Real-Time Engine Emission Prediction with Neural Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for engine emission detection, such as computational fluid dynamics simulation and gas analyzer measurements, are inaccurate, time-consuming, and costly, and fail to reliably predict transient emissions.
Innovation Solution
A real-time prediction method using neural networks (NNs) trained with historical test data, optimized by a mind evolutionary algorithm (MEA) and combined with an Adaboost algorithm, to predict engine emissions based on measurable engine parameters like rotation speed and fuel consumption, without simulating physical reactions or using expensive instruments.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If computational fluid dynamics simulation is used to predict engine emission, then emission data can be obtained, but the calculation is time-consuming and accuracy is dominated by grid quality
Solution Approach 1:
The patent pre-trains neural network models offline using CFD simulation data and experimental data to establish a predictive model. Once trained, the model can rapidly predict emissions without requiring real-time CFD calculations, thus resolving the contradiction between accuracy and calculation time.
Solution Approach 2:
The patent creates a virtual copy of the complex CFD simulation process through neural network models that replicate emission prediction functionality. These trained models serve as simplified copies that can predict emissions rapidly without requiring the computational resources and time of actual CFD simulations.
2Measurement precision
If gas analyzer instruments are used to measure engine emission, then accurate emission values can be obtained within limited ranges, but the cost of purchase and maintenance is expensive and the instruments are greatly affected by environmental factors
Solution Approach 1:
The patent replaces expensive, complex gas analyzer instruments with computationally lightweight neural network models that run on standard computing hardware. This substitution eliminates the need for costly instruments and their supporting devices while maintaining prediction capabilities across various operating conditions.
Solution Approach 2:
The patent replaces the physical measurement system (gas analyzers and supporting devices) with an information processing system (neural network models). This substitution transitions from mechanical/physical measurement to computational prediction, eliminating environmental sensitivity and high costs associated with physical instruments.
3Reliability
If gas analyzer instruments are used to measure engine emission, then emission values can be obtained, but the instruments cannot desirably reflect performance of transient emission of the engine
Solution Approach 1:
The patent pre-trains neural network models with extensive training data covering various transient operating conditions. This preliminary training enables the models to accurately predict transient emissions in real-time without being affected by environmental factors during actual measurement.
Data Source
AI summary
A real-time prediction method for an engine emission is provided, including: acquiring multiple known historical test data samples for engine emission, dividing the samples into a training set and a test set to train multiple neural network (NN), calculating mean square errors (MSEs) output by each of the NNs with the different numbers of hidden layer nodes to determine a topological structure of the NNs, optimizing initial weights and initial thresholds for each of the NNs with a mind evolutionary algorithm (MEA), and establishing a real-time engine emission prediction system with an Adaboost algorithm.