The PV systems are generally designed based on JIS C 8955 in Japan. However, no provision is given to the PV panels installed in the edge zones. It is not recommended to install them in such zones, because very large suctions are induced by flow separation. When installing PV panels in such zones, we have to evaluate the wind force coefficients appropriately. Because it is difficult to make wind tunnel models of the PV panels with the same geometric scale as that for the building, we apply a numerical simulation to evaluate the net wind forces on the PV panels, where the pressures beneath the PV panels, called ‘layer pressures’, are numerically simulated using the unsteady Bernoulli equation together with the time history of external pressures on the roof measured in a wind tunnel. Furthermore, we propose to install the PV panels with small gaps between them, which may reduce the net wind forces acting on the PV panels due to pressure equalization. It is expected that the external pressures on the roofing are also reduced significantly.

　The present paper consists of six sections. Following the introduction, Section 2 illustrates the building models under consideration. Two models are used; one is a square-roof building with a square plan of 11 m × 11 m and the other is a hip-roof building with a rectangular plan of 18.6 m × 9.4 m. The ridge height and the roof slope are respectively 10.7m and approximately 25° in both models. Section 3 describes the experimental models and procedure used in the wind tunnel experiment. In the square-roof model case the PV panels installed on the roof are also modeled. The wind pressures at many locations on the top surface of PV panels as well as on the roof surface are measured in a wind tunnel, the results of which are used for validating the simulation method of layer pressures. The hip-roof model is not equipped with PV panels. The wind pressures measured on the roof are used for simulating the layer pressures. Section 4 first explains the method of simulation. The layer pressures are numerically simulated by using the unsteady Bernoulli equation together with the time history of external pressures on the roof. The net wind forces on the PV panels are provided by the difference between the wind pressure on the upper surface and the layer pressure. The simulation is verified by comparing the simulated results of the layer pressures with those obtained from the wind tunnel experiment for the square-roof model. Then, the simulation is applied to the hip-roof building model in Section 5. The wind force coefficients of PV panels installed all over the roof are evaluated. The results indicate that negative wind force coefficients large in magnitude are induced on the PV panels located near the roof corner. However, the magnitude of such negative wind force coefficients decreases with an increase in the gap between PV panels up to approximately 5 mm. When PV panels are installed on the roof, the external pressures on the roofing are equal to the layer pressures. It is found that the peak external pressures on the roofing are significantly reduced in magnitude by the PV panels. Finally, Section 6 summarizes the main conclusions obtained in the present study.

　The tuned viscous mass damper (TVMD) and its optimum designing method based on the fixed-point theorem have been proposed by Ikago et al. The validity of the damper system as a supplemental energy dissipation device has also been proved analytically and experimentally alike. However, analyses revealed that an installation of inertial mass in the viscous mass damper, and an increase in input ground motions might result in excessive stresses on the supporting member, the damper body, and the primary structure connected to the damper system. Kida et al. then developed the rotation-slipping mechanism to limit the maximum damper force up to a given criterion, such as yield strength, and calling it the “force restriction mechanism.” The TVMD connected with the force restriction mechanism is called force-restricted tuned viscous mass damper (FRTVMD), and the FRTVMD is modeled with a TVMD and a coulomb slider connected in series. Although the effectivity of the FRTVMD is proven experimentally, there are no theoretical backups shown so far due to the complex non-linear response property of the coulomb slider.

　This paper firstly presents a way to equivalently-linearize the coulomb slider stochastically without repetitive numerical computations. In the process of linearization, an RMS of the damper force under a white noise excitation is utilized. Numerical examples show the validity of the linearization method under non-stationary ground motions.

　This paper then presents the performance evaluation of the FRTVMD from total energy dissipation in the damper system and the optimum design variable of the force restriction mechanism using the area of an energy transfer function. The energy transfer function is an expression of input energy to an energy-dissipation element in the frequency domain, and the area of it indicates the total input energy under a white noise excitation. Considering that the FRTVMD dissipates energy in two-part, the dashpot and the coulomb slider, maximization of the total damper dissipation energy may minimize the energy input to the primary structure, if the total earthquake input energy depends little on the damper specifications. Response analyses under a couple of recorded ground motions lastly show the effectiveness of the proposed optimum design variables.

　We estimated the source parameter, inhomogeneous attenuation structure and site effects of the Kyusyu region by using block inversion analysis [Tomozawa et al., 2019]. The obtained results are summarized as follows.

　1) Estimated inhomogeneous attenuation structure showed low-Q around the volcanic region. These results are consistent with previous studies which had evaluated the attenuation structure individually around the damaged earthquake.

　2) The short-period levels of large major earthquakes were estimated to be from about 1/2 to average of Dan et al.(2001). The source characteristics were estimated to be larger when the heterogeneity of Q values was taken into account.

　3) The site amplification factor was estimated to be smaller when inhomogeneous attenuation structure was taken into account for many stations. Considering that seismic waves were observed at stations where many records passed through highly attenuated zones around the volcano, the site amplification factor was estimated to be smaller than actual if a homogeneous attenuation structure was assumed for the entire region. We compared the theoretical amplification characteristics with those of bedrock sites, and found that the amplification factors were more consistent with the theoretical amplification factor when the inhomogeneous attenuation structure was considered.

　4) By considering the inhomogeneous attenuation structure, the value of variation was reduced to up to 15%, especially at high frequencies. The bias of the spatial distribution of residuals and standard deviation were also improved.

　Eigenvalue manipulation of a stiffness matrix is newly proposed. The method of the eigenvalue manipulation is defined by Equations 2 and 5 for symmetrical and asymmetrical matrix, respectively. This manipulation magnifies the unbalance forces in only directions of eigenvectors with negative eigenvalues, and does not affect in the other directions with positive eigenvalues. In a simple inverted bar model, first order equilibrium leads its results to an unstable balanced path under compression. Additionally, its stable balanced paths are based on a solution of a cubic equation. Analytical results of 3D cantilever column model with 4 elastoplastic springs are very successful.

A combinational optimization algorithm for the design of base-isolation devices (bearings and dampers) is proposed. Among multiple-type options in the devices, rational combinations with bearing and dampers are designed by applying the local search approach in the algorithm. The bearing sizes and numbers of dampers in addition to the device-types are treated as discrete design variables. For given structural conditions of superstructures, the total cost of the devices is minimized under the multiple constraints such as the maximum displacements in the isolation layers and floor accelerations in the superstructures. The responses under large earthquakes are simulated with multiple degree of freedom (MDOF) models of the superstructures in the response history analyses (RHA). The multiple Level 3 earthquake ground motions, which are recorded ground motions scaled and normalized as the maximum velocity of 75 kine, are used in RHA. The effects by higher modes are directly evaluated in RHA. The findings are as follows:

(1) A base isolation building with 21-story RC moment-frame superstructure was examined The natural rubber bearings (NRB) and sliding support with rubber-pad (SSB) are the design options for the bearing types. Also, the lead dampers (LD) and oil dampers (OD) are the options for the dampers. The effectiveness of using the proposed algorithm was confirmed by comparing the results with those by the alternative algorithm for fixed-device design. The obtained design solutions are defined as superior solutions, in which the constraints of maximum displacement in the isolation layer and bearing pressure under the sustained load are governing in the examined structure. The superior solutions are not globally optimized solutions but are rationally designed independent of the engineer’s skills.

(2) The effects in the responses induced by higher vibration modes were confirmed by a comparison of the responses with the models with rigid superstructure. While the difference between the MDOF and rigid models is less than 20 % in the responses of the maximum displacements at the isolation layers, the floor accelerations and story shear coefficients in higher stories are significantly greater (more than twice) in the MDOF models than rigid models. The proposed algorithm takes into account the higher-mode effect, which can significantly influence the floor accelerations and story shear coefficients.

(3) The influences in the superior solutions by the design constraints regarding the displacement in the isolation layer, floor accelerations and elastic base-shear coefficient were investigated. The ODs, whose cost is relatively high, are more used for desisting the floor accelerations at higher stories under the greater seismic ground motions and smaller displacement constraints at the isolation layer. In contrast, SSBs are more used rather than LDs for increasing the elastic lateral strength of the isolation layers, which is required for the design against wind loads. The usefulness of the proposed algorithm was confirmed through this study, identifying the governing design constraints for various superior solutions.

　Recently, prefabricated piles tend to be used for the pile foundation of medium and low storied buildings, because of high workability and cost efficiency in comparison with cast-in-place concrete piles. On the other hand, the ductility of prefabricated piles is lower than cast-in-place concrete piles, because prefabricated piles are produced by centrifugal molding.

　It is expected that demand for pile foundations with higher ductility will increase, because revised “AIJ ：Recommendations for Design of Building foundations” has shown the ultimate strength design method of the foundation structure. However, the evaluation method of hysteresis model for SC piles has not been established.

　In the previous paper, it was clarified that the ductility of SC and PRC piles can be improved by infilled concrete, and the evaluation method of backbone curve for those piles is supposed. On the other hand, the infilled concrete is filled in full length of piles. Considering workability, it is necessary to evaluate the performance of piles that infilled concrete is filled only near the pile head.

　Therefore, this paper presents the experimental data to improve the ductility of SC piles under the earthquake and the evaluation method of restoring force model, when the filling section of infilled concrete is 1.0D (D: pile diameter) for SC piles.

　In the second chapter, the method of setting the section length of infilled concrete and the experimental results are shown. The parameter of the experiment is the section length of infilled concrete, axial force, steel thickness and shear span ratio. The plasticization of the section with infilled concrete is earlier than the section without infilled concrete by the method shown in Fig. 5. Then, the results of the experiment are shown. When the length of infilled concrete is 1.0D (D: pile diameter) for SC piles, infilled concrete improved the ductility of pile under the cyclic loading

　In the next chapter, the evaluation method of the hysteresis model of SC piles is shown. As in the previous report, experiments are analyzed by fiber analysis model. Fig. 10 and Table 5 show the hysteresis curves of concrete and steel. The pile head spring is evaluated as shown in Fig. 11 and Fig. 12, and the origin-oriented type shown in Fig. 13 is set for the hysteresis curves.

　Then, a comparison of experimental results and analysis results is shown. In each of the specimens, initial stiffness, maximum moment, and ultimate deformation of analysis corresponds to the those of experiments. In addition, equivalent viscous damping ratio evaluated by the fiber model analysis agree well with those experiments. This analysis method can be applied up to the deformation whose strength is dropped by 10% of maximum strength.

　Many fire resistant tests were performed to adopt the conventional fire curve such as ISO fire curve. Residual properties of concrete were typically exposed to elevated temperature only for several hours with unloaded condition. To consider concrete structure sustained upper structure in compression and possibilities to be exposed to elevated temperatures for more than 1-Day, combination between the presence of loaded condition and the long hours’ heating effect needed to be evaluated.

　This experimental study focused on the presence of loaded condition and the long hours’ heating effect. Initially, residual properties of siliceous and calcareous concretes, which were exposed to elevated temperatures only for several hours, were evaluated to confirm the applicability of conventional codes.

　Secondly, long hours’ unloaded heating test was performed. Based on the test results in this study and conventional test results, the formula to evaluate the long hours’ heating effect was proposed.

　Thirdly, 2-hours loaded heating was performed. Based on the test results in this study together with conventional test results, the formula to evaluate the loaded heating effect was proposed. Finally, the combination between long hours’ and loaded heating effect was evaluated together with both the conventional and experimental test results in this study. Major findings were as follows;

1)　The formulas whose maximum temperature were 167-800 ˚C and whose maximum temperature holding time were limited to 2 hours to 30 days with unloaded condition to estimate long hours’ heating effect on compressive strength ratio(α_C1) and young’s modulus ratio (α_E1) was proposed. The effect to estimate long hours’ heating was considered to have correlations with maximum temperature(T ).

　　α_C1(T )= − 0.00048T+1.08

　　α_E1(T )={α_C1(T )}²

2)　The formula whose maximum temperature was 300-500 ˚C and whose exposure time was limited to 2 hours with loaded condition(0.1≦k≦0.2 to estimate loaded heating effect on compressive strength ratio(α_C2) and young’s modulus ratio (α_E2) was proposed. The effect to estimate loaded heating was considered to have correlations with squares of maximum temperature(T ).

　　α_C2(T )=1+5.0×10^-7T ²

　　α_E2(T )={α_C2(T )}²

3)　Loaded and long hours’ heating tests whose maximum temperature were 200-800 ˚C and whose maximum temperature holding time were limited to 2 hours to 7 days with loaded condition(0.1≦k≦0.2) were confirmed by multiplying the loaded heating and long hours’ heating effects of corresponding compressive strength ratio and young’s modulus ratio or the conventional formula whose exposure time was limited to several hours with unloaded condition.

　The present paper discusses statically equivalent seismic force distributions for single layer reticulated shells. To analyze the relationship between a shell geometry and its vibrational properties, a vibration database is constructed. The database contains the results of natural vibration analysis for about 200 structural models, each of which is generated by parametrically varying the span, slenderness ratio of the member, number of partitions of the shell, and so on. The overall shape of the structure is based on three basic types: cylindrical shell, spherical shell, and free form shell. From the analysis of the database, the two-mode SRSS method reveals the effect of the effective mass ratio and strain energy ratio on the response evaluation of shell structures. Finally, the range of applicability of the two-mode static seismic load distributions, as shown in eq.(5.1), is investigated. The conclusions of this study can be summarized as follows:

1)　The response estimation using SRSS with two modes e1with the largest strain energy ratio and e2 with the second largest strain energy ratio can evaluate the response accurately in the range where the sum of the strain energy ratios of the two modes _εe1+_εe2 is greater than 0.8. In addition to this condition, it is confirmed that the bending moment can be evaluated accurately if the sum of the bending strain energy ratios of mode e1 and mode e2 _bεe1+_bεe2 is more than 0.2.

2)　It is confirmed that the response evaluation based on two modes using mode e1with the largest strain energy ratio and mode e2 with the second largest strain energy ratio is more accurate than that using mode r1 with the largest effective mass ratio and mode r2 with the second largest effective mass ratio. This is because the vibrational modes of in-plane deformation that appear in very short natural periods may have large effective mass ratio, and such modes do not contribute significantly to the overall response.

3)　For some cylindrical and free-form shapes, it is possible to evaluate the results of CQC considering all modes with two-mode SRSS. On the other hand, it is difficult to simulate the response of spherical shapes with two modes.

4)　For structures where _εe1+_εe2 >0.8 and _bεe1+_bεe2 >0.2, it is clarified that the two-mode based seismic load distribution can evaluate seismic response with high accuracy. Furthermore, the sum of the effective mass ratios of the two modes varies from 0.1 to 0.8. For this reason, it is concluded that when calculating two-mode based seismic load distributions, it is necessary to focus on the magnitude of strain energy ratio and bending strain energy ratio rather than effective mass ratio.

　Since there is a close relationship between the form and force in the large span spatial structures, it needs to design a suitable structural form corresponding to the stress transmission. Furthermore, it is desired to construct efficiently with saving resources for reducing environmental loading. Structural engineers need to judge totally by considering various requirements (e.g. structural safety, aesthetics, constructability, and economics) for a short time.

　A structural rational form can be found easily using the optimization method. There are various studies of computational form-finding methods for large-span spatial structures. Recently, it has been applied for the realization of practical design. Generally, optimal shape tends to be a complex shape. According to the construction reports of its application, it can be confirmed that issues about constructability of complex shapes and reducing scaffolding material have been solved in the construction phase. It is significant to solve the construction problems during the early design phase by using optimization methods.

　Removing supports is one of the important construction processes for spatial structures. Generally, it is mentioned that depending on the support conditions during the removal process, the internal stress may be higher than those in the completed state in the RC large-span structures. From a point of view of safety, it is necessary to plan to remove supports carefully. Furthermore, planning for the construction process depends largely on the experience of the contractors/designers. In the case of complex shapes, it can be imagined that it involves a lot of trial and error and is extremely difficult. If the construction plan can be reasonably designed at the stage of the form-finding process, it will be possible to realize further resource-saving and efficiency of construction materials. There are a lot of studies about optimization for removing supports in the construction process of the truss or tensile structures. However, to the author's knowledge, there are a few studies for RC spatial structures.

This paper presents a simultaneous optimization method for the large span spatial structures obtaining the process of removing supports in the construction and the shape of the completed state. An optimization problem is formulated to minimize the summation of the strain energy during removing supports. The optimization algorithm consists of two methods. The coordinates of B-spline control points are optimized using Sequential Quadratic Programming (SQP). Furthermore, Local Search (LS) is used for obtaining the order of removing supports.

　It is shown in the numerical examples that not only obtaining strain energy minimized shape, react force and stress are suppressed during the process of removing through optimization. In the case of a 2D arch, the optimal order is to start from the end with removing the center at last. In the removal phase, it is effective to leave the center support during the removal to reduce bending deformation. By using this method, it can realize the construction plan for the supports considering both structural safety and constructive efficiency. Moreover, the proposed methods require less computational cost than the heuristic method shown in the numerical example. In the optimization using NP2, it is possible to obtain the solution with less computational cost than using NP1. However, the order of removal of supports becomes complicated. From a point of view of practical design, this result needs caution to avoid mistakes in construction.

　The response analysis was conducted considering the site amplification effect on Japanese conventional wooden houses in metropolitan area in this paper. Then, the general tendency of seismic response of the Japanese conventional wooden houses was investigated from the analytical results. Finally, the wooden houses damage estimation map in the metropolitan area was shown for a disaster prevention material.

　The major findings obtained in this paper were as follows.

1）　If the base shear coefficient was 0.1, the first story's maximum drift angle exceeded the drift angle of 1/20, judged as collapse at all analytical sites in the metropolitan area. If the base shear coefficient was 0.4, response drift angles showed values below the safety drift angle (1/30(rad.)) judged as severe damage in most sites.

2）　Even if the base shear coefficient was 0.6, the maximum response drift angle was larger than the safety limit drift angles (1/30(rad.)) at some sites in Site class 2 and Site class 3.

3）　In the Site class 3, it might be possibly less than acceptable drift angle even for houses whose base shear coefficient is not so high, if the site acceleration response spectrum is small within the shorter period less than about 1 second.

4）　It was not easy for wooden houses to reduce their response at the sites in Site class 2, and the sites judged as Site class 3 closer to Site class 2.

5）　The phenomenal above mentioned in items 1)～4) were possibly understood from the acceleration response spectrum considering the site amplification effect.

6）　From the damage estimation map presented in this paper, it may be understood that the maximum response drift angle was large near Tokyo Bay and Chiba's coast and that the sites where the maximum response drift angle was relatively small were the northern Kanto region, such as western Saitama, southern Ibaraki, Tochigi, and Gunma.

　In this study, the influence of shapes of hysteresis dampers using bent steel plates for base-isolated structures (BSPD) on the deformation capacity is investigated. In chapter1, conditions which improve deformation capacity is suggested with reference to the Ref.13). In chapte2, the influence of steel plate shapes on the mechanical behavior (e.g., yield strength, initial stiffness, and equivalent plastic strain) is investigated by FE analysis. In chapter3, the cyclic shear loading tests are conducted to investigate the deformation capacity of BSPDs. The experimental parameters are steel plate shapes selected in Chapter2. The conclusion of these chapters are as follows.

1.　The yield strength and initial stiffness increase as α increases, or h decreases.

2.　When α = 0, the yield strength differs between 0° and 45° loading direction. When α = 0.4 or more, the loading

direction has little influence on the yield strength.

3.　Failure modes are classified according to the load amplitude and direction. In the case of 0° loading direction, the crack propagates along the circular plates when d = 100mm, and the crack propagates from diagonal plates when d = 200mm. On the other hand, in the case of 45° loading direction, the crack propagates perpendicularly to the loading direction from the circular plates, regardless of the loading amplitude.

4.　The deformation capacity increases as α decreases, or h increases.

5.　Compared to the previous BSPD¹³⁾, the deformation capacity is largely improved by 1) eliminating of the bolt holes on end plates, and 2) using LY225.

6.　There is a correlation between the magnitude of equivalent plastic strain obtained by FE analysis and the deformation capacity of BSPD.

　　Singly-symmetric open sections in pure compression reveals two fundamental buckling modes: distortional and local, except for overall buckling such as bending buckling. Though these buckling modes reveal independently or in combination in channel members, this paper focuses on only independent local buckling. Traditionally, local buckling strength is calculated as a simply supported plate or fixed supported plate because its methodology is easy and plain. However, when such a simple buckling strength calculation method is applied to a member whose purpose is to thin the member and make advanced use, such as light gauge steel member, the local buckling strength must be calculated precisely. There is a concern that the advantages of these members will be lost. Therefore, in this paper, the buckling displacements of the web/flange/lip are expressed by a series of displacement functions. The elastic local buckling evaluation formula, which considers the element interactions, is proposed via the energy method. Also, the relationship between the cross-sectional shape of the channel member and the buckling behavior is clarified. Finally, comparisons between the proposed and conventional formulae are described.

　　From this research, the following are found.

1)　To consider the restraint effect due to the element interaction of each plate that constitutes the member, the buckling waveform of the web/flange/lip is expressed by a series of displacement functions. Using this displacement function, the elastic local buckling strength formula for the channel member is proposed by the energy method.

2)　The lip width ratio's effect on the elastic local buckling strength differs depending on the difference in the flange width ratio b_f/b_w. When the flange width ratio becomes large, local buckling is relatively likely to occur in the flange; however, as the lip width ratio c/b_w increases, the flange's local buckling is restrained so that the lip restraining effect appears in a range where the flange width ratio is large. However, the effect of the lip width ratio on the local buckling resistance is not large.

3)　Since the plate thickness in the cross-section is uniform regardless of the presence or absence of lips when the flange width ratio decreases, the flange's width-thickness ratio decreases relatively, and local buckling of the flange is suppressed. As a result, the elastic local buckling strength increases.

4)　The flange coefficient is almost 4.0 or less in the channel member's realistic shape even when local buckling occurs in the channel member. The current design manual for light gauge steel members stipulates the required lip width for the flange buckling coefficient to satisfy 4.0 to suppress distortion buckling and cause local buckling. However, since the range in which the flange buckling coefficient satisfies 4.0 within the channel member's manufacturing range is limited, it is not appropriate as a basis for setting the required lip width.

5)　The average value of the finite element elastic eigenvalue analysis results against the elastic local buckling strength obtained from the proposed formula (19) is 0.967, the standard deviation is 0.00744, and the maximum error is 4.49 %. The approximate formula (20) can also obtain reasonably calculated results, which is a simple notation of the proposed formula (19).

　The purpose of this experimental study was to understand and improve the heat resistance of the glued-in rod (GIR) method of joining wooden members. Experiments were carried out with the conventional GIR method (Con), in which the inner surface of the reinforcing bar insertion hole is smooth, and an improved GIR method (Imp), in which the inner surface of the insertion hole is grooved with the expectation of improved mechanical bond strength.

　The experimental procedure can be broadly divided into three steps, as follows.

　Step 1: Assuming a building containing wooden members jointed using the GIR method, the internal temperature of a GIR joint in the event of a building fire was calculated by unsteady heat transfer analysis. This temperature, 80℃, was then set as the target temperature for pull-out experiments.

　Step 2: The heating required to achieve the target 80℃ boundary temperature between adhesive and wood in the GIR joint specimen was determined and experimentally verified by analytically examining the temperature-time relationship. It was determined that the temperature in the electric furnace used to heat the specimen should be raised to 180°C over a period of 35 minutes.

　Step 3: Based on the determinations made in steps 1 and 2, the following pull-out experiments were carried out:

　•　Experiment I: Pull-out from unheated specimens at room temperature.

　•　Experiment II: Pull-out under constant load from specimens during heating.

　•　Experiment III: Pull-out from specimens when the boundary temperature reaches the target of 80℃. The loaded or unloaded state during heating was a parameter in these pull-out experiments.

　•　Experiment IV: Pull-out from specimens after initially raising the boundary temperature to 80℃, holding it there for about 2 to 3 hours, and cooling again.

　The results obtained from these various pull-out experiments were as follows.

　•　Experiment I: The pull-out strength of Imp was lower (about 90%) than that of Con. The cause of this has not been clarified, and further experiments and data accumulation will be required in the future.

　•　Experiment II: The boundary temperature at the time of pull-out failure was 96°C for Imp and 61℃ for Con. The loading in this experiment was fixed at 37% of the room temperature pull-out strength, so further experiments and data accumulation will be required for other loadings.

　•　Experiment III: The average pull-out strength of Imp was 97.7 kN for unloaded heating (49 kN for Con) and 74.5 kN for loaded heating. Thus the pull-out strength of Imp under unloaded heating was about twice as high as that of Con. The pull-out strength of Imp under loaded heating was about 1.5 times as high as the pull-out strength of Con under unloaded heating. The high-temperature pull-out strength of Imp is clearly an improvement over that of Con.

　•　Experiment IV: The pull-out strength after heating and cooling greatly exceeded the high-temperature pull-out strength in both Imp and Con, and was almost equal to the unheated room-temperature pull-out strength. One of the reasons for the exhibited strength recovery may be that the moisture content near the boundary between adhesive and wood after heating and cooling was almost the same as that of the unheated wood.

　Cold-formed steel square hollow section columns are widely used for steel-structured buildings in Japan. Regarding the fire resistant design, the buckling and local buckling temperatures are evaluated by "AIJ Recommendation for Fire Resistant Design of Steel Structures (AIJ,2017)", respectively. For the evaluation of fire redundancy for an overall steel frame, the post-buckling residual strength of heated column is considered as a contribution factor to the load-redistribution capacity. However, the interactional effect between the flexural and local buckling for the post-buckling residual strength was not considered. In particular, the large strength degradation might be caused by the local buckling occurred in the heated column, resulting in the overall collapse of the steel frame owing to decrease in the fire redundancy. There is no experimental research to discuss the interactional behavior between the flexural and local buckling at the elevated temperature for the cold-formed steel square hollow section column. In this study, the axially loaded compression column experiments under both steady and transient state conditions were conducted to clarify the interactional buckling performance of square hollow section column at the elevated temperatures. To evaluate the maximum axial strength and the residual strength after that, the column specimens with both width-thickness ratios B/t of 21 and 28 were used for the steady state experiment under each constant temperature(400, 500, 600, 700, and 800 ℃). The former specimen (B/t=21) was used for the transient state experiment under the constant axial load (the load ratio values were given by 0.28, 0.33 and 0.40, respectively), to evaluate the collapse temperatures. The value of slenderness ratio of all the column specimens was unified by 16.4. From the experimental results and the discussion based on those, the following knowledge was obtained.

1) The single local buckling collapse mode and the interactional collapse mode of both flexural and local buckling were observed, respectively. The former collapse mode was dominant for the steady state experiments under 500℃, however, the latter collapse mode was dominant under the high temperature region (600-800℃), excluding some exceptions. The strength degradation of axial load was caused by the occurrence of local buckling, even if the interactional buckling behaviors were observed. For the transit state experiments, the collapse mode was changed from the single local buckling to the interactional buckling collapse mode with increasing the collapse temperature.

2) For the column specimen that exhibited the interactional buckling behaviors, the residual strength after the peak axial load can be approximately quantified by using the existing evaluation formula, which has been constructed based on the local buckling behavior of stub-column test specimens. When evaluating the contribution effect of the heated column to the load-redistribution capacity for the practical many steel buildings in Japan, the residual strength is strongly affected by the local buckling behavior, because the general steel column possesses the small value of slenderness ratio and the local buckling performance relatively becomes dominant, such as the column specimen used in this experiment.