In past, failures were analyzed after the fault/failure had already occurred. This resulted in unambiguity and non-determination of the precise parameters, factors and conditions that caused the failure(s). Over a period, with maturity in processes, it was identified that the failure rate can be reduced or minimized by performing periodic maintenance checks; however, the real failure cause and conditions were still not identified.
With advancement in semiconductor industry, high performance micro-controllers and miniature size memories were developed allowing product manufacturers to implement and experiment on additional features such as Built In Test Equipment (BITE), periodic self-checks within software, implementing fault isolation and storing the parameters and conditions (fault snapshot data) during failure. This data helped in diagnostics of product, improving design, and reducing failure rate.
In recent years, with advancement in data modelling, simulation, data processing algorithms, data analytics, communication, networking and cloud storage technologies, the data received from the product in conjunction with product life cycle data, is fed to product model/Simulation and various diagnostics and prognostics algorithms. Test scenarios based on field data received, are exercised, to identify deviation from normal pattern. New behavior is predicted to see impact on aircraft safety, failures and remaining useful life, trends can be plotted for critical parameters and can be used in providing further insight into failures, by implementing Prognostic and Health Management (PHM) framework.
Some of the advantages of implementing PHM in the aerospace industry include:
There are two different approaches available to assess the degradation or extent of deviation from expected performance to assess the reliability of system and to predict the remaining useful life of the system:
This approach uses monitored or historical data to learn the system behavior and to perform prognostics. Data driven approach do not require system models or specific system knowledge to start the prognostics and are suitable for complex systems as their behavior cannot be assessed and derived from first principles. For example, in aerospace products manufactured at production line undergoes test to check functional and parameter test. Based on the data collected from each test, mean, mode, median of the data and trends can be determined, correlation analysis can be performed to determine the impact of one variable on remaining variables. These analyses will help in predicting future state of product.
A model driven approach uses system knowledge to develop system model and based on the conditions exercised on model the system behavior is determined and data is collected. In this approach, since conditions and system are simulated, various worst-case scenarios can be exercised in order to understand the system behavior that sometime might be difficult on real system. For example, Digital Twin which is a virtual representation that serves as the real-time digital counterpart of a physical object or process.