پیش‌ بینی نرخ نفوذ tbm در انواع سنگ با استفاده از پارامترهای سیستم طبقه‌بندی rmr به روش آنالیز رگرسیون مبتنی بر یادگیری ماشین

علیرغم استفاده گسترده از ماشین ‌های حفر تمام مقطع تونل در صنعت تونل‌سازی، برآورد دقیق عملکرد آنها به ویژه در شرایط پیچیده زمین‌شناسی هنوز هم می‌تواند چالش‌برانگیز باشد. هدف از این مطالعه، بررسی امکان استفاده از سیستم‌های طبقه‌بندی توده‌سنگ معمول برای تخمین عملکرد ماشین حفر تونل در انواع مختلف سنگ، با استفاده از آنالیز رگرسیون مبتنی بر الگوریتم‌های یادگیری ماشین است. بدین منظور داده‌های واقعی عملکرد ماشین و همچنین داده‌های زمین‌شناسی از 10 پروژه تونل‌سازی در یک پایگاه داده جامع با 523 مقطع تونل در انواع مختلف سنگ جمع‌آوری و برای توسعه روابط جدید برای تخمین اندیس نفوذ صحرایی fpi بر اساس پارامترهای ورودی سیستم طبقه‌بندی rmr استفاده شد. از آنجا که انواع مختلف سنگ‌ها، بافت، ساختار و ترکیب کانی‌شناسی متفاوتی دارند و به نیروهای برشی ماشین پاسخ متفاوتی می‌دهند، ترکیب اثرات نوع سنگ در مدل‌های پیش‌بینی عملکرد می‌تواند دقت تخمین‌ها را بهبود بخشد. این روابط به ویژه در مرحله طراحی و برنامه‌ریزی یک پروژه تونل‌سازی، می‌توانند مفید واقع شوند.

prediction of tbm performance in different rock types using input parameters of rmr by applying ml-based regression analysis

despite the widespread use of tunnel boring machines (tbms) in tunneling projects, accurate estimates of their performance, especially in complex geological conditions, can still be challenging. the aim of this study is to investigate the possibility of using rmr rock mass classification system parameters to predict the tbm performance in different rock types, using regression analysis based on machine learning algorithms. therefore, real machine performance data, as well as geological and geomechanical data, were collected from 10 tunneling projects in a comprehensive database with 523 tunnel sections in different rock types and were used to develop new relationships to predict the field penetration index (fpi) based on input parameters of the rmr classification system. since different rock types have different textures, structures, and mineralogical compositions and respond differently to machine shear forces, combining the effects of rock type in the models for prediction of tbm performance can improve the accuracy of estimates. these relationships can be especially useful in the design and planning of a tunneling project. this study aims to investigate the possibility of using the input parameters of the rmr rock mass classification system to develop tbm performance prediction relationships using regression analysis based on machine learning (ml) algorithms. new equations can lead to estimating the performance of the tbm under different geological conditions and taking into account the main and effective parameters.introductionmany efforts have been made by researchers to predict the tbm performance based on rock mass classification systems. the rmr classification system shows a better correlation with the tbm penetration rate that is due to the use of uniaxial compressive strength (ucs) as an input parameter in this classification system. in most tunneling projects, input parameters of the rmr classification system are available. it should be noted that the rmr system, like other rock mass classification systems, has been developed to classify the conditions of the rock mass in terms of tunnel stability and support design, and the selection of effective parameters has also been based on this, and if the goal is to use the input parameters of the rmr system to predict the tbm performance, changes should certainly be made to the input parameters and weights. methodology and approachesto develop empirical equations based on rmr parameters, ml methods, which use supervised learning techniques, have been implemented. in this regard, first, the classic regression algorithm, and then, more modern algorithms, including stepwise regression and regularization algorithms, including lasso and ridge regression algorithms have been implemented.results and conclusionsthe input parameters of the rmr system, which are often available in tunneling projects, have been used to estimate field penetration index (fpi) as a result of applying ml methods, and new relationships have been developed. taking into account the rock type in modeling, the tbm performance has increased significantly, indicating the importance of the lithology in boring.