Machine Learning CO Emission Prediction Using Exhaust Sensors
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for measuring carbon monoxide (CO) emissions from vehicles are costly and require technical expertise, making them impractical for fleet-scale monitoring, and do not account for the effects of aftertreatment systems, resulting in inaccurate emission predictions.
Innovation Solution
A system using exhaust emission sensors to input data into a machine learning model trained on vehicle operating parameters and PEMS measurements, allowing for real-time prediction of CO concentrations and masses without additional sensors or expert oversight, enabling continuous monitoring and adjustment of emissions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If PEMS systems are used to measure CO emissions during on-road operation, then measurement accuracy and real-world applicability are improved, but system cost and operational complexity increase significantly
Solution Approach 1:
The patent creates a virtual copy of the PEMS measurement capability through a machine learning model that replicates CO emission measurement functions using data from existing vehicle sensors. The model is trained on PEMS data to learn the relationship between sensor readings and actual CO emissions, then uses this learned knowledge to predict CO emissions without requiring physical PEMS hardware, thereby achieving measurement accuracy while eliminating system complexity
Solution Approach 2:
The patent replaces the physical PEMS measurement system with a computational machine learning model. Instead of using complex hardware sensors and measurement equipment, the system uses software-based predictions derived from standard vehicle sensors (NOx, NH3, O2 sensors and ECU parameters), substituting mechanical measurement infrastructure with information processing
2Productivity
If PEMS systems are deployed for fleet-scale monitoring, then CO emission data collection is improved, but cost and requirement for technical expertise increase
Solution Approach 1:
The patent makes the CO emission measurement system universal by using existing multi-functional vehicle sensors (NOx, NH3, O2 sensors and ECU parameters) that already serve other engine control functions. The machine learning model processes data from these universal sensors to provide CO emission measurements, allowing the same sensor infrastructure to serve multiple purposes without requiring dedicated CO measurement hardware for each vehicle
Solution Approach 2:
The system enables vehicles to self-measure their CO emissions using their own existing sensors and onboard processing capabilities. The machine learning model runs on vehicle controllers, allowing each vehicle to independently generate its own emission data without requiring external measurement infrastructure, technical expertise for operation, or additional specialized hardware
3Device complexity
If machine learning models predict engine out emissions without aftertreatment effects, then computational simplicity is maintained, but emission prediction accuracy deteriorates
Solution Approach 1:
The patent incorporates feedback from downstream sensors (NOx, NH3, O2 sensors positioned after the aftertreatment system) into the machine learning model. The model uses these downstream measurements along with ECU parameters to infer the actual CO emissions that exit the aftertreatment system. This feedback mechanism allows the model to account for aftertreatment effects and achieve accurate predictions of real-world emissions without requiring overly complex physical chemistry simulations
Data Source
AI summary
Methods and systems are provided for a vehicle. In one example, a method may include acquiring measurements from at least one exhaust emission sensor of the vehicle, the exhaust emission sensor positioned to measure one of NOx, NH3, and O2 levels in exhaust gas of the vehicle. The measurements may be input into a machine learning model trained to output a predicted real-time amount of at least one exhaust gas constituent in the exhaust gas and operations of an emissions aftertreatment system may be assessed and adjusted based on the predicted real-time amount of the at least one exhaust gas constituent.