Date: Thu, 21 Nov 1996 20:19:19 GMT Server: NCSA/1.4.2 Content-type: text/html Last-modified: Sat, 27 Jul 1996 06:00:13 GMT Content-length: 5742
|
Department of Computer Science and Engineering University of Washington Box 352350 Seattle, WA 98195
|
I am interested in the problem of safety-critical software. We are all aware of devices, systems, and structures that can cause catastrophes when they fail. As computer hardware becomes less expensive and as more expectations are placed on software, computers are being placed in control of a wider range of these applications. Software has advantages over conventional technologies, e.g., flexibility. But these advantages come at a price. Software behavior can be very complex and unpredictable. Perhaps the best publicized example is the Three Mile Island incident, in which operators had great difficulty in diagnosing the state of the system during an emergency.The requirements stage of software development is key to a project's success, especially with respect to safety. For that reason, my colleagues and I have concentrated on requirements, especially the communication of requirements via a requirements specification. To that end we have developed a state-based language called Requirements State Machine Language, or RSML. We have validated the usefulness of this language by specifying TCAS II, an avionics system.
For my doctoral thesis I invented a hazard analysis procedure based on the Hazard and Operability (HAZOP) study. The most significant concept that this new procedure borrows from HAZOP is that of a deviation, hence the name deviation analysis. See below for a link to an HTML transcription of the dissertation.