a machine learning model with better accuracy with deeper understanding of the data and multiple levels of training. Through the application of machine learning techniques, virtual sensors that can continuously monitor the performance of Aero jet engines, even in the presence of deployed sensor failures.

In contrast to previous approaches, the proposed general methodology at its end-state will be a fully automated data-driven approach incorporating Datacivet that will accommodate the entire data life cycle from data ingestion to virtual sensor generation. Whether automating the complete process to make it more efficient, or scaling it up, or getting valuable insights, or increasing productivity, or reducing expenditure, DataCivet has been developed to allow people with limited technological knowledge, and minimal previous statistical experience, to build a Machine Learning model that fits the needs of a customer.

Get in Touch

International Virtual Assistance
Pvt Ltd. B-10, Park Center Building,
Phase I, Technopark Campus,
TVM 695 582.

Ph: +91 6238 642 100,
8089 070 851.
[email protected]
83 Morgan Street,

Ph: +1646 750 4241
Opole, Poland.

Ph: +48 690 425 600.

© 2021 Copyright: IVA