Search

Platform atomics provide memory consistency and atomicity across the heterogeneous architecture. They allow the latency compute units (CPU cores) and the throughput compute units (graphics processing unit, GPU cores) to simultaneously access the same memory locations. This chapter describes three case studies of algorithm patterns that benefit from platform atomics. The first case study is a task queue system in which the CPU produces tasks that are processed by the GPU. It avoids kernel relaunch and ensures load balancing. The second case study implements the breadth-first search as an example of an application that identifies dynamically the most appropriate cores (CPU or GPU) to execute a task. Thus, program execution can occasionally swap between the CPU and the GPU, depending on workload characteristics. The third case study is a data layout conversion routine that simultaneously uses the CPU and the GPU to shift elements of an array in-place. It represents a pattern of close collaboration between the CPU and the GPU to process a pool of fine-grained tasks.

In order to help programs exploit the power of systems compatible with the HSA, compilers and runtimes of high-level programming languages or domain-specific languages should target HSA for portable code generation and runtime management. In this chapter, we use C++ AMP, a parallel programming extension to C++, to show how efficient HSAIL code can be generated from a higher-level programming model. We discuss important translation techniques, such as mapping between C++ constructs and HSA features, compilation processes for GPU and CPU codes, tiling support, memory segment annotation. In addition, we explain how to access HSA-specific features such as shared virtual memory and platform atomics from C++, and we cover how they are mapped to HSAIL. These techniques can be useful for compiler writers interested in mapping other programming models to HSA. The implementation, Kalmar, which consists of a compiler, a set of header files, and a runtime library, is publicly accessible as an open source project.

Autonomous systems are becoming more commonly used, especially in hazardous situations. Such systems are expected to make their own decisions about future actions when some capabilities degrade due to failures of their subsystems. Such decisions are made without human input, therefore they need to be well-informed in a short time when the situation is analysed and future consequences of the failure are estimated. The future planning of the mission should take account of the likelihood of mission failure. The reliability analysis for autonomous systems can be performed using the methodologies developed for phased mission analysis, where the causes of failure for each phase in the mission can be expressed by fault trees. Unmanned Autonomous Vehicles (UAVs) are of a particular interest in the aeronautical industry, where it is a long term ambition to operate them routinely in civil airspace. Safety is the main requirement for the UAV operation and the calculation of failure probability of each phase and the overall mission is the topic of this paper. When components or sub-systems fail or environmental conditions throughout the mission change, these changes can affect the future mission. The new proposed methodology takes into account the available diagnostics data and is used to predict future capabilities of the UAV in real-time. Since this methodology is based on the efficient BDD method, the quickly provided advice can be used in making decisions. When failures occur appropriate actions are required in order to preserve safety of the autonomous vehicle. The overall decision making strategy for autonomous vehicles is explained in this paper. Some limitations of the methodology are discussed and further improvements are presented based on experimental results.

A number of accident databases have been developed including both public and commercial ones. However, these databases implement static data models that are used to classify each accident. It is then difficult to dynamically expand the underlying database schema if for example new causes are identified. In this paper we propose an approach for accident database development based on knowledge engineering techniques. In particular, we explore the use of ontologies and Formal Concept Analysis.

The VIRTHUALIS project aims to develop a number of virtual reality applications for improving safety in the process industries. The applications allow human factors experts to study how operators interact with plant, and provide a safe environment in which new safety actions can be tried and tested. Safety applications are built on the SafeVR technology platform, a distributed clientserver virtual reality system. This paper describes how two external modules - a process simulator and a rule-based system - are interfaced to the platform and the benefits they provide both separately and together. The two modules communicate with the platform's server by exchanging messages, conforming to a simple syntax. pSimProxy provides a generic interface to an external process simulator, which in turn delivers the realistic plant behaviour. It handles bidirectional data exchange with and control of the external simulator. It can be configured at run time to use whichever available mechanisms are supported by the actual process simulator that models the plant being simulated. ClipsClient is an expert or rule-based system, based on NASA's CLIPS expert system software that can make inferences about the information contained in the messages. It consists of a set of facts, a number of rules and an inference engine. It can be provided with a number of rules that monitor how operators are running the plant, and react in useful ways to these events. The simulator notifies the server of changes in process parameters through a message. The values may be displayed, for example as gauge readings, in the virtual environment. As operators control the plant, their actions, say opening a valve, are also reported by messages via the server to the process simulation. Messages can also be read by the rule-based system, allowing it to maintain its own representation of the plant. This in turn permits automated expert reasoning on the state of the plant and the actions of its operators which can cause further message to be sent to the server. The rule-based system is therefore, a powerful mechanism for rapidly reconfiguring the application and general rules can be written that only require new facts at run-time to change the behaviour of the entire virtual environment. The message syntaxes, the system architecture and the interfacing of the external modules are described along with examples showing their individual and joint benefits.

The rapid expansion of the optical fiber telecommunication industry due to the explosion of the Internet has substantially driven down the cost of optical components, making fiber optic sensors more economically viable. In addition, the rapid development of fiber-optic sensors, particularly the fiber Bragg grating (FBG) sensors offers many advantages and capability that could not be achieved otherwise. In the past few years, fiber Bragg grating sensors have attracted a lot of interest and they are being used in numerous applications. This paper describes the FBG sensors developed for structural health monitoring, and were installed on Hong Kong's landmark Tsing Ma bridge (TMB), which is the world longest (1377 m) suspension bridge that carried both railway and regular road traffic. Forty FBG sensors divided into three arrays were installed on the hanger cable, rocker bearing and truss girders of the TMB. The objectives of the study are to investigate the feasibility of using the developed FBG sensors for structural health monitoring, via monitoring the strain of different parts of the TMB under both the railway and highway loads as well as comparing the FBG sensors' performance with the conventional structural health monitoring system - Wind and Structural Health Monitoring System (WASHMS) that has been operating at TMB since the bridge's commissioning in May 1997. The experimental observations in this project show that the results using FBG sensors were in excellent agreement with those acquired by WASHMS. © 2005 Elsevier Ltd. All rights reserved.

There is an enormous amount of information available on past accidents in the form of incident reports available as documents and managed by databases. Engineers who perform safety analysis can benefit from this information. However, incident reports and accident databases are written the form of textual natural language descriptions. This paper presents a knowledge-based approach to extract knowledge directly from such sources while reducing the number of mismatches.

In the process industries batch plants are attracting attentions because of their suitability for producing small-volume, high-value added commodity chemicals. Pipeless plants have also been developed and built to increase plant flexibility. Unexpected events, such as the failure of a processing unit, sometimes happen during operations. To avoid risk and to utilise the remaining resource, it is important to reschedule the production operation quickly. The constraint model in BPS has been extended to include constraints for rescheduling. These additional constraints are described in this paper and a case study is used to demonstrate the feasibility of the approach.

This chapter focuses on the hazard identification aspect of safety assessment and discusses hazard and operability (HAZOP) studies applied to computer-controlled process plants. In the process industry, HAZOP is a long-established methodology used for identifying hazards in chemical plant design. Some attempts have been made to modify conventional HAZOP for computer-related systems. Modified versions of HAZOP are generally referred to as CHAZOP (computer HAZOP) or PES (programmable electronic systems) HAZOP in the literature. This chapter provides a brief description of the conventional HAZOP as used in the process industry and an overview of the different CHAZOP frameworks/guidelines suggested by engineers and researchers over the past few years. The overview shows that there is so far no agreed format on how CHAZOP should be done and that the different approaches were made ad hoc. The chapter presents a new CHAZOP methodology, which is systematically developed and based on incident analysis. The chapter discusses the strategy used to develop the methodology and illustrates the application of the methodology with examples.