Portable Soil UCS Prediction Using LIBS and Machine Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for determining soil unconfined compressive strength are time-consuming, costly, and prone to inaccuracies, and existing machine learning approaches are limited by oversimplification and computational expense.
Innovation Solution
A field portable device using laser-induced breakdown spectroscopy (LIBS) combined with a decision tree regressor and adaptive boosting classifier to predict soil unconfined compressive strength, incorporating bulk density and water content as input features.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional UCS measuring methods (manual digging, physical property measurement setup, GPR, ERT) are used, then measurement precision is improved, but productivity deteriorates due to time-consuming and costly procedures
Solution Approach 1:
The patent replaces mechanical and physical measurement systems (manual digging, physical property measurement setup, GPR, ERT) with a laser-based optical system. The LIBS device uses laser-induced breakdown spectroscopy to analyze soil samples, substituting mechanical extraction and physical measurement methods with a non-contact optical analysis approach that provides both high precision and rapid results.
Solution Approach 2:
The patent changes the measurement parameter from mechanical/physical properties to spectral emission intensities. By measuring the intensity of light emitted during laser-induced plasma breakdown, the system derives UCS values through machine learning models, transforming the measurement approach from direct mechanical testing to optical spectroscopy-based indirect measurement.
2Productivity
If in situ soil testing methods are used, then productivity is improved by eliminating sample extraction and transport, but measurement precision deteriorates due to limited applicability and monitoring difficulty
Solution Approach 1:
The patent replaces traditional in situ testing equipment with a portable LIBS device that uses laser-induced breakdown spectroscopy. This optical system can be deployed directly at the construction site, eliminating the need for sample extraction and transport while maintaining high measurement precision through spectral analysis and machine learning-based UCS prediction.
Solution Approach 2:
The patent introduces machine learning models as an intermediary between the raw spectral data and UCS values. The trained models (decision tree regressor, random forest regressor, support vector regressor) translate the spectral emission intensities into accurate UCS predictions, bridging the gap between optical measurement and mechanical property determination.
3Productivity
If statistical equations are used for soil analysis, then productivity is improved through rapid analysis, but measurement precision deteriorates due to oversimplification and assumptions
Solution Approach 1:
The patent replaces simple statistical equations with advanced machine learning models that can capture complex non-linear relationships in soil data. The trained regressors (decision tree, random forest, support vector) provide rapid predictions while maintaining high accuracy by learning from training datasets, substituting oversimplified statistical methods with intelligent algorithms.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
The device achieves high accuracy in predicting soil unconfined compressive strength, with R2-scores up to 99.03%, enabling efficient and precise geotechnical engineering applications.
Implementation Method 1
performing, with a laser induced breakdown spectrometer, laser induced breakdown spectroscopy (LIBS) on the soil sample to generate spectral emission intensities of the soil sample
Data Source
AI summary
A field portable device for determining the unconfined compressive strength of a soil sample includes a sample holder, a heating device, a scale, a spectrometer, and a microprocessor. The sample holder receives a soil sample. The heating device dries the soil sample for a specified time. The scale measures a weight of the soil sample and a dried weight of the soil sample. The spectrometer performs laser induced breakdown spectroscopy on the soil sample and generates spectral emission intensities of the soil sample. The microprocessor calculates a bulk density and a water content of the soil sample, actuates the spectrometer to generate the spectral emission intensities. The microprocessor applies the spectral emission intensities, the bulk density and the 10 water content to a trained machine learning regressor combined with an adaptive boosting classifier to predict the unconfined compressive strength of the soil sample.