An expert system approach for generating test sequence for CTCS-3 train control system

As a unified technology platform for high speed railway in China, CTCS-3 train control system can ensure the safety and high efficiency of train operation. In order to verify whether a CTCS-3 system is consistent with the system requirements specification, it is necessary to carry out a series of testing, including laboratory testing, field testing, integrated testing and commissioning, as well as interoperability testing, for whose purposes a set of test sequences must be prepared. The manual preparation of test sequence is characteristic of heavy work load, low efficiency and high demand for professional expertise. Although a computer-aided tool has been developed to provide an integrated graphical editing environment, with such functions as automatic generation of WORD test sequence files and XML test scripts, as well as electronic management of test cases, test sub-sequences and test sequences, however its intelligence and flexibility is limited because human efforts and expertise are stilled needed in such works as preparing pre-defined test sub-sequences, choosing proper test sub-sequences and concatenating them to form a test sequence, as well as setting up various test conditions. To address these issues, this paper proposes a new approach for generating the test sequence by building an expert system, with the aim to increase the degree of automation in concatenating test cases, determining test location and setting up test condition for each test case. Firstly on the basis of analyzing the process of manually generating test sequence, the related domain knowledge are summarized; secondly the knowledge tree method is adopted for compiling the knowledge about test cases and their test method, the typical test scenarios and some rules of thumb; thirdly production rules are adopted to represent the knowledge and relational database is utilized to build the knowledge base; finally a hybrid inference mechanism by combining model-based forward inference and rule-based backward inference is adopted to generate the test sequence.