Raul Fuentes

Deeper Networks for Pavement Crack Detection

Pavement crack detection using computer vision techniques has been studied widely over the past several years. However, these techniques have faced several limitations when applied to real world situations due to for example changes of lightning conditions or variation in textures. But the recent advancements in the field of artificial neural networks, especially in deep learning, have paved a new way for applying computer vision methods to pavement crack detection. In this paper we demonstrate the effectiveness of deep networks in computer vision based pavement crack detection. We also show how variations in location of training and testing datasets affect the performance of the deep learning based pavement crack detection method.

Field monitoring of static, dynamic, and statnamic pile loading tests using fibre Bragg grating strain sensors

Pile loading test plays an important role in the field of piling engineering. In order to gain further insight into the load transfer mechanism, strain gauges are often used to measure local strains along the piles. This paper reports a case whereby FBG strain sensors was employed in a field trial conducted on three different types of pile loading tests in a glacial till. The instrumentation systems were configured to suit the specific characteristic of each type of test. Typical test results are presented. The great potential of using FBG sensors for pile testing is shown.

A Smoothed Particle Hydrodynamics Study of an Experimental Debris Flow

Open image in new window A comparison between a physical model debris flow of a monodispersed sand and a numerical model is presented. The physical experiment consisted of the instantaneous release of a saturated sand mass via a lock gate in a 220 × 160 mm rectangular channel of length 1.75 m. The sand mass had a 20% moisture content and was placed in the same position behind the lock gate before each experimental run. Additional water was added to the mass to keep the sand particles in suspension. The inclination of the flume was 31° which meant that the sand reached runout velocities of 1 m/s, approximately. The free surface of the flow was monitored in order to compare the readings to a Smoothed Particle Hydrodynamics (SPH) numerical model where a frictional Mohr-Coulomb model was implemented. This model was that of Voellmy, including the turbulent Voellmy coefficient. The free surface of the experimental flow was predicted well by the numerical model, showing the ability of the SPH model to predict the run out behaviour of small scale debris flow experiments with physically viable rheological parameters. This validation of the model is essential for its use in predicting landslide behaviour in the field. The prediction of the behaviour of landslides in areas that are susceptible to them is a key step in reducing the damage that they cause.

HYD verifications using numerical methods

ABSTRACT HYD, as described in Eurocode 7 (EC7), is related to the upward flow of water through the soil towards a free surface, such as in front of a retaining wall or in the base of an excavation. The HYD verification, using numerical analysis, can be performed with two different approaches. The first approach is the conventional soil block approach where safety may be checked by calculating the equilibrium of a rectangular block of soil. The second approach is the integration point approach where stability can be verified at every integration point in the numerical analysis by checking that the equilibrium is satisfied for a soil column of negligible width above each point. In this paper, the two approaches are described and their advantages and disadvantages are discussed. Comparisons made using benchmark geometries, extensively studied and discussed between the members of the EC7 Evolution Group 9, on Water Pressures, illustrate that the HYD verification using numerical methods seems very promising. Thorough comparisons between the factors from the two approaches allow designers to better understand the benefits of using more advanced and robust approaches for such stability verifications.

An improved robot for bridge inspection

This paper presents a significant improvement from the previous submission from the same authors at ISARC 2016. The robot is now equipped with low-cost cameras and a 2D laser scanner which is used to monitor and survey a bridge bearing. The robot is capable of localising by combining a data from a pre-surveyed 3D model of the space with real-time data collection in-situ. Autonomous navigation is also performed using the 2D laser scanner in a mapped environment. The Robot Operating System (ROS) framework is used to integrate data collection and communication for navigation.

Design of Deep Supported Excavations: Comparison Between Numerical and Empirical Methods

This paper focuses on the derivation of design prop loads for supported excavations in stiff clay with increasing excavation depth and number of prop levels. For multi-propped walls there are a number of empirical graphs to obtain the design prop forces. CIRIA C517 (Twine & Roscoe, 1999) enhancing Terzaghi & Pecks work (Terzaghi & Peck (1967) and Peck (1969)) and making it more relevant in the UK practice, suggests the Distributed Prop Load (DPL) method based on 81 case histories and field measurements of prop loads. Similar guidance and empirical graphs exist in other countries such as the EAB Recommendations in Germany (Recommendations on Excavations: EAB, 3rd Edition, 2014). The design prop loads derived by empirical graphs (both CIRIA and EAB which are widely used in the UK and Germany respectively) and Finite Element methods are compared in the context of Eurocode 7 requirements. The German recommendations give prop loads in better agreement with the numerical analysis results. Suggestions are made to update the CIRIA guidance in line with the German recommendations and give different shapes of pressure distribution for supported walls with different number of prop levels. This can result in more realistic predictions of prop loads for upper layers, particularly in deep excavations, and hence more economic design.