Steel moment-resisting frames (MRFs) with strong column and weak beam characteristics are widely utilized in earthquake-prone countries. The seismic capacity of the frames is known to be significantly affected by the low-cycle fatigue behaviour of beam-to-column connections in MRFs. Several types of steel beam-to-column connections, e.g., bolted flange and web connections and welded connections, have been utilized in MRFs. However, many beam-to-column connections, especially welded connections, were fractured during the 1994 Northridge earthquake and the 1995 Kobe earthquake [1–4]. After these earthquakes occurred, significant steps were taken to improve the construction materials and characteristics of connections to prevent such fractures in the future. Moreover, structural steel with higher ductility and better welding characteristics was implemented, and improved beam-to-column connections were developed. For example, in the United States of America, the reduced beam section (RBS) beam system was developed [5, 6]. In an RBS moment connection, portions of the beam flanges are selectively trimmed in regions adjacent to the beam-to-column connection to induce plastic hinging within the reduced section, thereby reducing the likelihood of brittle fractures at the beam flange groove welds and surrounding base metal regions [7]. Furthermore, after the occurrence of the Kobe earthquake, one of the most significant improvements implemented in Japan was an alteration in the shape of weld access holes. Wide flange beam-to-rectangular hollow section column connections are commonly used in Japan to prevent the weak-axis problem mentioned in [8]. A detailed description of welded-flange-welded-web connections of this type has been presented in several works [9, 10]. The beam flange is attached to columns via through-diaphragms using full penetration welding, and the web is attached to the skin plates of columns in the absence of shear connection plates using fillet welding. Figure 1 depicts the pre-Kobe (conventional type R35) and post-Kobe (improved type R35 + R10) weld access holes [4]. Because of the smaller curved structures with a radius of only 10 mm, the new weld access holes prevent early fractures by decreasing the strain concentration at the flange. In recent years, the occurrences of brittle fractures at the connections have become rare. However, ductile fractures that happen after adequate plastic deformations occur remain one of the most typical failure modes for beam-to-column connections during earthquakes.
Damage evaluations of beam-to-column connections have been conducted for decades. The deformation amplitude and low-cycle fatigue life of steel beam-to-column connections have been reported to follow the Manson–Coffin relationship [11] and Miner’s rule [12] by Krawinkler and other researchers [13–18]. Other than classical methods, Iyama et al. [19] studied crack growth behaviours in welded beam-to-column connections by considering the effects of the stress triaxiality on critical plastic strains. In addition, some micromechanics fracture models, e.g., the void growth model and the stress-modified critical strain model, were introduced to predict ductile crack initiations under monotonic loadings [20, 21], whereas the cyclic void growth model [22] was adopted to predict the low-cycle fatigue behaviours of steel connections [23, 24]. Except considering crack developments, Jiao and Yamada et al. [25] developed a damage evaluation method of steel beam-to-column connections based on the decomposition of hysteresis loops of the structural members or materials. The plastic deformation capacity of the connections is evaluated by studying the relationship of the cumulative damage calculated from the skeleton curves and Bauschinger parts. Moreover, it is difficult to predict fractures based solely on the present analyses. All these abovementioned evaluation methods are empirical rules based on multiple component experiments and have relatively limited applicability. To cover important variables, a large number of experiments that require much time and effort would be necessary. Recently, evaluation methods based on both experimental and analytical methods were introduced [26, 27], where the influence of some parameters can be considered through analysis to reduce the number of experiments.
On the other hand, the strain amplitude at the critical cross-section is one of the key indices that affect the fracture of the material. Ono et al. [28] reported that, for the most used structural steel, the strain-life relationship can be expressed using a single Manson–Coffin equation when the strain amplitude lies between 0.2 and 8%. Furthermore, Yamada et al. [29] empirically demonstrated that the strain-life of structural steel follows a Manson–Coffin relationship over a wider strain range, ± 2 to ± 12%. At the component level, it is worth discussing whether there is a similar Manson-Coffin relationship between the average flange strain amplitude at the critical cross-section and the number of loading cycles until fractures occur. The deformation capacity of the beam-to-column connections can be evaluated based on the average flange strain at the critical cross-section if the above relationship is true. This evaluation is the main motivation of this study. To accomplish this goal, determining the strain history at the critical cross-section is necessary. However, it is difficult to obtain the local critical cross-section strain through experiments because large strain is likely to occur near the critical cross-section of the components, which would soon damage the strain gauges. Moreover, the accuracy of the strain gauges commonly used in strain measurements would decrease with an increasing number of loading cycles even before the damage occurred. Therefore, in the current study and in addition to experiments, analyses are also introduced to simulate component behaviours and to obtain the local strain at the critical section.
In the current study, the WF beam-to-RHS column connections with weld access holes on the beam web that are commonly used in Japan are selected as the subject to prove the viability of the abovementioned concept. For this type of connection, several variables affect the plastic deformation capacity [30–32]. Material characteristics are one of the main parameters that influence the strain history of the connection. Additionally, different member cross-sections would change the moment distribution in the beam-end flange and web. Therefore, the contribution of the flange and the strain at the beam-end flange are highly affected. Moreover, when the connections are subjected to cyclic loadings, ductile cracks initiated at the toes of the weld access holes are observed to propagate during loading, leading to the eventual fracture of the connection. Regarding connections with different shapes of weld access holes, the crack initiation locations and local strain history would also be different. The current study focuses on shop welded beam-to-column connections made of 400 N/mm2 class structural steel with conventional or improved types of weld access holes illustrated in Fig. 1. In the first phase of this study, cyclic in-plane beam analysis is performed based on several experiments where the beam-to-column connections were tested under constant amplitude cyclic loadings. With the strain histories at the toe of weld access holes that were obtained via analyses, the relationship between the average flange strain amplitude at the critical cross-section and the number of loading cycles until fractures occurred was discussed. The experimental and analytical results show that similar to the material level, the Manson-Coffin relationship between the average strain amplitude and the number of loading cycles until fractures occurred is confirmed on the component level. In the following phase, the results of previous beam-to-column connections loaded under variable amplitudes were studied. To expand the size of the database, experiments were conducted on two beam-to-column connections corresponding to random loading histories obtained via the response analysis of a three-story steel frame. As a result, the accuracy of the obtained Manson-Coffin relationship under random loading histories was verified with the experimental data. The findings of this study allow a damage evaluation method on typical Japanese shop-welded beam-to-column connections with analysis and a limited number of experiments.