ABSTRACT The design of TLP tendons and tendon connectors to resist fatigue failures requires consideration of many factors, such as mean and alternating stresses, cycles, welding, residual stresses, base metal fatigue, and weld profile control. Many assumptions are necessary to do this. This paper describes the development of a machine which can validate these assumptions by testing full scale sections of tendons to fatigue loadings similar to those seen in service. INTRODUCTION One of the significant design considerations of Tension Leg Platforms is potential fatigue failures of the tendons. [1]* Tendons have a high mean tension and experience a large number of primarily axial tension variations (bending is also present) throughout the design life. In addition, all tendons have welds which have a lower fatigue strength than the base metal. Many tendon designs also incorporate connectors which may result in additional stress concentrations. Unfortunately, there is a relatively large amount of uncertainty in fatigue design. To account for this, API RP 2T [1] recommends that tendon components be designed for a life of 10 times the tendon design service life and this recommendation was followed in the Conoco Jolliet TLWP. [7] There has been a considerable amount of work done in developing and fatigue testing connectors for TLP tendons [3,4,5,6], but this work has been primarily analysis and testing at scales of 1/3 and less. Aside from mean stress, alternating stress, and number of cycles (the effects of these variables are customarily accounted for in the fatigue curves), there are a number of factors which influence fatigue life. These are:Metallurgy of base metalWelds and post weld stress relief of residual stressesProfile control of weldsStress concentrations at connectors and weldsSize of structure and amount of material actually stressedCorrosion/stress corrosion; speed of cycles With the exception of item 6, which is accounted for by cathodic protection, these factors are considered in design by accurate calculation of peak stresses and selection of a fatigue curve which accounts for items 13. Item 5, the size of the structure, is accounted for by the fact that the fatigue curve is a lower bound to actual test data taken from a large number of small test coupons. However, this process has a number of assumptions and approximations which it may be prudent to verify based on the large amount of welds and connectors in tendons which are subject to fatigue and the relatively small amount of actual test data. For example, a TLP in 3000' of water with twelve tendons and a connector every 80' will have approximately 450 connectors and 900 circumferential butt welds to connect the connector to the base pipe. If a tendon has 100 square inches of cross sectional area and 1" long welds, there are about 90,000 cubic inches of weld metal in all the tendons. Most fatigue data is taken with specimens which are on the order of 0.25" diameter and with a cross sectional area of about 0.05 square inches.