RC railway viaducts supporting entire railway systems are deteriorating due to age, increased loads and frequency of use, and frequent mild to strong earthquakes. These structures link many cities and prefectures in Japan, playing a vital role in the socio-economic life of the country, so that maintenance and monitoring of these structures are of crucial importance.

Without prior knowledge of the existence of damage, structural health monitoring (SHM) methods are used to detect, locate and assess damage. Many SHM methods use vibration-based procedures and algorithms for damage detection. SHM is a wide research field that includes development of new measurement systems and effective excitation methods, modal analysis and system identification techniques, determination of damage-sensitive features, damage quantification and finally, structural evaluation.

Past earthquakes showed that RC railway viaducts are susceptible to different forms of damage. Damage to structural members above ground are easily found by visual inspection but damage to substructures below ground will require excavation to be seen. This requirement is difficult to meet as it will not only be expensive but will require an extended amount of time, keeping the railway system out of service. Further, damage in foundation elements after an earthquake event may even go undiscovered, become worse over time and slowly reduce the carrying capacity of the structure. With these considerations, a vibration-based method that can detect and quantify the severity of unseen damage by measuring the vibration of structures above ground alone will be very useful.

In this dissertation, two sensor systems with different characteristics are used to measure the vibration of RC viaducts in service. A system using Laser Doppler Vibrometers (LDV) for monitoring Shinkansen RC railway viaducts is first presented. The LDV with its high accuracy, frequency resolution, and scanning capability, is an ideal non-contact tool for monitoring large structures. This system is shown to be superior to other sensor systems in terms of safety for monitoring personnel on site. Wireless and ordinary cable-connected accelerometers and velocimeters are also used to measure vibration of the viaduct in one direction and three-perpendicular directions.

The RC viaducts are subjected to diverse excitation sources that cause it to vibrate in distinct ways. A particular form of excitation may be more effective in detecting damage. Thus, structural vibrations under three excitation sources are investigated. Ambient vibration due to wind or other naturally occurring vibration source is measured using the LDV system and velocimeters. Ambient vibration measurement simplifies the measurement process because controlled input excitation or input force measurement is not involved. However, ambient vibration is small compared to surrounding noise because the viaducts are very stiff. Compensation for tripod vibration is achieved by removing from the recorded data, vibrations measured by a velocimeter attached to the LDV. Train-induced vibrations in three-perpendicular directions are measured as well using accelerometers to gain more insight in the vibration behavior of a viaduct caused by a passing train. Free vibration responses of columns after an impact excitation using a wooden hammer are also measured because the ambient vibration level of the columns is found to be very small to excite high frequency modes. Structural state of the RC viaducts are investigated by analyzing structural vibrations under these excitation sources revealing dynamic behavior that is unexpected and providing valuable insight on how to best pursue development of a damage detection method suited for RC railway viaducts.

Two system identification techniques are used in the analysis: Peak-picking method and Eigensystem Realization Algorithm (ERA). From ambient vibration, the Natural Excitation Technique (NExT) was used to derive cross-correlation functions which were used as input for ERA. Thus, with these techniques, the global mode natural frequencies and mode shapes of a viaduct are derived from ambient vibration data measured using the LDV system. It was observed that the global mode shapes derived from modal analysis are not pure torsional and lateral modes but a combination of the two. To begin with, the RC viaducts have unsymmetric cross-sections so that the center of mass does not coincide with the center of the viaduct. This already complicates the basic mode shapes of the viaduct. Moreover, the complicated mode shapes derived from ambient vibration reveal the presence of dynamic interaction between adjoining viaducts because of the continuity of the rail tracks connecting them. This discovery implies that it will be difficult to use changes in global mode shapes of the viaduct as an indicator of damage because the effect of damage will be masked by the existing dynamic interaction.

The same system identification techniques were used in deriving local column mode shapes from free vibration response after applying an impact load using a wooden hammer. The local column mode shapes derived are stable. Also, the local column mode shapes of a regular RC column and a steel jacket retrofitted column were compared. The results show that the local column mode shapes are sensitive to changes in the stiffness distribution of the column. Moreover, the first (single curvature) local column modes of three unretrofitted columns from two viaducts are shown to be the same, allowing the adoption of a standard undamaged mode shape for monitoring. Thus, the local column mode shape is a good index for damage detection and localization.

Analysis of train-induced vibrations show that the frequency peaks are mainly dependent on the train speed and that resonance in the global modes of vibration of the viaduct is reached at the current maximum speed of Shinkansens. The root-mean-square (RMS) of train-induced acceleration in the vertical, lateral and longitudinal directions were compared for two adjacent RC viaducts. Comparison shows that the RMS accelerations in the longitudinal direction for the two viaducts are nearly the same. If damage occurs, this parameter may change so that it may be a useful indicator of damage. Similarly, the RMS accelerations between two opposite columns were compared. The comparison showed that the lateral RMS acceleration is almost equal for the opposite columns and may serve as an indicator of damage.

A new method for detecting damage on substructures below ground by examining changes in the modal properties of the exposed part of the viaduct alone is developed. From the different damage-sensitive features identified, the method uses the curvature of the first local column mode shape. There are three reasons: the first local column mode is the easiest to excite among the local column modes, it is not affected by dynamic interaction between viaducts, and a standard undamaged mode shape can be used for monitoring. This requires many measurement points to derive an accurate mode shape curve fit. In order to model damage, a three-dimensional elastic solid finite element (FE) model of an RC viaduct was made, calibrated with modal properties identified from field measurements. In the FE model, impact loading is simulated and the resulting free vibration response is measured. To simulate actual analysis steps subsequent to field measurements, velocity responses of the FE model were recorded at twelve measurement points in two opposite columns and then analyzed using the peak-picking method and ERA. Different damage cases were considered by reducing the modulus of elasticity of concrete at different locations on the viaduct model. Once the modal properties are estimated, the Modal Strain Energy Change Ratio (MSECR) quantifies the severity of the existing damage. The results show that minor to medium-scale damage occurring within the active pile length can be detected. However, after adding random noise to the simulation results and analyzing using peak-picking and ERA, the method can no longer distinguish between minor damage and effect of noise. Medium-scale damage equivalent to 33.3% reduction in modulus of elasticity at the bottom layer of a column is the least amount of damage that can be distinguished from noise effects. Damage equivalent to 66.7% reduction in modulus of elasticity at the top layer of piles can also be distinguished from noise effects. These localized damages do not substantially change the natural frequencies of the viaduct. Thus, this new method, coupled with the LDV measurement system shows promise in detecting medium-scale localized damage in substructures below ground without the need for numerous sensors.

Finally, an experiment using a two-storey steel model was conducted to demonstrate the principle used in the new damage detection method. The experiment also demonstrated the applicability of using free vibration response at many points, measured by scanning with the LDV measurement system, in deriving column modal natural frequencies and mode shapes. The results show that the modal properties of a column changes by loosening its supporting bolt. This verifies the assertion that damage at foundations, acting like changes in boundary condition of the column, can be detected by measuring changes in its modal properties.

新幹線をはじめとする鉄道高架橋は年数を経ているものが多く,列車走行の安全性から健全度判定,特に地震後の損傷判定のニーズは極めて高い.特に,地震後の損傷において最も検知が難しいのは,高架橋下部の杭基礎やフーチングの損傷である.

本研究は,高架橋上部工の動特性から下部工の損傷検知を試みたものである.

論文は七章から構成されている.第一章では,RC鉄道高架橋のモニタリングの現状と構造ヘルスモニタリングに関する既往の研究を述べ,さらに本研究の目的を記述している.

第二章では,実際の東海道新幹線RC鉄道高架橋において,加速度計ならびにレーザードップラー速度計により常時微動ならびに自由振動を実施し,そこでの測定データからのシステム同定を行っている.その結果,高架橋の全体振動モードとして橋軸方向,直角方向,ねじれ方向の振動モードの同定に成功し,3つのモードの固有振動数が近接していること,軌道レールを介して隣接するRC高架橋との連成が顕著であることを指摘している. このことは,従来から鉄道で使われている重錘を使った振動測定はどのモードを測定してうるか明確でなく,その精度に問題があることを示唆するものである.

第三章では,新幹線通通過時の列車通過時の振動の測定結果を論じている.梁部,柱部における測定応答から,新幹線の速度により応答が系統的に変化すること.応答には,非常に高次の振動も含まれること,ほぼ同一とみなされる柱,梁においても応答が異なっており,個々のRC高架橋の構造特性はかなりばらつきがあることなどを測定結果の上から明らかにしている.

第四章では,基礎部までを取り込んだRC高架橋のFEMモデル構築を行っている.特に地盤バネの選定には配慮し,また隣接する高架橋との相互作用についても検討をくわえ,実測された高架橋の同促成と固有振動特性が良く一致することを明らかにしている.

第五章では,四章で構築されたFEM高架橋モデルを使って基礎部の損傷が高架橋の全体振動モードならびに局部振動モードに及ぶ影響を定量的に明らかにしている.その結果,全体モードに及ぼす影響は小さく,局部振動モード,たとえば柱部に及ぼす影響を数値的に明らかにしている.柱部の局部振動モードにおいてもモード形状に及ぼす影響は小さいものの振動モードの曲げひずみエネルギーの点に直目すると,基礎の損傷により曲げひずみエネルギーが有意に変化することを見出した.実際のRC高架橋にこの手法を適用する場合,ERA法もしくは,ビーク相関法が用いられるがそれによる同定誤差,計測点の個数等が局部振動モードのひずみエネルギー評価に及ぼす影響を調べ,本手法が実用的にも使えることを明らかにしている.

第六章では,局部振動ひずみエネルギーに直目した損傷同定を屋内実験より確認している.実験に用いたものは,二層のラーメン構造でうち一部の柱の幹部のボルトによる締め付けを変えた実験を行っている.振動モードはレーザードプラ振動計により行い,その結果を整理し柱の局部振動において損傷を与えてもモード形はほとんど変化しないが,モードのひずみエネルギーについては優位な変化があることを明らかにしている.

第七章は,論文の結論と今後の展望について述べている.

以上のように本論文では,鉄道RC高架橋を対象とし,外からは見えない下部構造の損傷が及ぼす上部坑の振動特性変化に着目し,柱部の局部振動のひずみエネルギーが変化することに明確にし,下部杭の損傷同定の検知に成功した.現在行われている振動モニタリングを使った損傷同定に比べ圧倒的に精度が高く実用的価値も高いと判断される.今後つめるべき課題も数多く残しているが,工学上多大な知見を提示していると判断される.よって,博士(工学)の学位請求論文として合格と認める.