Purge Canister Flow Prediction for Stable Vehicle Fuel Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current vehicle systems rely on inaccurate delay models for estimating purge flow, leading to vehicular stalls and increased evaporative emissions, as they struggle to predict purge vapor characteristics effectively.
Innovation Solution
Implementing a neural network, specifically a convolutional neural network (CNN) or hybrid deep CNN with recurrent neural network (RNN), to classify inputs from vehicle sensors and predict purge flow, enabling proactive adjustments in injector fueling to manage purge vapor fluctuations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a delay model is used to estimate purge flow, then the system complexity is low, but the measurement precision of purge flow estimation is poor
Solution Approach 1:
The patent replaces the traditional mechanical delay model with a neural network-based predictive system. The neural network processes multiple sensor inputs (mass airflow, manifold pressure, engine speed, purge valve duty cycle) to predict purge vapor concentration and timing, substituting simple time-delay calculations with an intelligent system that learns complex purge flow patterns, thereby significantly improving estimation accuracy while accepting increased computational complexity.
Solution Approach 2:
The patent introduces a neural network as an intermediary between sensor inputs and fuel control decisions. This intermediary processes and interprets multiple sensor signals to predict purge vapor characteristics, acting as a bridge that translates raw sensor data into actionable fuel compensation commands, thereby improving the overall system's measurement and control precision.
2Reliability
If purge command is lowered to reduce vehicular stalls, then drivability improves, but evaporative emissions increase
Solution Approach 1:
The patent implements preliminary action by using the neural network to predict purge vapor concentration and timing before the vapor actually enters the intake system. This advance prediction allows the fuel control system to proactively adjust injector timing and quantity to compensate for upcoming purge vapor, preventing drivability issues before they occur while maintaining optimal evaporative emissions control.
Solution Approach 2:
The patent establishes a feedback loop where the neural network continuously monitors sensor inputs, predicts purge vapor characteristics, and the fuel control system adjusts injector commands based on these predictions. This closed-loop feedback mechanism enables real-time optimization of fuel delivery to counteract purge vapor effects, simultaneously improving drivability stability and reducing evaporative emissions.
3Measurement precision
If traditional delay model is used for purge prediction, then response time is fast, but prediction accuracy is poor
Solution Approach 1:
The patent applies preliminary action by pre-training the neural network offline with extensive sensor data to learn purge flow patterns. During actual vehicle operation, the pre-trained network requires minimal computational resources to process sensor inputs and generate predictions, thereby achieving high prediction accuracy without significant real-time computational delay.
Solution Approach 2:
The patent implements dynamics by using a neural network that can adapt to varying operating conditions. The system dynamically adjusts its predictions based on real-time sensor inputs (mass airflow, manifold pressure, engine speed, purge valve duty cycle), allowing it to maintain high accuracy across different driving scenarios while optimizing computational efficiency for each specific condition.
Data Source
AI summary
In various embodiments, methods, systems, and vehicle apparatuses are provided. In one exemplary embodiment, a method is provided that includes obtaining a set of inputs, by a processor, pertaining to one or more features that are used to predict the purge flow of a purge canister system of an intake system of a vehicle; obtaining data, by the processor, from sensors about the vehicle's intake system for use by a neural network to enable the processor to classify the set of inputs including the one or more features for purge flow control for use in predicting a presence of purge content in the vehicle's intake system; and obtaining, by the processor, an output from the neural network wherein the output is configured as a binary or continuous output to instruct a vehicle controller to execute an action to fueling control by letting fueling controller choose different gain sets and adaption strategy based on the binary output flag in a case of the binary-output model, or apply an adjustment factor to fueling command in case of a continuous model.