Description:
More than a quarter of Australian urban areas are covered by impermeable surfaces. This shift from "naturally-permeable" to "impervious" surfaces results in an increase in the surface runoff during rainfalls, which contributes to an increase in flash flooding. Permeable pavements are among potential solutions to reduce the amount of surface runoff and the risk of flash floods; however, the main barriers in their uptake is their perceived lack of capacity to perform under traffic loads, rate of infiltration, and perceived reduction in permeability over time caused by clogging. This study reports on the hydro-mechanical performance of a large-scale permeable pavement trial site, constructed by the combination of tyre-and rock-derived aggregates, bonded together using a polyurethane-based binder, located at a car park in the City of Mitcham, South Australia. An area of approximately 400 m 2 , consisting of 24 parking bays, was paved using a variety of waste tyre-based blends. A series of in-situ double-ring infiltration tests were carried out over a nine-month period to study the evolution of clogging over time. A practical maintenance program, involving the use of a conventional street sweeper vehicle, was also developed and successfully implemented to cope with the potential adverse effects of pavement clogging. The infiltration rates, even nine months after construction, were consistently greater than the 0.4 cm/s-requirement commonly suggested for porous asphalt systems. The mechanical sweeper using one sweeping pass was found to be an effective solution for permeability recovery. Furthermore, the intense spot-sweeping approach, which involved five sweeping passes over highly-clogged locations, led to a further increase in the permeability; however, the magnitude of improvement was relatively small compared with the single-sweeping technique
Description:
The choice of state and action representation in Reinforcement Learning (RL) has a significant effect on agent performance for the training task. But its relationship with generalization to new tasks is under-explored. One approach to improving generalization investigated here is the use of language as a representation. We compare vector-states and discreteactions to language representations. We find the agents using language representations generalize better and could solve tasks with more entities, new entities, and more complexity than seen in the training task. We attribute this to the compositionality of language
Description:
The choice of state and action representation in Reinforcement Learning (RL) has a significant effect on agent performance for the training task. But its relationship with generalization to new tasks is under-explored. One approach to improving generalization investigated here is the use of language as a representation. We compare vector-states and discreteactions to language representations. We find the agents using language representations generalize better and could solve tasks with more entities, new entities, and more complexity than seen in the training task. We attribute this to the compositionality of language
Description:
In this paper, a high‐capacity fabricated hollow section is proposed as a lightweight replacement for concrete‐filled composite members which are commonly used as beam‐columns in seismic zones. The lateral cyclic performance of the fabricated hollow beam‐column comprising of mild steel plates (grade 250) and high strength steel (HSS) tubes (grade 750) is investigated. A finite element model is developed and validated against available experimental data. To incorporate the cyclic hardening and softening of both steel materials, a combined plasticity model obtained from low‐cycle material tests is incorporated in the finite element model. A comprehensive sensitivity analysis is undertaken to propose an optimum slenderness design for the thin‐walled steel elements with ductile failure. While the high strength steel tube elements can significantly increase the axial and lateral capacity of the hybrid section, the cyclic failure mechanism is governed by mild steel plate elements, rather than a sudden failure in the high strength steel tubes.