This study evaluates the performance of alkali-activated slag-fly ash blended concrete made with recycled concrete aggregates (RCA) and reinforced with steel fibers. Two blends of concrete with ground granulated blast furnace slag-to-fly ash ratios of 3:1 and 1:1 were used. Natural aggregates were substituted with RCA, while macro steel fibers with 35 mm of length and aspect ratio of 65 were incorporated in RCA-based mixtures at various volume fractions. Fine aggregates were in the form of desert dune sand. Mechanical and durability characteristics were investigated. Experimental results revealed that RCA replacement decreased the compressive strength of plain concrete mixtures with more pronounced reductions being perceived at higher replacement percentages. Mixtures made with 30%, 70%, and 100% RCA could be produced with limited loss in the design compressive strength upon incorporating 1%, 2%, and 2% steel fibers, by volume, respectively. In turn, splitting tensile strength was comparable to the NA-based control while adding at least 1% steel fiber, by volume. Moreover, higher water absorption and capillary sorptivity and lower ultrasonic pulse velocity, bulk resistivity, and abrasion resistance were reported during RCA replacement. Meanwhile, incorporation of steel fibers densified the concrete and enhanced its resistance to abrasive forces, water permeation, and water transport. Analytical regression models were developed to correlate hardened concrete properties to the 28-day cylinder compressive strength.

The recent trend toward increasingly deep convolutional neural networks (CNNs) leads to a higher demand of computational power and memory storage. Consequently, the deployment of CNNs in hardware has become more challenging. In this paper, we propose an Intra-Kernel Regular (IKR) pruning scheme to reduce the size and computational complexity of the CNNs by removing redundant weights at a fine-grained level. Unlike other pruning methods such as Fine-Grained pruning, IKR pruning maintains regular kernel structures that are exploitable in a hardware accelerator. Experimental results demonstrate up to 10× parameter reduction and 7× computational reduction at a cost of less than 1% degradation in accuracy versus the un-pruned case.

No abstract

No abstract

No abstract