NOx Sensor Falsification Detection via Probability Distribution
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for remote emission monitoring of diesel engines struggle with accurately detecting NOx sensor data falsification in real-time, leading to high calculation complexity and instability, especially when dealing with big data processing.
Innovation Solution
A method involving preprocessing and discretizing vehicle data sets to form probability distribution models, screening for reliable data domains, and calculating distribution probabilities to determine if NOx sensor readings from tested vehicles meet falsification conditions, including average distribution probability and urea consumption thresholds, while maintaining low calculation complexity and improving accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional remote monitoring based on OBD is used to acquire NOx sensor data, then real-time monitoring capability is achieved, but the system cannot distinguish between genuine and falsified sensor data
Solution Approach 1:
The patent introduces an intermediary statistical analysis layer that mediates between the raw sensor data and the monitoring system. By using probability distribution models as an intermediary, the system can indirectly detect falsification without directly modifying the sensor or ECU, thus maintaining low device complexity while improving data authenticity verification
Solution Approach 2:
The patent implements feedback mechanisms where the statistical characteristics of normal sensor data are continuously learned and used to feedback into the detection model. This feedback loop enables the system to adapt to different vehicle types and operating conditions, improving reliability without requiring complex reconfiguration of the monitoring infrastructure
2Measurement precision
If complex statistical analysis methods are used to detect data falsification, then detection accuracy is improved, but calculation complexity increases
Solution Approach 1:
The patent transforms the complex problem of falsification detection into a simpler parameter comparison task. By changing the approach from analyzing raw sensor signals to comparing statistical parameters (mean, variance, distribution shape), the system achieves high detection accuracy with reduced calculation complexity. The key parameter transformation is from time-domain sensor readings to frequency-domain probability distribution characteristics
Solution Approach 2:
The patent replaces complex mechanical/physical analysis methods with statistical mathematics. Instead of using sophisticated signal processing algorithms or physical models of sensor behavior, the invention substitutes these with probability distribution analysis, which is computationally simpler while maintaining or improving detection accuracy
3Measurement precision
If probability distribution models are built using reference vehicle data, then detection accuracy is improved, but data processing time increases
Solution Approach 1:
The patent applies preliminary action by pre-processing and storing the statistical characteristics of reference vehicle data before actual monitoring begins. The probability distribution models are built in advance during a learning phase, so that during real-time monitoring, only simple parameter comparisons are needed. This preliminary preparation significantly reduces the processing time required during actual detection operations
Solution Approach 2:
The patent segments the data processing task into distinct phases: offline model building using reference data, and online detection using the pre-built models. By segmenting the work between offline preparation and online execution, the system can use extensive computational resources during model building without impacting real-time detection performance
Data Source
AI summary
A method for determining NOx sensor data falsification based on remote emission monitoring, includes the steps of: acquiring a plurality of vehicle data sets and urea level data of to-be-tested reference vehicles, wherein vehicle data include NOx sensor readings and corresponding engine data vectors; acquiring urea level data of reference vehicles; calculating standard urea consumption per kilometer; (2) acquiring an average distribution probability of the vehicle data of the to-be-tested vehicles through a probability distribution evaluation step; counting a total proportion of invalid or negative NOx sensor readings in the plurality of vehicle data sets; determining whether the data of the to-be-tested vehicles satisfy one or more falsification conditions; if so, determining that the data from the NOx sensors of the to-be-tested vehicles are falsified; otherwise, determining that the data from the NOx sensors of the to-be-tested vehicles are not falsified.