Vehicle Emissions Prediction Using Latent Driving Cycle Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for diagnosing internal combustion engine emissions are limited by requiring specific load states on a test bench and long diagnostic times, making it cumbersome to predict emissions during practical driving operations, especially with varying driving cycles.
Innovation Solution
A method using a machine learning system to generate realistic time curves of operating variables by transforming measured data into latent variables, allowing for the prediction of emissions across different driving cycles, including the use of autoencoders and Gaussian process models to extract essential features and reduce dimensionality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If emissions are measured during practical driving operations with multiple driving cycles, then measurement precision and reliability are improved, but loss of time and productivity deteriorate
Solution Approach 1:
The patent applies preliminary action by training a machine learning model in advance using emission measurements from multiple different driving cycles. Once trained, the model can rapidly predict emissions for new driving cycles without requiring actual measurement campaigns, thus saving time while maintaining measurement precision through the model's ability to generalize across different driving conditions.
2Adaptability or versatility
If a machine learning system is trained to generate time curves of operating variables, then adaptability to different driving cycles is improved, but device complexity increases
Solution Approach 1:
The patent uses copying by training a machine learning model to learn and replicate the complex relationships between operating variables and emissions across multiple driving cycles. The model creates a virtual copy of the emission measurement process that can generate realistic time curves for any driving cycle, providing adaptability without requiring physical measurement equipment for each cycle.
Solution Approach 2:
The patent applies parameter changes by using the machine learning model to transform input parameters (operating variables from different driving cycles) into output predictions (emissions). The model learns the underlying patterns and adjusts its internal parameters during training to accurately map diverse driving cycle characteristics to corresponding emission profiles, enabling versatile prediction across varying conditions.
3Device complexity
If only test bench measurements are used for emissions prediction, then device complexity is reduced, but measurement precision and reliability deteriorate
Solution Approach 1:
The patent introduces an intermediary approach by using a machine learning model as a bridge between test bench measurements and real-world driving emissions prediction. The model is trained on controlled test bench data but learns to generalize to uncontrolled real-world conditions, acting as an intermediary that translates simplified measurements into accurate real-driving emission predictions without requiring complex measurement systems.
Data Source
AI summary
A method for ascertaining emissions of a motor vehicle driven with the aid of an internal combustion engine in a practical driving operation. A machine learning system is trained to generate time curves of the operating variables with the aid of measured time curves of operating variables of the motor vehicle and/or of the internal combustion engine, and to then ascertain the emissions as a function of these generated time curves.