Swarming is a collective flagella-dependent movement of bacteria across a surface that is observed across many species of bacteria. Due to the prevalence and diversity of this motility modality, multiple models of swarming have been proposed, but a consensus on a general mechanism for swarming is still lacking. Here, we focus on swarming by Pseudomonas aeruginosa due to the abundance of experimental data and multiple models for this species, including interpretations that are rooted in biology and biophysics. In this review, we address three outstanding questions about P. aeruginosa swarming: what drives the outward expansion of a swarm, what causes the formation of dendritic patterns (tendrils), and what are the roles of flagella? We review models that propose biologically active mechanisms including surfactant sensing as well as fluid mechanics-based models that consider swarms as thin liquid films. Finally, we reconcile recent observations of P. aeruginosa swarms with early definitions of swarming. This analysis suggests that mechanisms associated with sliding motility have a critical role in P. aeruginosa swarm formation.
The importance of electric arc furnace (EAF) steelmaking is expected to increase worldwide as parts of the industry transition to lower carbon dioxide emissions. This work analyzed one year’s operational data from an EAF plant that uses a large proportion of direct-reduced iron (DRI) in the furnace feed. The data were used to test different approaches to quantifying the effects of process conditions on specific electricity consumption (kWh per ton of crude steel). In previous work, inputs such as the proportion of DRI, fluxes, natural gas, and oxygen were linearly correlated with the specific electricity consumption. The current work has confirmed that conventional multiple linear regression (MLR) reproduces electricity consumption trends in EAF steelmaking, but many model coefficients deviated significantly from expected values and appeared unphysical. The implementation of engineered features—the slag volume and total carbon input—in an MLR model resulted in coefficients that were closer to expectations, but did not improve prediction accuracy. Further improvement was obtained by applying the engineered features to a non-linear machine-learned model (based on XGBoost), yielding both physically reasonable trends and smaller prediction errors. Trends from Shapley dependence analysis (applied to the XGBoost model) are quantitatively consistent with theoretical trends. These include the energy needed to melt slag, and the endothermic effect of carbon additions. The fitted models demonstrate the potential to diagnose poor slag foaming by showing an increase in electricity consumption with increased oxygen use. This example demonstrates that practically important steelmaking process insights inferred via a linear regression approach can be improved by applying Shapley analysis to a machine-learned model based on engineered features.