Inverting Simulation Models for Event Factor Estimation

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

It is challenging to calculate input-equivalent data from output-equivalent data using simulation models inversely, making it difficult to estimate factors related to events detected in targets, particularly in abnormal conditions.

Innovation Solution

An information processing device and method that acquires training data including time-series data of measurement items and influencing items, and uses this data to train a model that receives input time-series data of measurement items and outputs time-series data of influencing items, enabling the estimation of event factors.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If simulation models are used to estimate target behavior, then forward prediction capability is improved, but inverse calculation capability to estimate event factors deteriorates

Engineering Contradiction:

Improveprediction accuracyVSAvoidinverse calculation difficulty

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The patent inverts the traditional simulation approach by training a neural network model to perform inverse calculations. Instead of using simulation models forward to predict target behavior from input factors, the invention trains the model to estimate event factors from observed target behavior, effectively reversing the calculation direction to solve the inverse problem

Inventive Principle:

Principle #13The other way round (Inversion)

Solution Approach 2:

The patent transforms the simulation model into a trainable neural network model by changing its parameters through learning. The model parameters are adjusted during training to minimize the difference between predicted and actual target behaviors, enabling the model to learn the inverse mapping from output to input space

Inventive Principle:

Principle #35Parameter changes

2Reliability

If traditional simulation models are used, then forward modeling capability is improved, but factor estimation capability for detected events deteriorates

Engineering Contradiction:

Improvesimulation reliabilityVSAvoidevent factor detection difficulty

Core Design Contradiction:

ReliabilityVSDifficulty of detecting and measuring

Solution Approach 1:

The patent transforms the rigid simulation model parameters into learnable parameters that can be adjusted through training data. This allows the model to adapt to specific event scenarios and accurately estimate event factors by learning the relationship between target behavior and influencing factors from historical data

Inventive Principle:

Principle #35Parameter changes

Solution Approach 2:

The patent introduces a feedback mechanism where the model's predictions are compared with actual observed data, and the parameters are adjusted based on this feedback. This iterative training process enables the model to improve its factor estimation capability by continuously learning from the difference between predicted and actual outcomes

Inventive Principle:

Principle #23Feedback

Data Source

PatentUS20230376727A1Information processing device, information processing method, and recording medium

Publication Date: 2023.11.23 NEC CORP

AI summary

An information processing device acquires training data. The training data includes time-series data of a measurement item regarding a target and time-series data of an item that influences the target. The information processing device trains, by using the training data, a model that receives an input of time-series data of the measurement item and outputs time-series data of the item that influences the target.