Seismic experiments on welded steel moment connections (WSMCs) demonstrated that the welding sequence can influence the failure responses by influencing the localized low-cycle fatigue responses. It is widely known that welding residual stresses relax with a few inelastic loading cycles, and hence, there is a notion that the welding sequence or resulting residual stress does not influence the seismic response of WSMCs. Because of this notion, investigations on the influence of weld sequence on WSMC failure responses have been limited. This study developed a numerical scheme based on a sequentially uncoupled thermomechanical analysis technique to simulate the weld-induced residual stress and its influence on the WSMC seismic responses. The simulation technique is validated against the recorded temperatures and strains. The simulations indicate that high residual stresses tend to concentrate near the weld toe and around the weld access holes. Subsequent seismic loading prescribed to the WSMCs in the presence of weld residual stress simulates the influence of residual stress on the localized strain responses. Strain accumulations are simulated at locations where cracks were observed in the seismic experiments of WSMCs. The numerical results also demonstrate the influence of weld sequences on the local strain responses of WSMCs. Finally, the residual stresses at critical locations are found to relax with a few inelastic loading cycles, as widely known; however, strain accumulation persisted with inelastic loading cycles at the locations where cracks were observed in the experiments. The reason for this phenomenon is demonstrated to be related to the influence of the multiaxial stress state at these locations.
Plasma assisted catalysis of post consumer plastics remains an intensive area of research with the potential to provide a method for a closed loop lifecycle for plastics waste. This work focuses on applying a nonthermal plasma (NTP) pretreatment of polypropylene (PP) prior to catalytic deconstruction to modify the product distribution. With the addition of NTP pretreatment, a conversion percentage of up to 94.5%, an increase of 2.9% from catalytic deconstruction without a NTP pretreatment, was observed. Propene was observed as the largest product in all pretreatments tests with the maximum propene yield with no pretreatment. With the addition of different pretreatment conditions, the yield of different major products including propane, butane, methane, and carbon monoxide could be increased with little decrease in conversion. The effects of the plasma pretreatment on the PP were investigated with FTIR and XPS studies. The plasma conditions were investigated with optical absorption spectroscopy and optical emission spectroscopy for ozone and gas temperature measurements, respectively. This research provides a potential method for the tuning of catalytic deconstruction products to better meet market fluctuations in product prices.
Purpose This study examines how machine learning (ML) adoption reshapes consumer purchasing behavior and value creation across major e-commerce platforms within the broader context of the information society. Specifically, it investigates the impact of ML technologies on conversion rates and Average Order Value (AOV) in digitally mediated marketplaces. Methods Using secondary data from Amazon, Alibaba, and Etsy covering the period 2020–2023, the study applies descriptive statistics, t-tests, analysis of variance (ANOVA), regression analysis, and difference-in-differences techniques to compare platform performance before and after ML adoption. These methods allow for cross-platform comparison while controlling for marketing expenditure and seasonality. Results The findings reveal statistically significant increases in both conversion rates and AOV following ML adoption across all three platforms. However, the magnitude of these effects differs significantly by platform, reflecting variations in market structure, consumer access mechanisms, and platform-specific personalization strategies. Conclusion The results demonstrate that machine learning functions as a critical infrastructural force shaping consumer access, engagement, and economic outcomes in contemporary digital marketplaces. These findings contribute to understanding how algorithmic systems influence value formation in the information society and raise important implications for platform governance, ethical personalization, and digital inclusion.
This paper applies a travel cost model combining revealed and stated preference data to estimate the recreational value of a disc golf course near New York City, where such facilities were limited at the time of the study. Using intercept survey data from Franklin D. Roosevelt and Heckscher State Parks, we estimate individual consumer surplus per trip with a fixed-effects Poisson regression. The survey collects travel costs and contingent behavior responses, capturing how trip frequency would change with reduced travel time. By combining revealed and stated preferences, we control for individual-specific factors influencing demand. Supplementary data from the UDisc scorekeeping app provides an estimate of total annual visits, allowing us to calculate aggregate consumer surplus. We estimate individual surplus at approximately $68 per trip and total annual surplus for the course at about $6 million. These results can inform policymakers considering the allocation of public land.
This paper measures how a seasonal vehicle entry fee affects disc golf course choice in Westchester County, New York. We use ten years of user-level data from the UDisc app, covering 30,531 visits by 758 players from 2015 to 2024, in a setting with two nearby and broadly comparable courses. One course is located within a state park that charges a predictable $10 per-vehicle fee on a seasonal calendar, while the other course is free year-round. We estimate a within-user probit model of course choice, controlling for weather differences and a linear time index, with fixed effects for user, day of week, holiday, and year. When the fee is in effect, the probability of choosing the fee course falls by about 1.5 percentage points, a result that is robust to logit and linear probability models.