Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “prediction”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 811 records · Page 45

Theoretical prediction and impact of fundamental electric dipole moments

The predicted Standard Model (SM) electric dipole moments (EDMs) of electrons and quarks are tiny, providing an important window to observe new physics. Theories beyond the SM typically allow relatively large EDMs. The EDMs depend on the relative phases of terms in the effective Lagrangian of the extended theory, which are generally unknown. Underlying theories, such as string/M-theories compactified to four dimensions, could predict the phases and thus EDMs in the resulting supersymmetric (SUSY) theory. Earlier one of us, with collaborators, made such a prediction and found, unexpectedly, that the phases were predicted to be zero at tree level in the theory at the unification or string scale ~O(10 16 GeV). Electroweak (EW) scale EDMs still arise via running from the high scale, and depend only on the SM Yukawa couplings that also give the CKM phase. Here we extend the earlier work by studying the dependence of the low scale EDMs on the constrained but not fully known fundamental Yukawa couplings. The dominant contribution is from two loop diagrams and is not sensitive to the choice of Yukawa texture. The electron EDM should not be found to be larger than about 5 × 10 –30 e cm, and the neutron EDM should not be larger than about 5 × 10 –29 e cm. These values are quite a bit smaller than the reported predictions from Split SUSY and typical effective theories, but much larger than the Standard Model prediction. Also, since models with random phases typically give much larger EDMs, it is a significant testable prediction of compactified M-theory that the EDMs should not be above these upper limits. The actual EDMs can be below the limits, so once they are measured they could provide new insight into the fundamental Yukawa couplings of leptons and quarks. As a result, we comment also on the role of strong CP violation. EDMs probe fundamental physics near the Planck scale.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

Predicting Fracture Network Development in Crystalline Rocks

The geometric properties of fractures influence whether they propagate, arrest, or coalesce with other fractures. Thus, quantifying the relationship between fracture network characteristics may help predict fracture network development, and perhaps precursors to catastrophic failure. To constrain the relationship and predictability of fracture characteristics, we deform eight one centimeter tall rock cores under triaxial compression while acquiring in situ X-ray tomograms. The tomograms reveal precise measurements of the fracture network characteristics above the spatial resolution of 6.5 µm. We develop machine learning models to predict the value of each characteristic using the other characteristics, and excluding the macroscopic stress or strain imposed on the rock. The models predict fracture development more accurately in the experiments performed on granite and monzonite, than the experiments on marble. Fracture network development may be more predictable in these igneous rocks because their microstructure is more mechanically homogeneous than the marble, producing more systematic fracture development that is not strongly impeded by grain contacts and cleavage planes. The varying performance of the models suggest that fracture volume, length, and aperture are the most predictable of the characteristics, while fracture orientation is the least predictable. Orientation does not correlate with length, as suggested by the idea that the orientation evolves with increasing differential stress and thus fracture length. This difference between the observed and expected relationship between orientation and length highlights the influence of mechanical heterogeneities and local stress perturbations on fracture growth as fractures propagate, link, and coalesce.

58 GEOSCIENCES↗

A deep learning framework for layer-wise porosity prediction in metal powder bed fusion using thermal signatures

Abstract Part quality manufactured by the laser powder bed fusion process is significantly affected by porosity. Existing works of process–property relationships for porosity prediction require many experiments or computationally expensive simulations without considering environmental variations. While efforts that adopt real-time monitoring sensors can only detect porosity after its occurrence rather than predicting it ahead of time. In this study, a novel porosity detection-prediction framework is proposed based on deep learning that predicts porosity in the next layer based on thermal signatures of the previous layers. The proposed framework is validated in terms of its ability to accurately predict lack of fusion porosity using computerized tomography (CT) scans, which achieves a F1-score of 0.75. The framework presented in this work can be effectively applied to quality control in additive manufacturing. As a function of the predicted porosity positions, laser process parameters in the next layer can be adjusted to avoid more part porosity in the future or the existing porosity could be filled. If the predicted part porosity is not acceptable regardless of laser parameters, the building process can be stopped to minimize the loss.

42 ENGINEERING↗

Adaptive behavior and different thermal experiences of real people: A Bayesian neural network approach to thermal preference prediction and classification

Various observed and unquantifiable factors affect the thermal comfort of occupants in indoor environments and can lead to high uncertainty in the prediction and classification of their thermal preferences. The behavioral adaptation of occupants, by operating window systems for example, changes their thermal experience and expectations and therefore contributes to even higher prediction uncertainty. In this study, we applied a Bayesian neural network (BNN) algorithm to build a predictive model for occupant thermal preference using the ASHRAE Global Thermal Comfort Database II. The Bayesian method allows us to synthesize prior knowledge and available measurements into a unified modeling framework. It also offers a way to express and quantify uncertainty. Here we have performed a systematic study to test the efficiency and robustness of different BNN model configurations. In this study, the results show that the BNN model outperforms conventional thermal comfort models such as Predicted Mean Vote (PMV) and adaptive comfort model. The BNN model tends to produce more confident “prefer cooler” predictions with high possibility and low uncertainty. In contrast, the BNN model produces less certain predictions for “prefer no change” and “prefer warmer” across all occupants. Our findings suggest that linking occupants’ subjective evaluation measures and window opening/closing behavior to thermal comfort modeling effectively improves predictive performance.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗

Investigation of the effect of chemistry models on the numerical predictions of the supersonic combustion of hydrogen

In this numerical study, the influence of chemistry models on the predictions of supersonic combustion in a model combustor is investigated. To this end, 3D, compressible, turbulent, reacting flow calculations with a detailed chemistry model (with 37 reactions and 9 species) and the Spalart-Allmaras turbulence model have been carried out. These results are compared with earlier results obtained using single step chemistry. Hydrogen is used as the fuel and three fuel injection schemes, namely, strut, staged (i.e., strut and wall) and wall injection, are considered to evaluate the impact of the chemistry models on the flow field predictions. Predictions of the mass fractions of major species, minor species, dimensionless stagnation temperature, dimensionless static pressure rise and thrust percentage along the combustor length are presented and discussed. Overall performance metrics such as mixing efficiency and combustion efficiency are used to draw inferences on the nature (whether mixing- or kinetic-controlled) and the completeness of the combustion process. The predicted values of the dimensionless wall static pressure are compared with experimental data reported in the literature. The calculations show that multi step chemistry predicts higher and more wide spread heat release than what is predicted by single step chemistry. In addition, it is also shown that multi step chemistry predicts intricate details of the combustion process such as the ignition distance and induction distance. (author)

08 HYDROGEN↗

Predicting lake surface water phosphorus dynamics using process-guided machine learning

Phosphorus (P) loading to lakes is degrading the quality and usability of water globally. Accurate predictions of lake P dynamics are needed to understand whole-ecosystem P budgets, as well as the consequences of changing lake P concentrations for water quality. However, complex biophysical processes within lakes, along with limited observational data, challenge our capacity to reproduce short-term lake dynamics needed for water quality predictions, as well as long-term dynamics needed to understand broad scale controls over lake P. Here we use an emerging paradigm in modeling, process-guided machine learning (PGML), to produce a phosphorus budget for Lake Mendota (Wisconsin, USA) and to accurately predict epilimnetic phosphorus over a time range of days to decades. In our implementation of PGML, which we term a Process-Guided Recurrent Neural Network (PGRNN), we combine a process-based model for lake P with a recurrent neural network, and then constrain the predictions with ecological principles. We test independently the process-based model, the recurrent neural network, and the PGRNN to evaluate the overall approach. The process-based model accounted for most of the observed pattern in lake P; however it missed the long-term trend in lake P and had the worst performance in predicting winter and summer P in surface waters. The root mean square error (RMSE) for the process-based model, the recurrent neural network, and the PGRNN was 33.0 μg P L -1 , 22.7 μg P L -1 , and 20.7 μg P L -1 , respectively. All models performed better during summer, with RMSE values for the three models (same order) equal to 14.3 μg P L -1 , 10.9 μg P L -1 , and 10.7 μg P L -1 . Although the PGRNN had only marginally better RMSE during summer, it had lower bias and reproduced long-term decreases in lake P missed by the other two models. For all seasons and all years, the recurrent neural network had better predictions than process alone, with root mean square error (RMSE) of 23.8 μg P L -1 and 28.0 μg P L -1 , respectively. The output of PGRNN indicated that new processes related to water temperature, thermal stratification, and long term changes in external loads are needed to improve the process model. By using ecological knowledge, as well as the information content of complex data, PGML shows promise as a technique for accurate prediction in messy, real-world ecological dynamics, while providing valuable information that can improve our understanding of process.

54 ENVIRONMENTAL SCIENCES↗

Revenue prediction for integrated renewable energy and energy storage system using machine learning techniques

Revenue estimation for integrated renewable energy and energy storage systems is important to support plant owners or operators’ decisions in battery sizing selection that leads to maximized financial performances. A common approach to optimizing revenues of a hybrid hydro and energy storage system is using mixed-integer linear programming (MILP). Although MILP models can provide accurate production cost estimations, they are typically very computationally expensive. To provide a fast yet accurate first-step information to hydropower plant owners or operators who consider integrating energy storage systems, we propose an innovative approach to predicting optimal revenues of an integrated energy generation and storage system. In this study, we examined the performance of two prediction techniques: Generalized Additive Models (GAMs) and machine learning (ML) models developed based on artificial neural networks (ANN). Predictive equations and models are generated based on optimized solutions from a market participation optimization model, the Conventional Hydropower Energy and Environmental Resource System (CHEERS) model. The two predicting techniques reduce the computational time to evaluate annual revenue for one set of battery configurations from 3 h to 1 to 4 min per run while also being implementable with significantly less data. The model validation prediction errors of developed GAMs and ML models are generally below 5%; for model testing predictions, the ML models consistently outperform the regression equations in terms of root mean square errors. This new approach allows plant owners, operators, or potential investors to quickly access multiple battery configurations under different energy generation and market scenarios. This new revenue prediction method will therefore help reduce the barriers, and thereby promoting the deployment of battery hybridization with existing renewable energy sources.

13 HYDRO ENERGY↗

Development of simplified model for injection rate prediction of diesel injectors during transient and steady operation

Determining the injection rate of the diesel injector during the transient short injection operations is a critical matter to enhance the combustion process by improving air/fuel mixture formation. In this regard, numerous injection models have been suggested in the literature to predict the injection rate accurately. However, such algorithms have not precisely predicted the rate of injection (ROI) in transient short injection events. In addition, most of those models require many input parameters to characterize hydraulic and mechanical subsystems in the injector, which can make them inconvenient to tune for different injector specifications, in other words not suitable for universal usage. Therefore, our study aims to develop a simple injection model that can predict injection rate precisely in the short and long injection durations considering relatively fewer input parameters. The model traces the pressure variation inside the sac based on the bulk modulus theory to predict nozzle exit velocity, thus the injection rate of the injector. In the algorithm, the bulk modulus of fuel (E) and discharge coefficients of the nozzle hole and needle seat (C d,H and C d,N ) have been considered to vary rather than set constant during the injection process since such flow parameters would have a critical impact on the flow development at the transient stage, thus affecting the injection rate of diesel injectors. Here with this modelling strategy, the accuracy of the new model on ROI prediction was confirmed with acceptable model error for both short and long injection durations by validating against experimental results. To understand the effect of the new model scheme on ROI prediction, ROI results with the new model have been discussed by comparing it with the ROI results predicted with constant flow parameters.

42 ENGINEERING↗

Applying NIR and MIR spectroscopy for C and soil property prediction in northern cold-region ecosystems. Which approach works better?

Here, developing reliable predictions of soil attributes is necessary to understand northern cold-region climate-soil feedback. Calibration models using near-infrared (NIR) and mid-infrared (MIR) spectroscopy were developed to predict eight commonly measured soil properties for 119 soil samples representing a range of vegetation types, parent materials, and soil types spanning >23° of latitude from southeast Alaska to the Canadian high Arctic. In order to obtain a more accurate prediction, this study compared the performance of linear and non-linear calibration techniques, including lasso regression (Lasso), support vector machine (SVM), random forest (RF) and classic partial least squares (PLS) to predict different soil properties of these soils. Comparing the four models, we noticed that their performance was quite similar for MIR overall, while NIR achieved better results with a PLS model for our dataset. PLS coupled with MIR showed a better performance for soil parameters, such as total organic carbon (TOC), total nitrogen (TN), cation exchange capacity (CEC) and clay (R-squared of 0.9, 0.81, 0.80, and 0.84) when compared with NIR (R-squared of 0.85, 0.72, 0.81 and 0.68). However, using either MIR or NIR spectroscopy, PLS predictions for bulk density (BD) and sand content were not accurate. The variable importance analysis based on the PLS model successfully estimated the relative contribution of wavelengths influencing soil property predictions most. Overall, TOC, TN, CEC and clay mineral predictions are closely related to the occurrence of specific spectral bands in the MIR region. For example, wavelengths at 2978 and 1761 cm -1 for TOC and TN, as well as at 3064 cm -1 for CEC, were selected as the most influential predictor variables. We demonstrated that MIR spectroscopy is a powerful tool for more extensive monitoring in soils of the northern cold climate region; however, NIR could be utilized for rapid estimates when the highest accuracy is not essential.

54 ENVIRONMENTAL SCIENCES↗

External Validation of a Preoperative Nomogram for Prediction of the Risk of Recurrence After Radical Prostatectomy

Purpose: To test the validity of an updated version of the preoperative Kattan nomogram for prediction of recurrence after radical prostatectomy (RP), published by Stephenson et al. Methods and Materials: We relied on clinical and postoperative prostate-specific antigen follow-up data of 1978 patients treated with open RP at our institution between 1992 and 2006. The accuracy of the nomogram-derived recurrence-free survival predictions were separately tested at 1 to 10 years after RP. Moreover, the relationship between the nomogram-predicted recurrence-free survival rate and the observed recurrence-free survival rate was graphically explored in calibration plots. Results: Median follow-up of censored patients was 32 months. For nomogram-derived recurrence-free survival predictions at 1 to 10 years, the accuracy of the nomogram ranged from 70% to 76%. The calibration between the predicted and observed recurrence-free survival rates was good at 1 and 2 years after RP. However, at all other examined time points, departures from ideal predictions were recorded. Conclusions: Our findings indicate that the preoperative biochemical recurrence nomogram reported by Stephenson et al. can be applied to European patients with good accuracy. However, its calibration was only suboptimal for predictions made 3 or more years after RP, with a tendency to overestimate the probability of recurrence-free survival. This potential limitation should be considered when this tool is applied to European patients.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

PET-Based Treatment Response Evaluation in Rectal Cancer: Prediction and Validation

Purpose: To develop a positron emission tomography (PET)-based response prediction model to differentiate pathological responders from nonresponders. The predictive strength of the model was validated in a second patient group, treated and imaged identical to the patients on which the predictive model was based. Methods and Materials: Fifty-one rectal cancer patients were prospectively included in this study. All patients underwent fluorodeoxyglucose (FDG) PET-computed tomography (CT) imaging both before the start of chemoradiotherapy (CRT) and after 2 weeks of treatment. Preoperative treatment with CRT was followed by a total mesorectal excision. From the resected specimen, the tumor regression grade (TRG) was scored according to the Mandard criteria. From one patient group (n = 30), the metabolic treatment response was correlated with the pathological treatment response, resulting in a receiver operating characteristic (ROC) curve based cutoff value for the reduction of maximum standardized uptake value (SUV{sub max}) within the tumor to differentiate pathological responders (TRG 1-2) from nonresponders (TRG 3-5). The applicability of the selected cutoff value for new patients was validated in a second patient group (n = 21). Results: When correlating the metabolic and pathological treatment response for the first patient group using ROC curve analysis (area under the curve = 0.98), a cutoff value of 48% SUV{sub max} reduction was selected to differentiate pathological responders from nonresponders (specificity of 100%, sensitivity of 64%). Applying this cutoff value to the second patient group resulted in a specificity and sensitivity of, respectively, 93% and 83%, with only one of the pathological nonresponders being false positively predicted as pathological responding. Conclusions: For rectal cancer, an accurate PET-based prediction of the pathological treatment response is feasible already after 2 weeks of CRT. The presented predictive model could be used to select patients to be considered for less invasive surgical interventions or even a 'wait and see' policy. Also, based on the predicted response, early modifications of the treatment protocol are possible, which might result in an improved clinical outcome.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Individualized Prediction of Overall Survival After Postoperative Radiation Therapy in Patients With Early-Stage Cervical Cancer: A Korean Radiation Oncology Group Study (KROG 13-03)

Purpose: A nomogram is a predictive statistical model that generates the continuous probability of a clinical event such as death or recurrence. The aim of the study was to construct a nomogram to predict 5-year overall survival after postoperative radiation therapy for stage IB to IIA cervical cancer. Methods and Materials: The clinical data from 1702 patients with early-stage cervical cancer, treated at 10 participating hospitals from 1990 to 2011, were reviewed to develop a prediction nomogram based on the Cox proportional hazards model. Demographic, clinical, and pathologic variables were included and analyzed to formulate the nomogram. The discrimination and calibration power of the model was measured using a concordance index (c-index) and calibration curve. Results: The median follow-up period for surviving patients was 75.6 months, and the 5-year overall survival probability was 87.1%. The final model was constructed using the following variables: age, number of positive pelvic lymph nodes, parametrial invasion, lymphovascular invasion, and the use of concurrent chemotherapy. The nomogram predicted the 5-year overall survival with a c-index of 0.69, which was superior to the predictive power of the International Federation of Gynecology and Obstetrics (FIGO) staging system (c-index of 0.54). Conclusions: A survival-predicting nomogram that offers an accurate level of prediction and discrimination was developed based on a large multi-center study. The model may be more useful than the FIGO staging system for counseling individual patients regarding prognosis.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Predicting Esophagitis After Chemoradiation Therapy for Non-Small Cell Lung Cancer: An Individual Patient Data Meta-Analysis

Purpose: Concurrent chemoradiation therapy (CCRT) improves survival compared with sequential treatment for locally advanced non-small cell lung cancer, but it increases toxicity, particularly radiation esophagitis (RE). Validated predictors of RE for clinical use are lacking. We performed an individual-patient-data meta-analysis to determine factors predictive of clinically significant RE. Methods and Materials: After a systematic review of the literature, data were obtained on 1082 patients who underwent CCRT, including patients from Europe, North America, Asia, and Australia. Patients were randomly divided into training and validation sets (2/3 vs 1/3 of patients). Factors predictive of RE (grade ≥2 and grade ≥3) were assessed using logistic modeling, with the concordance statistic (c statistic) used to evaluate the performance of each model. Results: The median radiation therapy dose delivered was 65 Gy, and the median follow-up time was 2.1 years. Most patients (91%) received platinum-containing CCRT regimens. The development of RE was common, scored as grade 2 in 348 patients (32.2%), grade 3 in 185 (17.1%), and grade 4 in 10 (0.9%). There were no RE-related deaths. On univariable analysis using the training set, several baseline factors were statistically predictive of RE (P<.05), but only dosimetric factors had good discrimination scores (c > .60). On multivariable analysis, the esophageal volume receiving ≥60 Gy (V60) alone emerged as the best predictor of grade ≥2 and grade ≥3 RE, with good calibration and discrimination. Recursive partitioning identified 3 risk groups: low (V60 <0.07%), intermediate (V60 0.07% to 16.99%), and high (V60 ≥17%). With use of the validation set, the predictive model performed inferiorly for the grade ≥2 endpoint (c = .58) but performed well for the grade ≥3 endpoint (c = .66). Conclusions: Clinically significant RE is common, but life-threatening complications occur in <1% of patients. Although several factors are statistically predictive of RE, the V60 alone provides the best predictive ability. Efforts to reduce the V60 should be prioritized, with further research needed to identify and validate new predictive factors.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Comparison of fission product release predictions using PARFUME with results from the AGR-1 safety tests

Safety tests were conducted on fuel compacts from AGR-1, the first irradiation experiment of the Advanced Gas Reactor (AGR) Fuel Development and Qualification program, at temperatures ranging from 1600 to 1800 °C to determine fission product release at temperatures that bound reactor accident conditions. The PARFUME (PARticle FUel ModEl) code was used to predict the release of fission products silver, cesium, strontium, and krypton from fuel compacts containing tristructural isotropic (TRISO) coated particles during 15 of these safety tests. Comparisons between PARFUME predictions and post-irradiation examination results of the safety tests were conducted on two types of AGR-1 compacts: compacts containing only intact particles and compacts containing one or more particles whose SiC layers failed during safety testing. In both cases, PARFUME globally over-predicted the experimental release fractions by several orders of magnitude: more than three (intact) and two (failed SiC) orders of magnitude for silver, more than three and up to two orders of magnitude for strontium, and up to two and more than one orders of magnitude for krypton. The release of cesium from intact particles was also largely over-predicted (by up to five orders of magnitude) but its release from particles with failed SiC was only over-predicted by a factor of about 3. These over-predictions can be largely attributed to an over-estimation of the diffusivities used in the modeling of fission product transport in TRISO-coated particles. The integral release nature of the data makes it difficult to estimate the individual over-estimations in the kernel or each coating layer. Nevertheless, a tentative assessment of correction factors to these diffusivities was performed to enable a better match between the modeling predictions and the safety testing results. The method could only be successfully applied to silver and cesium. In the case of strontium, correction factors could not be assessed because potential release during the safety tests could not be distinguished from matrix content released during irradiation. Furthermore, in the case of krypton, all the coating layers are partly retentive and the available data did not allow the level of retention in individual layers to be determined, hence preventing derivation of any correction factors.

11 NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗

Real-time monitoring and prediction of water quality parameters and algae concentrations using microbial potentiometric sensor signals and machine learning tools

We report the overarching hypothesis of this study was that temporal microbial potentiometric sensor (MPS) signal patterns could be used to predict changes in commonly monitored water quality parameters by using artificial intelligence/machine learning tools. To test this hypothesis, the study first examines a proof of concept by correlating between MPS's signals and high algae concentrations in an algal cultivation pond. Then, the study expanded upon these findings and examined if multiple water quality parameters could be predicted in real surface waters, like irrigation canals. Signals generated between the MPS sensors and other water quality sensors maintained by an Arizona utility company, including algae and chlorophyll, were collected in real time at time intervals of 30 min over a period of 9 months. Data from the MPS system and data collected by the utility company were used to train the ML/AI algorithms and compare the predicted with actual water quality parameters and algae concentrations. Based on the composite signal obtained from the MPS, the ML/AI was used to predict the canal surface water's turbidity, conductivity, chlorophyll, and blue-green algae (BGA), dissolved oxygen (DO), and pH, and predicted values were compared to the measured values. Initial testing in the algal cultivation pond revealed a strong linear correlation (R 2 = 0.87) between mixed liquor suspended solids (MLSS) and the MPSs' composite signals. The Normalized Root Mean Square Error (NRMSE) between the predicted values and measured values were <6.5%, except for the DO, which was 10.45%. The results demonstrate the usefulness of MPSs to predict key surface water quality parameters through a single composite signal, when the ML/AI tools are used conjunctively to disaggregate these signal components. The maintenance-free MPS offers a novel and cost-effective approach to monitor numerous water quality parameters at once with relatively high accuracy.

54 ENVIRONMENTAL SCIENCES↗

Characterization and validation of an in silico toxicology model to predict the mutagenic potential of drug impurities*

Control and minimization of human exposure to potential genotoxic impurities found in drug substances and products is an important part of preclinical safety assessments of new drug products. The FDA's 2008 draft guidance on genotoxic and carcinogenic impurities in drug substances and products allows use of computational quantitative structure–activity relationships (QSAR) to identify structural alerts for known and expected impurities present at levels below qualified thresholds. This study provides the information necessary to establish the practical use of a new in silico toxicology model for predicting Salmonella t. mutagenicity (Ames assay outcome) of drug impurities and other chemicals. We describe the model's chemical content and toxicity fingerprint in terms of compound space, molecular and structural toxicophores, and have rigorously tested its predictive power using both cross-validation and external validation experiments, as well as case studies. Consistent with desired regulatory use, the model performs with high sensitivity (81%) and high negative predictivity (81%) based on external validation with 2368 compounds foreign to the model and having known mutagenicity. A database of drug impurities was created from proprietary FDA submissions and the public literature which found significant overlap between the structural features of drug impurities and training set chemicals in the QSAR model. Overall, the model's predictive performance was found to be acceptable for screening drug impurities for Salmonella mutagenicity. -- Highlights: ► We characterize a new in silico model to predict mutagenicity of drug impurities. ► The model predicts Salmonella mutagenicity and will be useful for safety assessment. ► We examine toxicity fingerprints and toxicophores of this Ames assay model. ► We compare these attributes to those found in drug impurities known to FDA/CDER. ► We validate the model and find it has a desired predictive performance.

60 APPLIED LIFE SCIENCES↗

Predicting chemically-induced skin reactions. Part I: QSAR models of skin sensitization and their application to identify potentially hazardous compounds

Repetitive exposure to a chemical agent can induce an immune reaction in inherently susceptible individuals that leads to skin sensitization. Although many chemicals have been reported as skin sensitizers, there have been very few rigorously validated QSAR models with defined applicability domains (AD) that were developed using a large group of chemically diverse compounds. In this study, we have aimed to compile, curate, and integrate the largest publicly available dataset related to chemically-induced skin sensitization, use this data to generate rigorously validated and QSAR models for skin sensitization, and employ these models as a virtual screening tool for identifying putative sensitizers among environmental chemicals. We followed best practices for model building and validation implemented with our predictive QSAR workflow using Random Forest modeling technique in combination with SiRMS and Dragon descriptors. The Correct Classification Rate (CCR) for QSAR models discriminating sensitizers from non-sensitizers was 71–88% when evaluated on several external validation sets, within a broad AD, with positive (for sensitizers) and negative (for non-sensitizers) predicted rates of 85% and 79% respectively. When compared to the skin sensitization module included in the OECD QSAR Toolbox as well as to the skin sensitization model in publicly available VEGA software, our models showed a significantly higher prediction accuracy for the same sets of external compounds as evaluated by Positive Predicted Rate, Negative Predicted Rate, and CCR. These models were applied to identify putative chemical hazards in the Scorecard database of possible skin or sense organ toxicants as primary candidates for experimental validation. - Highlights: • It was compiled the largest publicly-available skin sensitization dataset. • Predictive QSAR models were developed for skin sensitization. • Developed models have higher prediction accuracy than OECD QSAR Toolbox. • Putative chemical hazards in the Scorecard database were found using our models.

60 APPLIED LIFE SCIENCES↗

WearGP: A computationally efficient machine learning framework for local erosive wear predictions via nodal Gaussian processes

Computational fluid dynamics (CFD)-based wear predictions are computationally expensive to evaluate, even with a high-performance computing infrastructure. Thus, it is difficult to provide accurate local wear predictions in a timely manner. Data-driven approaches provide a more computationally efficient way to approximate the CFD wear predictions without running the actual CFD wear models. In this paper, a machine learning (ML) approach, termed WearGP, is presented to approximate the 3D local wear predictions, using numerical wear predictions from steady-state CFD simulations as training and testing datasets. The proposed framework is built on Gaussian process (GP) and utilized to predict wear in a much shorter time. The WearGP framework can be segmented into three stages. At the first stage, the training dataset is built by using a number of CFD simulations in the order of O (10 2 ). At the second stage, the data cleansing and data mining process are performed, where the nodal wear solutions are extracted from the solution database to build a training dataset. At the third stage, the wear predictions are made, using trained GP processes. Two CFD case studies including 3D slurry pump impeller and casing are used to demonstrate the WearGP framework, in which 144 training and 40 testing data points are used to train and test the proposed method, respectively. The numerical accuracy, computational effiency and effectiveness between the WearGP framework and CFD wear model for both slurry pump impellers and casings are compared. It is shown that the WearGP framework can achieve highly accurate results that are comparable with the CFD results, with a relatively small size training dataset, with a computational time reduction in an order of 10 5 to 10 6 .

42 ENGINEERING↗