JOURNAL OF STRUCTURAL ENGINEERING / APRIL 2000 / 445 RETROFIT OF PRE-NORTHRIDGE MOMENT-RESISTING CONNECTIONS By Scott A. Civjan, 1 P.E., Michael D. Engelhardt, 2 P.E., and John L. Gross, 3 P.E., Members, ASCE ABSTRACT: An experimental program was undertaken to evaluate methods to retrofit existing steel moment connections for improved seismic performance. Six full-scale subassemblages were tested under cyclic loading. Typical pre-Northridge connections were retrofit by the addition of either a bottom flange dogbone or a welded bottom flange haunch. Retrofitted specimens were tested both with and without a composite floor slab. The tests showed poor performance of the bottom flange dogbone retrofit when the existing low toughness welds were left in place. Somewhat improved performance was observed when the bottom flange dogbone was combined with replacement of beam flange groove welds with higher toughness welds. The welded bottom haunch retrofit showed excellent performance on specimens with a composite slab, even though the existing low toughness beam flange groove welds were left in place. The presence of the composite slab appeared to help prevent fracture of the existing top flange weld in the haunch retrofit. This paper provides a summary of the complete experimental program. INTRODUCTION Damage to steel moment resisting frames (SMRFs) due to the 1994 Northridge California earthquake was extensive, in- dicating that the typical SMRF welded flange-bolted web con- nection detail used from the early 1970s had inherent perfor- mance problems. The implications of this damage are far reaching, as the connection detail was used throughout Cali- fornia and other seismically active areas for over twenty years. After the Northridge earthquake, damage reports indicated that prevalence of bottom flange weld fractures in the absence of inelastic beam deformations (Youssef et al. 1995). Major factors contributing to observed damage include the low frac- ture toughness of pre-Northridge connection welds, the stress concentrations and general flange overstress present in the typ- ical connection, the influence of leaving steel backing bars in place (leading to stress concentrations, possible weld inclu- sions, and added difficulty inspecting the welds), the stop-and- start bottom flange weld procedure, and the presence of a com- posite slab (Proceedings 1994; Yang and Popov 1995; Miller 1996; Sabol and Engelhardt 1996; Kaufmann et al. 1997; Ojdrovic and Zarghamee 1997). Post-Northridge testing of pre-Northridge connection details corroborated the damage observed in the Northridge earth- quake (Engelhardt and Sabol 1994; Yang and Popov 1995; Shuey and Engelhardt 1996; Uang and Bondad 1996). Suc- cessful solutions for new construction design have been tested, including flange cover plate, ribbed connection, haunch, dog- bone designs and many others (Engelhardt and Sabol 1994; Plumier 1995; Uang and Noel 1995; Chen et al. 1996; Engel- hardt et al. 1996; Iwankiw and Carter 1996; Shuey and En- gelhardt 1996; Uang and Bondad 1996; Tremblay et al. 1997; Xue et al. 1997). These improved connections utilized higher toughness electrodes and other weld improvements, in addition to design modifications. Relatively little testing has studied effects of a composite slab on steel moment connection behavior under cyclic loads. 1 Asst. Prof., Univ. of Massachusetts at Amherst, Dept. of Civ. and Envir. Engrg., 139 Marston Hall, Box 35205, Amherst, MA 01003–5205. 2 Assoc. Prof., Univ. of Texas at Austin, Dept. of Civ. Engrg., Phil M. Ferguson Engrg. Lab., 10100 Burnet Rd., Bldg. 177, Austin, TX 78758. 3 Nat. Inst. of Standards and Technol., Struct. Div., 100 Bureau Dr., Stop 8611, Gaithersburg, MD 20899–8611. Note. Associate Editor: Brad Cross. Discussion open until September 1, 2000. To extend the closing date one month, a written request must be filed with the ASCE Manager of Journals. The manuscript for this paper was submitted for review and possible publication on July 1, 1999. This paper is part of the Journal of Structural Engineering, Vol. 126, No. 4, April, 2000. ASCE, ISSN 0733-9445/00/0004-0445–0452/$8.00 + $.50 per page. Paper No. 21298. Most testing of composite specimens to date has included a fully composite slab with significant amounts of slab rein- forcement. When subjected to cycled positive moments, the slab contribution under positive moment has been determined as a compression zone emanating from the face of the column flange. The compressive stress acting at this location was re- ported as 1.3 (Du Plessis and Daniels 1972), although a f ' c specimen subjected to reversed cyclic loads did not reach the estimated moment capacity based on this assumption (Lee and Lu 1989). Reversed cyclic testing of a full frame indicated a compressive stress at the column face of 1.8 Tagawa et al. f ' c 1989). These previously tested specimens were fully compos- ite with respect to gravity load conditions. A bare steel and two ‘‘partially’’ composite specimens were tested by Leon et al. (1998). Bottom weld failures occurred in all three speci- mens. Previous connection testing has concentrated on pre-North- ridge connections, new construction designs, and repair meth- ods. Less testing has been directed towards retrofit methods for existing moment connections. In addition, the presence and influence of a concrete slab is not fully understood for the case of a laterally loaded structure. This project was therefore un- dertaken to investigate the effects of relatively inexpensive and nonintrusive retrofit procedures on connection performance. The influence of a building slab on the connection retrofit was also studied. Additional information on the project can be found in Civjan (1998). This testing program was part of a larger research program on retrofit of existing steel moment connections coordinated by the National Institute of Standards and Technology. Details of the complete program can be found in Gross et al. (1999). As part of this larger coordinated program, additional tests on dogbone and haunch retrofit techniques were also conducted at the University of California at San Diego, as reported by Uang et al. (1998). EXPERIMENTAL TEST SETUP AND SPECIMENS Tests were performed on full-sized interior joint subassem- blages. Points of inflection were assumed at column story midheights and at beam midspans. A typical story story height and beam span were assumed. The overall test frame sche- matics can be seen in Fig. 1. The test specimens were chosen to be representative of building construction details in common use prior to the Northridge Earthquake and not to duplicate specimens inves- tigated elsewhere. Beams were W30  99 sections of A36 steel. Columns were W12  279 sections of A572 Grade 50 steel to provide strong column, weak beam action and to pro-