یکپارچه سازی داده کاوی با استفاده از روش KJ برای دسته بندی عیوب ساخت و ساز پل

This paper tries to analyze common bridge construction defects, classify them into appropriate groups, and redefine them as a precautionary measure and means to improve quality in bridge construction. For this purpose, data on bridge construction since January 2007 were obtained from the evaluation report of the Public Construction Committee (PCC) of Taiwan. Bridge construction defects were classified according to their characteristics. A constraint-based clustering method and affinity diagram (KJ method) are proposed and used. This method can simultaneously treat mixed data types; moreover, it can incorporate user-specified constraints. The quality or safety issues, the unit-in-charge (Government authorities/project owners/contractor), and the properties of the defects (construction/audit/documents/others) are the sorting attributes. The constraint is avoiding empty clusters or clusters having very few objects. The results revealed five major defect classifications: safety and environment, construction site defects, supervision/control process, construction quality documents, and others.

Construction defects are always a major concern in infrastructure projects. These defects directly influence the quality of the construction and may result in potential injury to the people. For example, the main 458-foot span of the 35-W interstate bridge that collapsed into the Mississippi river in Minneapolis; Part of the Big Nickel Road Bridge collapsed onto the roadway below it during construction in Canada. If constructors ensure construction quality, such disasters can be reduced. Therefore, the critical issue here is how recurring defects in such constructions can be avoided. Graves (1993) proposed practical prevention instead of correction, the collection and communication of data graphically to track continuous improvement, and the use of a plan-do-check-act cycle and pilot projects to test and refine processes. Thus, it can be seen that an effective defect feedback system can prevent recurring defects, thereby improving construction quality. In Taiwan, the Public Construction Committee (PCC) is a body that audits important construction projects each season. The evaluation reports compiled by the PCC describe many defects in bridge construction processes; these descriptions, however, are miscellaneous and disordered. Therefore, it is very important to arrange and classify common defects, translate them into useful information, and then enable project participants to understand the critical defects so that their recurrence can be prevented. With this background, this paper tries to integrate the KJ method (affinity diagram) and constrained k-prototypes (CKP) to analyze common bridge construction defects, classify them into appropriate groups, and redefine them as a precautionary measure and means to improve the quality of bridge construction. The integrated process is used to analyze the results of the PCC audit report and inspect the operation of the quality management system for public construction projects in Taiwan. Through this process, a general list of defects in bridge construction operations can be built and recurring defects can be prevented.

In order to prevent recurring construction defects, an effective defect feedback system is necessary. In this paper, an effective constraint-based clustering algorithm CKP was proposed. Although this algorithm has some shortcomings, it can simultaneously treat mixed data types; moreover, it can incorporate user-specified constraints. Besides, the CKP was integrated with the KJ method to analyze common bridge construction defects and classify them into five major groups: safety and environment, construction site defects, supervision/control process, construction quality documents, and others. In these five important groups, the percentage of construction site defects is the highest, approximately 33%. Construction site defects include some important issues such as backfill, structure and plan report. The probable factors influencing these issues include supervision by contractor, supplier involvement, personnel, technologies used, and so on. Contractors must remain alert about these issues. The construction quality document is the second most important group; its percentage is approximately 29%. Each unit-in-charge must pay attention to the document format, the content of the plan report, and carry out relational plans and maintain records. The contents of the remains groups are presented in the relational table. An effective feedback system can prevent recurring defects, thereby improving construction quality. In this paper, common bridge construction defects were analyzed, classified into appropriate groups, and redefined. Each unit-in-charge can refer to these common defects and prevent their recurrence, thereby improving quality in bridge construction.