低成本且非侵入式電梯運行狀態偵測系統

Abstract

現今電梯已成為一種非常普及且重要的垂直運輸工具，幾乎是每一棟住宅、飯店、百貨公司等建築的標準配備，因此為了以利於電梯的保養、維護、安全甚至是行程管理，監控電梯運行狀態的裝置便成為了一個重要的議題。但目前大部的電梯監測系統都是藉由安裝於電梯上的控制板所取得資料。雖然從電梯本身內建的感測器可拿到大量資料，且經過處理後更能提供豐富的資訊，但這只限於能順利跟電梯所屬廠商順利交涉的情況，因此許多研究若沒能得到廠商支持則會受到諸多限制。 本研究提供了一種非侵入式且低成本的方法，建設了一套設備以利於收集電梯運行的相關資料，並提供資料上傳至雲端分析。也透過遞迴選取特徵的方式，找出了適合代表電梯行為的特徵，為電梯狀態建立分類模型，並利用滑動視窗來進行即時性的電梯狀態辨識，最後更進一步根據電梯行為的特性，提出了利用有限狀態機對分類器進行校正的方法來提高分類器的辨識準確率。

Nowadays, elevator is become a common equipment and play an important role in residential, hotels, shopping malls and other buildings as a vertical transport. Thus, it is necessary to monitor the running status and failure conditions of elevator for the elevator management, maintenance and safe operation. However, most of elevators monitor system are using onboard controllers installed within the elevators. Although it can generate huge amount of data during day-to-day and providing informative after processing in the backend, everyone who want to knows the status of the elevator needs to contact with supervisor of the elevator. In this study, we propose a non-intrusive and low-cost method to build a device in order to collect data about the elevator’s behavior and transmit the raw data to analyzing on the server. Also, we find out the most suitable features set for the target elevator’s operations and build a classifier for it. Moreover, sliding window is being adapted and the most suitable window size was found so as to detecting the elevator’s operations online. Finally, an adjustment method for the classifier be proposed by using the finite state machine of the elevator.