区分
|
甲
|
氏 名
|
|
| 論
文
題名 |
VISUALIZATION OF UNDERGROUND AND
APPLICATION OF ARTIFICIAL INTELLIGENCE FOR THE UTILIZATION OF ROCK MASS |
| (岩盤利用のための地下の可視化と人工知能の適用性) |
| This dissertation is devoted to
two themes in rock mechanics. Two themes are the imaging of underground
and the methodology of handling the nonlinear and uncertain information
related to the rock engineering field. Imaging of underground needs the consideration about what information about rock masses should be selected, how it is expressed and what method can be implemented. The method of geophysical exploration has been utilized for seeking the information of rock masses for developing mines or land. The conventional geophysical exploration also gives widely good information about the persistence and its situation of discontinuities in rock mass, but another approach is requested for the visual interpretation through imaging operation by the effective expression on the properties of discontinuities. Therefore, a new approach for the visual interpretation was selected as a topic. As is generally known, much information obtained from monitoring the parameters and mechanical behavior of rock mass has nonlinear and uncertain properties. An approach using the conventional mechanical concept fundamentally has the limitation on the data with the above ambiguous. Therefore, the induction of Artificial intelligence (AI), was actively discussed in another field as an alternative approach, should be significantly considered. As the practical studies of AI application, two topics, which are the rational estimation of Joint roughness coefficient (JRC) values and the control of grout injection under grouting, were selected. According to the above scope, this dissertation consists of five chapters. Chapter 1 suggests the study outline how three topics will be investigated. Chapter 2 covers the visualization of rock masses and its application in local range. The contents consist of the field data obtained from an open-pit limestone quarry and a borehole for geological research, the visualization algorithm of three-dimensional image and its application to three places. The monitoring of borehole wall was accomplished through a portable type of borehole camera applicable to the field conditions. The invention of the visualization algorithm of three-dimensional image and the simulation technique based on these field data focused on the three-dimensional expression of discontinuities net and its geotechnical understanding. Application results give that the invented methodology has three capacities, which are to analyze the trace exposed on free face from the point of engineering geology, to find the profitable direction about the safety of area in the mechanical point and to understand visually the distribution of discontinuities. Chapter 3 covers the reasonable estimation of JRC through the application of AI. Prior to stating the application contents, a fundamental knowledge of AI was concisely given, which consists of the history of AI and the essential concept of two tools of AI, artificial neural network and fuzzy logic. And then the methodology using the image analysis of joint profile and the rational reasoning by fuzzy logic was introduced. In here, the rational estimation of JRC using AI focused on the rational reasoning process through the fuzzy inference system for avoiding the subjective determination of geotechnical engineers. The reasoning process firstly needs the image analysis on the digital photography of joint profiles. It results in the rational deduction of fuzzy logics about two essential factors of joint profiles, height of asperity and its slope. The fuzzy inference system was finally deduced This methodology for estimating of JRC values was applied to 15 pieces of natural joint profiles, which 13 pieces were used for the construction of the fuzzy inference system and other 2 pieces were applied to the fuzzy inference system. It was found that the determination of JRC value was strictly defined within 1 value of JRC. Chapter 4 covers the application of AI to the control of grout injection. As a more reasonable system against the usual of grouting operations depending excessively on the subjective decision and experience of operators, the methodology utilizing the database and a neural-fuzzy combined system was implemented. The neural-fuzzy combined system is composed of two fuzzy inference systems and an artificial neural network. The fuzzy inference systems play the part of the selection of the data type according to the behavior of pressure and flow rate on time during grout injection, also that of the opening range on grout valve corresponding to the difference between the flow rate of input data into the combined system and the value simulated by an artificial neural network. The artificial neural network plays the part of the training of database according to the selected type of input data, and as a result, it induces the general behavior corresponding to the data of each type. This control algorithm was applied to four case studies. It was found that the opening range of the grout valve sensitively responded to the behaviors of the flow rate and injection pressure and moreover, this response corresponded with the essential principle for controlling the grout valve. Conclusively, all results were synthetically summarized and process in the next improvement was concisely stated in chapter 5. |
| |