Engine State Estimation Guard Process for Neural Network Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Catalyst temperature prediction devices using neural networks may output values outside acceptable ranges when encountering unexpected parameters, leading to potentially inaccurate or unexpected results.
Innovation Solution
A state estimation device for internal combustion engines that includes a storage device with mapping data learned through machine learning, which adjusts estimated values to ensure they fall within acceptable ranges by executing a guard process, using upper and lower limit calculations based on input variables like outside air temperature and fluid energy variables.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a neural network is used for catalyst temperature prediction, then the prediction capability is improved, but the output may fall outside acceptable ranges when unexpected parameters are input
Solution Approach 1:
The patent applies preliminary action by pre-calculating and storing upper and lower limit values in lookup tables before the prediction process. These limit values are determined in advance based on expected operating conditions, allowing the system to quickly validate and adjust predictions without complex real-time calculations, thus ensuring reliability while maintaining prediction accuracy
Solution Approach 2:
The patent implements feedback by using the predicted catalyst temperature to calculate an expected upstream air-fuel ratio, comparing it with the actual measured value, and using this comparison to detect and correct prediction errors. This feedback mechanism ensures that predictions remain within acceptable ranges and improves overall system reliability
2Measurement precision
If machine learning mapping data is used for state estimation, then the estimation accuracy is improved, but the estimated value may become unexpectedly large or small
Solution Approach 1:
The system pre-calculates upper and lower limit values for estimated parameters and stores them in lookup tables before operation. These predetermined limits serve as guard rails that prevent the machine learning model from producing unexpected values, while still allowing high estimation accuracy within the valid range
Solution Approach 2:
The patent changes the parameter representation by transforming the estimated catalyst temperature into an expected upstream air-fuel ratio for validation. This parameter transformation allows the system to detect unexpected values indirectly and correct them, maintaining estimation accuracy while eliminating harmful outliers
3Reliability
If guard process is executed to adjust estimated values to acceptable ranges, then the reliability is improved, but the calculation complexity increases
Solution Approach 1:
The patent reduces processing complexity by pre-calculating and storing upper and lower limit values in lookup tables during the design phase. During operation, the system only needs to perform simple table lookups and comparisons rather than complex real-time calculations, thus improving reliability without significantly increasing processing complexity
Solution Approach 2:
The system uses simple, computationally inexpensive operations for the guard process - primarily lookup table access and basic comparisons - rather than complex algorithms. These lightweight operations provide reliable validation with minimal processing overhead, effectively treating the validation mechanism as a simple, disposable check
Data Source
AI summary
A state estimation device for an internal combustion engine includes: a storage device that stores mapping data, the mapping data being data defining a mapping that takes as an input an internal combustion engine state variable and that generates as an output an estimated value for estimating the state of the internal combustion engine; and an execution device that executes an acquisition process of acquiring the internal combustion engine state variable and an estimation process of calculating the estimated value based on the output of the mapping. The mapping data is data learned by machine learning. When the estimated value is out of an acceptable range, the execution device executes a guard process of adjusting the estimated value to a value close to or within the acceptable range. When executing the guard process, the execution device calculates the value after the guard process as the estimated value.