Probabilistic Load Model for Vehicle Component Simulation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for determining vehicle component loads fail to account for the variability in training drives and real-world conditions, leading to inaccurate load simulations due to unconsidered influencing factors such as driver behavior and route differences.
Innovation Solution
A method using a probabilistic load model created from training data to simulate a range of possible loads by approximating the probability distribution of loads on a training route, allowing for realistic simulations of load distributions across different conditions and vehicle configurations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If deterministic load models are used with predetermined speed profiles and routes, then the modeling process is simple and deterministic, but the load simulation does not account for real-world variability in driver behavior and route conditions
Solution Approach 1:
The patent transforms the load model from a deterministic parameter-based approach to a probabilistic approach by introducing probability distributions for load values. Instead of using fixed speed profiles and deterministic load calculations, the model now incorporates statistical variations to reflect real-world driving conditions, thereby improving reliability while managing complexity through structured probabilistic frameworks.
Solution Approach 2:
The patent replaces the traditional mechanical/deterministic load modeling system with a data-driven probabilistic system. By substituting fixed mathematical models with statistical models that learn from training data, the system captures real-world variability without requiring complex manual modeling of every possible driving scenario.
2Adaptability or versatility
If multiple speed profiles and routes are predetermined to simulate different loads, then various load conditions can be covered, but the process requires extensive manual preparation and does not capture actual driving variability
Solution Approach 1:
The patent performs preliminary action by collecting and storing training data from actual driving operations before the load simulation phase. This pre-collection of diverse driving data enables the probabilistic model to be trained in advance, so that when load simulations are needed, the model can quickly generate realistic load variations without requiring manual preparation of multiple speed profiles and routes.
Solution Approach 2:
The patent creates a probabilistic copy of real-world driving behavior based on training data. Instead of manually creating multiple deterministic scenarios, the system learns the statistical patterns from actual driving data and generates synthetic load scenarios that replicate real-world variability, thereby covering diverse load conditions efficiently.
3Measurement precision
If training drives are conducted by professional drivers on the same training track, then consistent baseline data can be obtained, but slight variations in driving behavior still result in different load courses that are not accounted for in deterministic models
Solution Approach 1:
The patent introduces dynamics into the load model by transitioning from static deterministic values to dynamic probabilistic distributions. The model now adapts to variations in driving behavior by using probability distributions that capture the range of possible load values, allowing it to accommodate natural variations even among professional drivers while maintaining measurement consistency through statistical frameworks.
Data Source
AI summary
A load prediction method for a component of a vehicle includes a model creation process. A load model is created using an identification method based on a training-drive data group, a training vehicle data group and a training load group. The training-drive data group contains training-drive data sets, each containing route data and/or accompanying drive data for a training drive of a training vehicle on a training route. The training vehicle data group comprises vehicle data from the training vehicle used on the training drive. The training load group includes training load data including a load of the component that corresponds to a training-drive data set. The load model approximates the occurring load on a predetermined training route or with predetermined accompanying drive data or according to predetermined vehicle data. In a model evaluation process, the load prediction is determined using the load model.