Roger W. Meier

USING ARTIFICIAL NEURAL NETWORKS AS A FORWARD APPROACH TO BACKCALCULATION

In recent years, artificial neural networks have successfully been trained to backcalculate pavement layer moduli from the results of falling weight deflectometer (FWD) tests. These neural networks provide the same solutions as existing programs, only thousands of times faster. Unfortunately, their use is constrained to the test conditions assumed during network training. These limitations arise from practical aspects of neural network training and cannot be circumvented easily. The goal of this research was to develop a backcalculation program combining the speed of neural networks and the flexibility of conventional programs to produce the same solutions as existing programs. This was accomplished by forgoing neural network backcalculation in favor of neural network forward-calculation, that is, using neural networks in place of complex numerical models for computing the forward-problem solutions used by the conventional backcalculation programs. A suite of neural networks, covering a range of flexible pavement structures, was trained using data generated by WESLEA, the forward-problem solver used in the WESDEF backcalculation program. When tested on 110 experimental FWD results, a version of WESDEF augmented by the neural networks provided statistically identical answers 42 times faster, on average, than the original. Provisions have been made for periodic upgrades as additional networks are trained for other pavement types and test conditions. Meanwhile, the original WESLEA can still be used when an appropriate network is unavailable. This preserves the flexibility of the original program while taking maximum advantage of the speed gains afforded by the neural networks.

Seasonal Temperature Effects on Flexible Pavements in Tennessee

Beginning in 1996, four newly constructed flexible pavements in Tennessee were instrumented with automated weather stations and various in situ devices to measure temperature and moisture conditions throughout the pavement system. Since their completion, falling weight deflectometer (FWD) tests have been performed periodically to assess the seasonal changes in pavement response. The collected data were analyzed to establish correlations between the pavement layer moduli (back calculated from the FWD results by using MODULUS, Version 5.0) and the asphalt layer temperature. The use of the BELLS3 model to relate asphalt temperature to air and surface temperatures is also explored.

Attempt at Resilient Modulus Modeling Using Artificial Neural Networks

The pitfalls inherent in the indiscriminate application of artificial neural networks to numerical modeling problems are illustrated. An example is used of an apparently successful (but ultimately unsuccessful) attempt at training a neural network constitutive model for computing the resilient modulus of gravels as a function of stress state and various material properties. Issues such as the quantity and quality of data needed to successfully train a neural network are explored, and the importance of an independent test set to verify network performance is examined.

An Initial Study of Surface Wave Inversion Using Artificial Neural Networks

An artificial neural network is proposed as an expeditious alternative to the trial-and-error and least-squares surface wave inversion techniques that are currently available. To use an artificial neural network for surface wave inversion, synthetic dispersion curves are calculated for representative shear wave velocity profiles using a theoretical wave propagation algorithm. An artificial neural network is then “taught” to map these dispersion curves back into their respective shear wave velocity profiles. Once the network has been successfully trained on these synthetic dispersion curves, experimental dispersion curves can be inverted by passing them through the neural network. Because the neural network requires only a single forward pass of the data, it performs inversions much more quickly than iterative procedures. To determine the feasibility of using an artificial neural network for surface wave inversion, a two-dimensional wave-propagation algorithm was used to create synthetic dispersion curves for 99 000 randomly generated, two-layer velocity profiles. A backpropagation neural network was then trained to associate the synthetic dispersion curves with their respective velocity profiles. The trained network was evaluated using synthetic dispersion curves for another 1000 randomly generated velocity profiles as surrogate experimental curves.

Observed Long-Term Water Content Changes in Flexible Pavements in a Moderate Climate

A comprehensive instrumentation system was installed at four sties across the state of Tennessee to monitor seasonal variations in the environmental factors affecting flexible pavement response. This paper presents some selected findings obtained from over five years of continuous data collection. The data included temperature and water content of the various pavement layers, and weather information. Falling weight deflectometes (FWD) tests were used to observe the pavement response during different seasons. The measured seasonal variations in subgrade and base water content were observed to be small. Consistent with these observations, the seasonal variations in FWD back calculated subgrade modulus were small, suggesting that these effects may not be important in the design of pavements in moderate climates. Since the pavement systems were new construction, little pavement distress was observed over the study period. However, as the pavements age, water infiltration may increase leading to greater water content changes in the unbound materials.

Effect of Temperature Averaging on Predicted Pavement Life

Mechanistic-empirical pavement design methods for flexible pavements are based on the assumption that the pavement life is inversely proportional to the magnitude of the traffic-induced pavement strains. These strains vary with the stiffness of the asphalt layer, which in turn varies with temperature. Because these relationships are nonlinear, the additional pavement life consumed at higher-than-average temperatures is not offset by savings at lower-than-average temperatures. As a result, whenever average pavement temperatures are used to determine the asphalt stiffness, pavement life is overestimated. Using hourly pavement temperature data from an instrumented pavement site in Tennessee, the effects of temperature averaging on predicted pavement life are examined and it is shown, for a typical full-depth asphalt pavement section supported by subgrade soils of different strengths, that pavement life can be overestimated by 50% to 75% if the temperatures are aggregated into monthly averages. It is also shown that even hourly average temperatures can produce errors if the hourly distribution of truck traffic is not taken into account.

SYSTEM CALIBRATION SAMPLING DESIGN BY GENETIC ALGORITHM

Most municipal water utilities use computerized numerical models of their water distribution systems for tasks such as planning expansions, sizing components, solving operations problems, and estimating operating costs. These models must be periodically calibrated to the existing system by imposing a known demand on the system, collecting pressure data at selected points, and adjusting the model parameters until the model predicts the same pressures. It is not currently possible to rationally design a data collection plan that yields the best calibration information given limited equipment and personnel. This research addresses the demand side of the calibration equation by using genetic algorithms to find the combination of open hydrants that causes water to flow at non-negligible velocities through as much of the pipe network as possible. The genetic algorithm, built around the hydraulic simulation package EPANET, has been applied to a network model for a small town in Ohio and was found to perform well.

Creep Behavior of Recycled-Content Expanded Polystyrene Geofoam Under Compressive Loading

This paper presents conventional creep test results for 0, 15 and 30% recycled-content expanded polystyrene (EPS)-block geofoam. The tests were performed on 50-mm (2-in.) cubical specimens with a nominal density of 21.6 kg/m3 under different stress levels for more than 9 months. It is concluded that the creep strain of recycled-content EPS geofoam increases with time and stress level in a manner similar to virgin EPS geofoam. As with virgin EPS geofoam, the elastic-limit stress appears to be a threshold stress level for creep strain development in recycled-content EPS geofoam.