Accurately predicting ground settlement is crucial for constructing safe and stable highways, but gathering sufficient data can be challenging. Researchers have developed a new technique using regression Kriging that can make accurate predictions even with limited data. By incorporating Box-Cox transformation and optimizing the trend structure, the team was able to significantly improve the accuracy of their ground settlement predictions compared to traditional methods. This innovative approach could pave the way for more reliable highway construction in areas with sparse soil data.
Unlocking the Secrets of Ground Settlement
Predicting ground settlement is a critical step in highway construction, as it helps engineers ensure the stability and longevity of the road. Excessive settlement can lead to cracks, potholes, and other structural issues that compromise safety and increase maintenance costs. While analytical and numerical models can provide estimates, they often rely on soil parameters obtained through experiments and tests, which can introduce uncertainties.
To overcome these limitations, researchers are increasingly turning to data-driven methods that leverage real-world monitoring data. However, a common challenge is the sparse nature of this data, particularly in the early stages of a project or in remote locations. Sparse data can significantly impact the accuracy of predictive models, leaving engineers with difficult decisions.
A Smarter Approach to Settlement Prediction
In their latest study, a team of researchers tackled the problem of ground settlement prediction with sparse data using a novel technique called regression Kriging (RK). This approach combines the power of polynomial regression and the best linear unbiased prediction of Kriging interpolation to capture the temporal trends and spatial correlations in the data.
The key innovations of the RK method lie in its ability to address two critical factors for accurate prediction: the stationarity of the sample residuals and the appropriate trend structure.
Achieving Stationarity with Box-Cox Transformation
Stationarity, which refers to the consistency of the statistical properties of the data over time, is a crucial prerequisite for Kriging interpolation. However, in the case of sparse data, the raw sample residuals often fail to meet this assumption.
To overcome this challenge, the researchers incorporated the neuralnetwork’>back-propagation neural network (BPNN) model.
The findings were remarkable: the RK method significantly outperformed all the other approaches, with substantially lower evaluation metrics like root mean square error, mean absolute error, and mean arctangent absolute percent error. Interestingly, the BPNN model, which is often touted for its powerful predictive capabilities, performed the worst among the methods tested, likely due to the inadequacy of the sparse training data.
Paving the Way for Reliable Highway Construction
This innovative study demonstrates the potential of regression Kriging for ground settlement prediction, particularly in situations where data is limited. By addressing the critical factors of stationarity and trend structure, the researchers have developed a robust and accurate method that could revolutionize highway construction practices.
As the global infrastructure continues to expand, the ability to make reliable predictions with sparse data will become increasingly valuable. The RK-based approach showcased in this research could pave the way for more efficient and cost-effective highway projects, ultimately contributing to safer and more sustainable transportation networks.
Author credit: This article is based on research by Lei Huang, Wei Qin, Guo-liang Dai, Ming-xing Zhu, Lei-Lei Liu, Ling-Jun Huang, Shan-Pian Yang, Miao-Miao Ge.
For More Related Articles Click Here
![neuralnetwork’>back-propagation neural network (BPNN)](https://en.wikipedia.org/wiki/Box%E2%80%93Cox<em>transformation’>Box-Cox transformation</a>, a flexible data-driven technique that can help transform the residuals to achieve stationarity. By applying this transformation, the team was able to significantly improve the prediction accuracy, as the transformed residuals better aligned with the assumptions of Kriging.</p>
<p><h3>Optimizing the Trend Structure</h3>
</p>
<p>In addition to the stationarity of the residuals, the researchers also investigated the influence of the trend structure on the prediction accuracy. They explored the use of different orders of polynomial functions as the trend component, ranging from first-order to third-order.</p>
<p><figure class="wp-block-image size-large"><img decoding="async" src="https://famefreq.s3.us-east-2.amazonaws.com/1729418566_41598_2024_75811_Fig3_HTML.png" alt="figure 2" class="wp-image-3" /><figcaption class="wp-element-caption">Fig. 2</figcaption></figure>
</p>
<p>The results showed that the first-order polynomial trend function, which is characterized by a constant slope, best captured the patterns of ground settlement in the case study. This aligned with the primary consolidation behavior of the saturated silt soil, where the settlement rate remained relatively stable over time.</p>
<p><h3>Outperforming Traditional Methods</h3>
</p>
<p>To assess the effectiveness of their RK-based approach, the researchers compared the predicted ground settlement results to those obtained from classical methods, such as the hyperbolic, exponential curve, Asaoka, and gray model (GM(1,1)) methods. They also included a comparison with a <a href=){kind=link}