Liquefaction Evaluation Model Using Vibration Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current liquefaction evaluation methods require extensive labor and data collection for geological and geographical features, making them inefficient for assessing liquefaction risk.
Innovation Solution
A liquefaction evaluation model generation device that uses machine learning to generate a model based on training data from seismometer and pore water pressure gauge readings, either from simulated or actual ground conditions, to predict liquefaction without the need for extensive data collection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional liquefaction determination method using neural network with geological and geographical features is used, then liquefaction evaluation can be performed, but extensive labor and data collection work is required
Solution Approach 1:
The invention extracts only the essential vibration data (acceleration time histories) from the complex set of geological and geographical features. By focusing solely on vibration observations from seismometers, the method removes the need for extensive boring, Swedish sounding tests, and surface wave exploration, thereby dramatically reducing data collection work while maintaining evaluation capability
Solution Approach 2:
The invention creates a simplified evaluation model that copies only the essential input-output relationship needed for liquefaction assessment. Instead of using the full complex neural network model requiring multiple geological parameters, a new model is trained that replicates liquefaction evaluation using only vibration data, thus preserving accuracy while eliminating data collection burdens
2Measurement precision
If neural network training with multiple geological parameters is performed, then comprehensive liquefaction assessment is achieved, but the complexity of data collection and processing increases
Solution Approach 1:
The invention extracts only the vibration acceleration data from the comprehensive set of geological parameters. By removing unnecessary parameters such as borehole data, Swedish sounding results, and surface wave exploration data, the system achieves precise liquefaction measurement using a simplified data collection framework
Solution Approach 2:
The invention changes the input parameters of the evaluation model from multiple geological and geographical features to solely vibration acceleration data. This parameter transformation simplifies the data collection system while the machine learning model learns the appropriate representations needed for accurate liquefaction assessment
Data Source
AI summary
A liquefaction evaluation model generation device includes a data acquisition unit configured to acquire training data in which learning vibration data indicating a physical quantity associated with vibrations observed in the ground is defined as a question and learning liquefaction data indicating the degree of liquefaction occurring in the ground where the vibrations associated with the learning vibration data have been observed is defined as an answer, and a machine learning execution unit configured to execute machine learning using the training data and generate a liquefaction evaluation model that is a machine learning model. A liquefaction evaluation device includes a data acquisition unit configured to acquire inference vibration data indicating a physical quantity associated with vibrations observed in the ground and an inference execution unit configured to input the inference vibration data to the above-described machine learning model and cause the machine learning model to output inference liquefaction data indicating the degree of liquefaction occurring in the ground where the vibrations associated with the inference vibration data have been observed.