Civil & Environmental Engineering Theses and Dissertations
Permanent URI for this collection
Browse
Recent Submissions
- ItemStress-Controlled Versus Strain-Controlled Triaxial Testing of Sand(1994) Alqutri, Samir Ahmed; Goodings, Deborah J.; Civil Engineering; University of Maryland (College Park, Md); Digital Repository at the University of MarylandThe purpose of this research was to compare the strength characterizations of Mystic White Silica Sands using stress-controlled loading versus strain-controlled loading in a standard compression triaxial tests. To this end one hundred sixty-six tests were conducted involving two types of quartz sand, one fine MWSS45 and one medium coarse MWSS18 , tested at three low to intermediate confining stresses of 14 kN/m2, 28 kN/m2 and 55 kN/m2 with only one specimen diameter size of 71.1 mm. Of the one hundred sixty-six tests, eighty-six were stress-controlled tests and eighty were strain-controlled tests. All specimens were dry, but both loose and dense specimens were tested. The results were evaluated individually and as group. Comparison of the two types of loading tests were evaluated for repeatability, stress-strain characteristics and strength parameters. The plots show that stress-controlled loading in general gives more reproducible results with smoother. steeper stress-strain plot s and a larger average deviator stresses at failure than strain-controlled loading at all three levels of confining stresses for both sands. This results in somewhat larger values of Φ' . Stress-controlled specimens were stiffer and failed with a clear cut failure surface while strain-controlled specimens mostly barreled.
- ItemUSING DEEP GENERATIVE MACHINE LEARNING METHODS TO GENERATE SYNTHETIC POPULATION(2022) Yang, Zhichao; Cinzia, Cirillo; Civil Engineering; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)Population synthesis is an important area of research aiming at generating synthetic data about households and individuals that would be representative of real large populations. Scholars in different fields have worked on synthetic population generation: statisticians, computers scientists, economists, social scientists, and engineers. In transportation modeling, synthetic agents are a key input for agent-based models, that are gradually replacing zone-based aggregate four steps models. Traditional methods for population synthesis include Iterative Population Fitting (IPF), that weights sample data until marginals for the variables of interest match official statistics (often from CENSUS) at a certain geographical area. Recently, Machine Learning algorithms have been tested and compared to IPF, which suffers from several well-known limitations. In this M.S. thesis, advanced deep generative machine learning methods are applied to generate synthetic populations, including CTGAN and TVAE. CTGAN is an advanced GAN algorithm that models tabular data distribution and sample rows from the underlying distribution. It has been shown that CTGAN can solve issues that challenge conventional GAN model, including mixed data types, non-Gaussian distributions, multimodal distributions, learning from sparse one-hot-encoded vectors and highly imbalanced categorical columns. TVAE is also an advanced VAE model that adapts VAE to tabular data by using preprocessing and modifying the loss function. As a case study, this research applies these two machine learning methods to generate synthetic population based on a sample from the American Community Survey relative to the State of Maryland. To demonstrate the performance of the proposed methods, we compare our results to those obtained with IPF and Bayesian Network using metrics that evaluate the ability of the population synthetizer to reproduce the dependency structure and the marginals in the real population and to solve the problem of zero cells in IPF.
- ItemANALYZING BID PRICES QUANTITATIVELY AND PROTEST DECISIONS QUALITATIVELY TO REDUCE PROJECT-RELATED DISPUTES IN ADVANCE(2022) Kim, Young Joo; Skibniewski, Miroslaw J; Civil Engineering; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)Parties to a construction contract can consume significant resources in dealing with project-related disputes. Therefore, it is advantageous for project stakeholders to identify potential issues earlier to avoid such problems as much as possible. This dissertation research explored evidence-based approaches to reduce project-related disputes before commencing construction projects. The research was carried out by examining a cost dataset from a state Department of Transportation that prioritizes the lowest-priced bid and by investigating a bid protest dataset from a Federal Government office that typically prioritizes the best value. With the coefficient of variation of bids as an independent variable of interest, the cost dataset was quantitatively studied using Welch’s t-test, correlation and regression analyses, and the K-nearest neighbors classification. Then, the Government Accountability Office’s decisions on denied bid protests against the U.S. Army Corps of Engineers were qualitatively meta-summarized. The observations showed the limited usefulness of collective intelligence provided by bidders at the time of bid opening in identifying projects likely to experience more significant project cost changes upon completion, as well as the effectiveness of the thematic findings in limitedly helping small businesses fore-test the validity of their cases before filing bid protests. The results could be applied beyond the Architecture, Engineering, and Construction industries as projects occur in all industries and industry sectors.
- ItemNUTRIENT MOVEMENT IN A VEGETATED COMPOST BLANKET AMENDING A VEGETATED FILTER STRIP ON A HIGHWAY SLOPE(2022) Forgione, Erica Rose; Davis, Allen P; Civil Engineering; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)Excess stormwater runoff caused by rapid urbanization and exacerbated by climate change generates many challenges for public safety and the environment. Large runoff volumes contribute to flooding and pollutants in stormwater runoff pose risks to human and environmental health, including toxicity to the aquatic environment caused by heavy metals and nutrient pollution leading to eutrophication, the cause of harmful algal blooms. An effort is being made to improve the efficiency of existing highway stormwater control systems which have limited performance in terms of volume reduction and pollutant removal. To address this issue, amendment of highway Vegetated Filter Strips (VFS) with a Vegetated Compost Blanket (VCB), a layer of seeded compost placed on an established slope, has been proposed. Compost has high water holding capacity and organic matter content which can immobilize contaminants of concern. However, the high nutrient content of compost poses a threat to net beneficial performance since excess nutrient leaching occurs after application. This research has posed the question: Can a VCB be used as a stormwater control measure (SCM) while avoiding excessive nutrient leaching?The VCB/VFS system was assessed through lab-scale, greenhouse-scale, and field-scale experiments. Hydrologic performance was evaluated in field and greenhouse experiments through evaluation of dynamic flow modification, event volume storage, and cumulative volume retention. Water quality performance was assessed through analysis of Total Suspended Solids (TSS), Nitrate + Nitrite (NOx), Total Kjeldahl Nitrogen (TKN), Total Nitrogen (TN), Total Phosphorus (TP), filtered and total Copper, and total Zinc concentrations. Nitrogen (N) and phosphorus (P) in compost are naturally transformed from organic to inorganic, soluble forms through the microbially-mediated process of mineralization. Nutrient removal occurs through adsorption as compost leachate passes through the VFS soil layer. To further investigate nutrient movement, small scale laboratory experiments were completed to determine the N and P compost mineralization rates and theoretical soil adsorption capacities. Nutrient data from greenhouse and field experiments were empirically evaluated using the lab-obtained mineralization data. Nutrient release was simulated and compared to experimental field data using a new open-source software, OpenHydroQual, which combines hydraulic and water quality modeling. VCBs were found to have a significant impact on both flow and volume reduction, though at the highest flowrates, VCBs were unable to significantly reduce flow and instead acted as conveyance. A useful design estimate for representative storage capacity using the saturated moisture content and wilting point of both the VCB and VFS was determined. Significant TSS removal was observed in both the field and greenhouse studies and particulate metals were largely removed; however dissolved copper leaching was observed in the field experiment, as has been observed previously for some compost in stormwater systems. Highly elevated concentrations of nutrients (as high as 100 mg/L TN and 12 mg/L TP) were observed in the effluent of both field and greenhouse experiments, resulting in net nutrient leaching and concentrations above recommended EPA freshwater limits even after 1-2 years. Additionally, mass loading rates at the field site (as high as 41 kg/ac/yr for TN and 14 kg/ac/yr for TP) were 1-2 magnitudes higher than observed influent mass loading rates (~3.8 kg/ha/yr for TN and ~0.47 kg/ha/yr for TP). Through laboratory mineralization studies, N and P mineralization rates were found to differ between compost batches, with initial nutrient content and age/leaching of compost being important factors. Adsorption experiments indicated increasing P adsorption from compost leachate with increasing soil Al+Fe content. Comparisons to greenhouse and field data showed differences in N speciation, likely due to differences in moisture content and temperature causing differing amounts of nitrification and volatilization. OpenHydroQual modeling showed modest results, with varying levels of accuracy for storm hydrograph simulation and mass release. VCBs are not currently recommended for use due to the risk of nutrient and metals pollution, especially in nutrient and metals sensitive watersheds. However, several impactful factors were identified that may reduce nutrient leaching, including compost composition, compost age/leaching, and VFS soil type.
- ItemANALYSIS OF THREE DIFFERENT MACHINE LEARNING ALGORITHMS FOR SWE ESTIMATION OVER WESTERN COLORADO USING SPACE-BASED PASSIVE MICROWAVE RADIOMETRY(2022) Yu, Bincheng; Forman, Barton; Civil Engineering; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)This study compares the performance of three different machine learning algorithms used for snow water equivalent (SWE) estimation. Inputs to these algorithms include passive microwave (PMW) brightness temperature (Tb) observations at 10.65 GHz, 18.7 GHz, and 36.5 GHz at both vertical and horizontal polarization as collected by the Advanced Microwave Scanning Radiometer (AMSR-2). The three algorithms include: 1) support vector machine (SVM) regression, 2) long short-term memory (LSTM) networks, and 3) Gaussian process (GP) regression. In-situ SWE measurements from the SNOTEL network collected across western Colorado is used as the training “targets” during the training procedure. The performance of the algorithms is evaluated using a number of different metrics including, but not limited to correlation coefficient, mean square error (MSE), and bias. The evaluation is conducted over a range of different elevations and different land cover classifications in order to assess algorithm performance across a broad range of snowpack conditions. Preliminary results suggest the LSTM algorithm is computationally more efficient during the training process as compared to the other algorithms, yet yields a similar level of performance. Some limitations, however, have been found in the study, including poor performance during deep snow conditions, which is likely related to signal “saturation” within the PMW Tb’s used during the supervised training process. Additionally, algorithm performance is strongly dependent on the amount of training data such that too little training data results in poor performance by the algorithm at successfully reproducing inter-annual variability. The strengths and limitations of these different machine learning algorithms for snow mass estimation will be discussed.