Welding-induced corrosion and protective measures for clad rebars in neutral chloride environments.

Corrosion-resistant steel plays a vital role in marine steel structures. This study developed an SS304/HRB400 stainless-steel-clad rebar for application in a cross-sea bridge in Zhejiang Province. CO2 gas shielded welding was employed in the prefabricated steel structure, with SS304 steel as the welding wire. This study investigated the welding on the corrosion resistance of clad rebars and explored corrosion protection measures for welded joints.The results indicated that refined grains appeared in both stainless steel and carbon steel due to distinct dynamic recrystallization (DRX) during welding. The corrosion resistance, as determined by potentiodynamic polarization curve analysis of the material's interaction with the solution ranked as follows: clad rebar (polished) > clad rebar welding (CRW) > painting the clad rebar after welding (PCRW) > clad rebar (unpolished) > carbon-steel welding (CSW) > carbon-steel bar > cold spraying zinc after clad rebar welding (ZCRW). However, an accelerated corrosion test with four samples for 600 s with a corrosion current of 0.8 A revealed minimal corrosion damage on zinc-coated surfaces. Hence, welding joints for clad steel structures are considered feasible and must be subject to cold zinc spraying after polishing to enhance their corrosion resistance.

Interface Microstructure and Properties of Vacuum-Hot-Rolled 55#/316L Clad Rebars.

The existing process for the preparation of cladded rebars is too complicated for large-scale industrial production. Therefore, this paper proposes a 55#/316L rebar preparation method based on vacuum hot rolling. The microstructure and mechanical properties of the composite interface of the rebar, along with the connecting technique, were studied using transmission electron microscopy, X-ray diffraction, and Vickers hardness testing. The obtained results showed that the minimum thickness of the 55#/316L rebar cladding was 0.25 mm, which was twice that of the M 329M/M 329-11 design standard used in the United States of America. Due to the diffusion of carbon, large numbers of second-phase particles were precipitated on the stainless-steel side, which resulted in intergranular chromium depletion. After multi-pass hot rolling, the minimum bonding strength of the composite interface reached 316.58 MPa, which was considerably higher than the specified value of 210 MPa. In addition, we designed three different types of rebar connection joints: sleeve, groove-welded, and bar-welded. According to the tensile test, the bar-welded joint had higher yield strength (385 MPa) and tensile strength (665 MPa) than the base rebar (376.6 MPa and 655 MPa), as well as a very high corrosion resistance.

A Study on the Acquirement Method of Mechanical Property Parameters in the Different Base Materials Composite Region of Clad Rebar.

Clad rebar is one of the key structures of marine and construction services. Therefore, it is of great importance to acknowledge the mechanical property parameters of the composite region in the structural integrity evaluation of clad rebar. The different base materials of clad rebar (20MnSiV/316L steel, 35#/316L steel, 45#/316L steel, and 55#/316L steel) are researched in this study. The composite area is further refined, and simultaneously, a refinement model of the composite region of clad rebar is established. In view of the fact that a surface hardness experiment is quite easy to conduct, a proposed method consists of obtaining the mechanical property parameters of materials using the surface hardness test. The mechanical property parameters are acquired; moreover, the relationship between yield stress and surface hardness of the stainless steel clad rebar is set up. We used this method to acquire the mechanical parameters of a composite surface uneven area of clad rebar, and we established a mechanical parameters mathematics model of clad rebar, it is a significant basis for a structural integrity evaluation of cladding materials.

Profile Change Law of Clad Rebars and the Formation Mechanism of Composite Interfaces during Hot Rolling.

Rough- and intermediate-rolled composite billets and finished clad rebars were cut using flying shears. The law of metal rheology and the mechanism of composite interface generation during clad rebar formation were then investigated using metallographic microscopy, electron backscatter diffraction, and scanning electron microscopy. The radial deformation trend of the clad rebars was greater than that of HRB400 rebars and "ears" were more likely to appear during the rolling process. The widths of the decarburization and composite zones and diffusion distances of each element decreased as the cumulative reduction rate increased. Furthermore, as deformation increased, the number of oxides on the composite interface significantly decreased, the proportion of recrystallized grains increased, and the grains became more refined. These changes led to increases in the bond and tensile strengths of the composite interface. According to the research above, the pass filling degree should be within 0.85-0.9 and the cumulative reduction rate should be over 80% when rolling clad rebars.

Non-Structured DNN Weight Pruning-Is It Beneficial in Any Platform?

Large deep neural network (DNN) models pose the key challenge to energy efficiency due to the significantly higher energy consumption of off-chip DRAM accesses than arithmetic or SRAM operations. It motivates the intensive research on model compression with two main approaches. Weight pruning leverages the redundancy in the number of weights and can be performed in a non-structured, which has higher flexibility and pruning rate but incurs index accesses due to irregular weights, or structured manner, which preserves the full matrix structure with a lower pruning rate. Weight quantization leverages the redundancy in the number of bits in weights. Compared to pruning, quantization is much more hardware-friendly and has become a "must-do" step for FPGA and ASIC implementations. Thus, any evaluation of the effectiveness of pruning should be on top of quantization. The key open question is, with quantization, what kind of pruning (non-structured versus structured) is most beneficial? This question is fundamental because the answer will determine the design aspects that we should really focus on to avoid the diminishing return of certain optimizations. This article provides a definitive answer to the question for the first time. First, we build ADMM-NN-S by extending and enhancing ADMM-NN, a recently proposed joint weight pruning and quantization framework, with the algorithmic supports for structured pruning, dynamic ADMM regulation, and masked mapping and retraining. Second, we develop a methodology for fair and fundamental comparison of non-structured and structured pruning in terms of both storage and computation efficiency. Our results show that ADMM-NN-S consistently outperforms the prior art: 1) it achieves 348× , 36× , and 8× overall weight pruning on LeNet-5, AlexNet, and ResNet-50, respectively, with (almost) zero accuracy loss and 2) we demonstrate the first fully binarized (for all layers) DNNs can be lossless in accuracy in many cases. These results provide a strong baseline and credibility of our study. Based on the proposed comparison framework, with the same accuracy and quantization, the results show that non-structured pruning is not competitive in terms of both storage and computation efficiency. Thus, we conclude that structured pruning has a greater potential compared to non-structured pruning. We encourage the community to focus on studying the DNN inference acceleration with structured sparsity.

GRIM: A General, Real-Time Deep Learning Inference Framework for Mobile Devices Based on Fine-Grained Structured Weight Sparsity.

It is appealing but challenging to achieve real-time deep neural network (DNN) inference on mobile devices, because even the powerful modern mobile devices are considered as "resource-constrained" when executing large-scale DNNs. It necessitates the sparse model inference via weight pruning, i.e., DNN weight sparsity, and it is desirable to design a new DNN weight sparsity scheme that can facilitate real-time inference on mobile devices while preserving a high sparse model accuracy. This paper designs a novel mobile inference acceleration framework GRIM that is General to both convolutional neural networks (CNNs) and recurrent neural networks (RNNs) and that achieves Real-time execution and high accuracy, leveraging fine-grained structured sparse model Inference and compiler optimizations for Mobiles. We start by proposing a new fine-grained structured sparsity scheme through the Block-based Column-Row (BCR) pruning. Based on this new fine-grained structured sparsity, our GRIM framework consists of two parts: (a) the compiler optimization and code generation for real-time mobile inference; and (b) the BCR pruning optimizations for determining pruning hyperparameters and performing weight pruning. We compare GRIM with Alibaba MNN, TVM, TensorFlow-Lite, a sparse implementation based on CSR, PatDNN, and ESE (a representative FPGA inference acceleration framework for RNNs), and achieve up to 14.08× speedup.