Analyzing the combustion characteristics, engine performance, and emissions pathways of the Internal Combustion (IC) engine requires management of complex and an increasing quantity of data. With this in mind, effective management to deliver increased knowledge from these data over shorter timescales is a priority for development engineers.
At cmcl innovations we have combined conventional engine research methods with the latest developments in process informatics and statistical analysis. Process informatics enables engineers to combine data, instrumental and application models to carry out automated model development including optimisation and validation against large data repositories of experimental data.
These developments are complemented with the inclusion of experimental error and model parameter uncertainty, to yield confidence regimes on the final model result, hence the impact of specific shortcomings of the model and/or experimental dataset can be identified in a systematic manner.
THE RESULTS
A methodology to couple models and experiments has been developed
The data model, engineML enables engineers and their modelling tools to have easy and reliable access to experimental data obtained from a variety of sources formats and structures.
A model has been optimised with respect to experimental data
In the example below, forty-two model parameters have been optimised with respect to numerous engine operating points.
Model uncertainty propagation
Experimental and model uncertainties have been included, in this case preventing model over fitting to experimental data of low certainty, and highlighting those aspects of the model where model parameter uncertainty is limiting the overall model performance.
A model has been optimised with respect to experimental data
Now that the model can be re-optimised as soon as new data comes online, our tool can identify the next most-effective experiment to carry out for intelligent design of experiments.
|
