Bending tests are by far the most widely used in order to evaluate creep deformations in cracked conditions. The testing procedure is normally similar to the one described in Sect. 3 with reference to uniaxial tests (see Fig. 2). The main difference is, of course, the crack opening or deformation measure being used in the pre-cracking and long-terms testing phases (e.g. CMOD, CTOD or mid-span displacement). Typical setups make use of lever systems in order to maintain constant loads for the whole duration of the tests . Figure 3 shows the setup used by the Authors and derived from . Daviau-Desnoyers et al.  adapted a hydraulic loading frame for creep tests in order to perform long-term bending tests. Kurtz and Balaguru  used a setup in which clamped specimens are loaded by a lever in a cantilever behaviour. Buratti and Mazzotti  carried out tests using 200 cm long beams with free hanging weights in a four-point bending setup. In the literature it is possible to find examples of tests on individual specimens and of tests on piles of specimens (normally 3) stacked one on top of each other (e.g. [5, 8, 40]). In this case the load on the specimens is obviously larger at the bottom of the pile, but
Fig. 3 Experimental setup used by the authors for long-term bending test 
often specimens can be ordered in such a way that the ratio of applied force versus residual strength at pre-cracking is constant. This is typically easier on MSFRC specimens because of smaller variability in the post peak behaviour.
In most of the tests published notched specimens were used [4, 6-8, 12, 40, 41] but examples of tests on un-notched specimens can be found in the literature [1, 5]. The most typical setup adopted while testing staked specimens is four-point bending, for obvious stability advantages. One of the shortcoming of this setup is that it is not consistent with the short-term testing procedure proposed by RILEM  which is typically used in the pre-cracking stage. However, Zerbino et al.  showed that use of three or four-point loading configuration in bending does not affect creep tests results in terms of the COD rate or the stress levels where stable creep behaviour takes place.
In the literature concerning long-term bending test there is not yet agreement on the extent to which some parameters might influence the test results. The Creep load ratio (see Sect. 3) is very often set to 50 % of the residual strength measured at the end of pre-cracking. Many results seem to indicate that creep load ratios between 60 and 70 % might represent an upper limit to the long-term load (see Sect. 1). Defining the load ratio as a percentage of the residual strength at one point might be non-representative for FRCs with strong softening or hardening behaviour. In these cases, in fact, the ratio of creep load versus actual residual strength during the long-term tests might change significantly because of the long-term crack opening. Other criteria were used by some researchers (e.g. ) but they are very limited in the literature on the topic. Another important parameter is the crack opening used in the pre-cracking stage, which controls the damage level at the beginning of the long-term tests. According to the previous considerations on the load ratio it might be important to consider, together with the serviceability behaviour, the shape of the stress crack opening curve in the post-peak region in order to define this parameter. Other factors that affect the long term behaviour are environmental conditions, i.e. temperature and relative humidity. Buratti and Mazzotti  found that moderate temperature variations could trigger tertiary creep failures in MSFRC specimens. Another factor on which there is no wide agreement in the literature is the minimum test duration. Zerbino et al.  suggested that the duration should be around 90 days and proposed to use the 30-90 days crack opening rate, defined as the secant of the COD versus time curve from 30 to 90 days, as synthetic parameter for characterizing the behaviour of different FRCs. Comparing the durations published bending tests of FRCs with typical creep tests in compression these latter are much longer, if fact they have normally a duration of at least 6-12 months. Test duration might be critical because Kusterle observed tertiary creep failures after years of testing . Zerbino et al.  found that performing loading-unloading cycles does not contribute to the reduction of testing time. Daviau-Desnoyers et al.  performed loading-unloading cycles during long-term tests in order to evaluate the progression of damage based on the observed compliance of the specimens.
Finally, another point on which different authors adopted different approaches is related to the deformation parameters to be measured during the tests. Typical examples are CMOD, CTOD and mid-span deflection. Relationships among those parameters (in particular crack opening and deflection) for long-term loads are not yet fully defined and therefore it might be complicated to compare tests in which different parameters were used. As an example, the creep coefficient in terms of mid-span deflection is normally larger than the one in terms of CMOD  because the former deformation parameter is affected not only by the delayed behaviour of the cracked cross-section but also by the creep of the un-cracked parts of the specimen.