An Autonomous Robotic Cladding Inspector for High-rise Buildings in Hong Kong

Abstract

Debonding of wall tiles to structure is one of the key concerns for building maintenance both in terms of serviceability of the buildings and the safety to the general public. The problem of debonding is not only serious in Hong Kong but in almost all countries around the world. Current practice in Hong Kong of identifying the possible debonding problem can be by visual inspection followed by manual hammering and audible interpretation or by the infrared scanning. However, manual analysis is very subjective in nature and infrared scanning is very limited in the applications. It is believed that the hammering approach should be the ultimate answer to the problem but it needs to be improved from a subjective system to an objective system. An accessory incorporating a hammering action with fixed force and material of impact has been designed, which can be attached onto any standard climbing robot. A directional, high-gain microphone is used to record the sound produced by each impact for further analysis. The full spectrum is generated by Fast Fourier Transform using standard hardware circuitry to achieve real time response. A new algorithm based on auto-correlation on the generation spectra has been devised to characterise each full spectrum, resulting in six important parameters. These parameters are used to train up an artificial neural network to form an expert database for a certain type of concretelrenderftile combination. The neural network becomes the critical tool to make decisions on similar combinations for cladding inspections at other sites. A histogram of goodness factor for a comprehensive inspection process is also constructed for immediate screening purpose.