數十年來,結構物健康檢測一直是個受到矚目的研究領域,各種識別損傷的方法相應而生。本論文首先將角鋼構架簡化成4個自由度的系統,於數值模擬中利用ARX模型對結構物進行系統識別工作。在獲得與精確解所差無幾的自然頻率、模態振形識別結果,證明了系統識別方法的正確性後,我們還觀察到有無考慮輸入歷時及外力頻率數目會影響識別所需的最低模型階數。 然而,直接使用實際試驗的歷時記錄進行識別無法得到良好的識別結果,必須對歷時記錄進行濾除高頻段和取其平緩區段兩道前處理,並逐步提高模型階數才能得到前三個模態的良好識別結果。 利用系統識別方法獲得結構的模態振形後,我們以此為基礎,選定層間偏移模態與層間模態相對轉角作為檢測指標。數值模擬結果顯示前者雖在檢測微小損傷方面表現不佳,但若指標變化率為正,可確定該樓層發生損傷狀況。而變化率最大且為正值者,必為損傷程度最大之處,整體表現較層間模態相對轉角為優。選定層間偏移模態作為檢測指標後,實際試驗所得的結果與數值模擬相符,可由損傷前後的層間偏移模態變化率判斷該樓層是否發生損傷狀況。
For decades, structure health monitoring has been of great interest and many reachers have devoted to develop many methods to detect damage. In this paper, we simplify the steel frame as a 4-DOF system first of all, then used ARX model to identify the nature frequencies and mode shapes of the structure in numerical simulation. After obtaining the identification results very close to the exact solutions, we prove that system identification method is credible. Also, we find that the same system parameters are identified by the ARX model with or without inputs (external forces), as long as the correct model order is used. However, when the ARX model is applied to the actual time-history data measured from a structure testing, good identification results can not be obtained. In order to obtain good results, besides increasing the model order, filtering and selecting the adquate range from data are also necessary. Based on the mode shapes from system identification results, inter-story drift mode shape (IDMS) and inter-story mode shape rotation angle (IDMSRA) are used to detecte damage locations for structure health monitoring. Both the results of the experiment and numerical simulation show that IDMS ans IDMSRA can correctly identify the damaged floors when the damage extent of the floors are similar. However, for other damage combinations, IDMS and IDMSRA may fail to detect all damage locations.