No abstract

Deterministic parallelism has become an increasingly attractive concept: a deterministic parallel program may be easier to construct, debug, understand, and maintain. However, there exist many different definitions of "determinism" for parallel programming. Many existing definitions have not yet been fully formalized, and the relationships among these definitions are still unclear. We argue that formalism is needed, and that history-based operational semanticsas used, for example, to define the Java and C++ memory models-provides a useful lens through which to view the notion of determinism. As a first step, we suggest several history-based definitions of determinism. We discuss some of their comparative advantages, note containment relationships among them, and identify programming idioms that support them. We also propose directions for future work.

Determinism is an appealing property for parallel programs, as it simplifies understanding, reasoning and debugging. It is particularly appealing in dynamic (scripting) languages, where ease of programming is a dominant design goal. Some existing parallel languages use the type system to enforce determinism statically, but this is not generally practical for dynamic languages. In this paper, we describe how determinism can be obtained-and dynamically enforced/verified-for appropriate extensions to a parallel scripting language. Specifically, we introduce the constructs of Determinis-tic Parallel Ruby (DPR), together with a run-time system (TARDIS) that verifies properties required for determinism, including correct usage of reductions and commutative operators, and the mutual independence (data-race freedom) of concurrent tasks. Experimental results confirm that DPR can provide scalable performance on mul-ticore machines and that the overhead of TARDIS is low enough for practical testing. In particular, TARDIS significantly outperforms alternative data-race detectors with comparable functionality. We conclude with a discussion of future directions in the dynamic enforcement of determinism.

No abstract

Fault diagnosis of the planetary gearbox (PGB) of wind turbines (WTs) plays an important role in the normal operation of WTs. Current studies commonly focus on the diagnosis of fault types of WT PGBs. Nevertheless, in addition to identifying the fault type, the current severity of the fault is also instructive for the maintenance and repair of WT PGBs. Thus, a novel optimized stacked diagnosis structure (OSDS) is proposed for the identification of fault type and severity. Compressed sensing is adopted to implement the compressed sampling of original vibration signals collected by the wireless sensor. Then, the compressed samples are input into first- and second-layer deep belief networks (DBNs) for the separate identification of fault type and severity. In order to realize the best feature extraction performance of DBNs, every single DBN in the OSDS is optimized with the chaotic quantum particle swarm optimization (CQPSO) algorithm. For OSDS, which is a hierarchical diagnosis system, the misdiagnosis results of the first layer will bring irreversible influence to the diagnosis of the second layer. That is to say, an incorrect fault type diagnosis will mean that these signals are wrongly classified, making them unable to judge the severity of the fault. Because the first-layer DBN is optimized with PGB historical data and the CQPSO algorithm, it shows an excellent performance in identifying fault types. Therefore, the diagnostic performance of OSDS has not been affected by the absence of diagnosis, and still shows an excellent recognition performance of fault type and severity in the experiment. This verifies its excellent role in the fault diagnosis of WT PGBs.