Transient Local State Space Models for NOx Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing NOx estimation models for internal combustion engines fail to accurately account for transient conditions, leading to inaccuracies in NOx emission measurements, which can propagate errors in reductant dosing, fault detection, and catalyst management within aftertreatment systems.
Innovation Solution
The implementation of discrete transient local state space models to modify steady state NOx estimation models by subdividing the engine operating space into discrete local state spaces and developing transient models for each, providing a NOx estimation correction value to improve the accuracy of final NOx estimates, which are then used to control engine and aftertreatment systems or diagnose sensor malfunctions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If steady state NOx estimation models are used, then the system is simple to operate, but measurement precision deteriorates under transient conditions
Solution Approach 1:
The engine operating space is divided into multiple discrete local state spaces based on speed and injection quantity maps. Each local state space has its own transient model with specific parameters, allowing the system to capture transient behavior accurately without requiring a single complex global model. This segmentation enables precise NOx estimation during transient conditions while keeping individual local models relatively simple.
Solution Approach 2:
The patent implements dynamic modeling by developing transient state-space models that adapt to changing operating conditions. The model transitions between different local state spaces as engine operating parameters change, capturing the dynamic behavior of NOx emissions during transient conditions. This dynamic approach significantly improves measurement precision compared to static steady-state models.
2Measurement precision
If discrete transient local state space models are implemented, then NOx estimation accuracy improves, but device complexity increases
Solution Approach 1:
The complex transient modeling task is segmented into multiple discrete local state spaces, each with its own simplified model. This segmentation reduces the complexity of individual models while collectively capturing the full range of transient behaviors through model selection based on current operating conditions.
Solution Approach 2:
The framework uses a universal state-space model structure that can represent different local operating conditions through parameter variations. The same mathematical framework serves multiple functions by adapting parameters based on the current local state space, reducing the need for entirely separate models for different conditions.
3Reliability
If steady state models are used, then computational resources are conserved, but reliability deteriorates during transient operations
Solution Approach 1:
By segmenting the operating space into discrete local state spaces, the system can select from a finite set of pre-characterized models. This segmentation allows the system to achieve high reliability during transient operations by selecting the appropriate local model, while keeping computational energy use moderate through efficient model selection rather than continuous complex calculations.
Solution Approach 2:
The transient models for each local state space are pre-developed and characterized offline. During operation, the system only needs to identify the current local state space and retrieve the corresponding pre-computed model parameters, significantly reducing real-time computational energy requirements while maintaining high reliability.
Data Source
AI summary
Implementations described herein relate to utilizing discrete transient local state space models to modify an output from a steady state NOx estimation model to provide a final estimated NOx value. An engine operating space, defined by a speed and injection quantity map, is subdivided into several discrete local state spaces and a transient model is developed for each discrete state space to output a NOx estimation correction value. The transient models may be a regression model or the transient model may be an empirical map of data values. The NOx estimation correction value modifies a steady state NOx value to calculate the final NOx estimated value. The final NOx estimated value is then output to another component, such as a controller as an input for controlling one or more components of an engine, an aftertreatment system, and/or a diagnostic system.