Assessment of concrete pavement support conditions using distributed optical vibration sensing fiber and a neural network Hongduo Zhao a,b , Difei Wu a,b,⇑ , Mengyuan Zeng a,b , Yu Tian a,b , Jianming Ling a,b a The Key Laboratory of Road and Traffic Engineering, Ministry of Education, Tongji University, Shanghai 201804, China b The Key Laboratory of Infrastructure Durability and Operation Safety in Airfield of CAAC, Tongji University, Shanghai 201804, China highlights  Distributed optical vibration sensing technology was used for vibration monitoring.  Wavelet packet transform was used to process the captured vibration signals.  A probabilistic neural network was developed to assess the support condition.  Results of N-fold cross-validations demonstrated the effectiveness of this method. article info Article history: Received 6 August 2018 Received in revised form 22 April 2019 Accepted 24 April 2019 Keywords: Concrete pavement Support condition Distributed optical vibration sensing Vibration testing Wavelet packet transform Probabilistic neural network abstract This paper presents a new method that uses distributed optical vibration sensing (DOVS) technology and a neural network, which can access the concrete pavement support condition with lower cost and higher efficiency. This method relies on the vibration properties of the concrete pavement slabs, which was captured by an embedded DOVS system and extracted using the wavelet packet transform. Based on the extracted properties, a grading method that using probabilistic neural network (PNN) to assess the support condition was proposed. Through field tests, the accuracy of this method can reach 92.8% for 2-grade assessment and 81.4% for 3-grade assessment, respectively. Ó 2019 Elsevier Ltd. All rights reserved. 1. Introduction The support condition is an important property that affects the structural performance of concrete pavements. Severe support loss in a concrete pavement slab increases structural stress and eventu- ally results in multiple pavement distresses such as cracking and joint faulting [1]. Therefore, assessment of support conditions is crucial to pavement management and maintenance. Researchers have developed many technologies and methods for support condition detection and assessment, including deflection-based methods [1,2] and other nondestructive methods such as ground penetrating radar [3], ultra-sonic arrays [3], impact echo and infrared technology [4]. The deflection-based methods are the most common-used methods in recent decades. The basic principle behind these is that drop-weight-generated deflections would significantly increase when support loss is present. There- fore, the corresponding deflection values can be used to identify whether local support loss exists or not. Although deflection- based methods have been widely used in practice, they always have high operational costs because they require a high-accuracy force transducer and deflectometer to obtain reliable results. Another method that can be used to measure support condi- tions is vibration testing. Vibration tests are based on the idea that support loss or poor support conditions will change the restraint condition of the pavement slab and further result in changes in structural modal properties, which are measurable. Vibration tests usually use accelerometers to obtain the vibration signal which is induced by a given or ambient excitation. However, similar to the deflectometers in deflection-based methods, this accelerometer-based methods is expensive and inconvenient to use as they need to be installed and removed many times. The pavement condition can also be assessed using smart sensing tech- nology [5]. Distributed optical sensing technology, which is an emerging sensing technology for large-scale monitoring, has https://doi.org/10.1016/j.conbuildmat.2019.04.195 0950-0618/Ó 2019 Elsevier Ltd. All rights reserved. ⇑ Corresponding author. E-mail address: 1994wudifei@tongji.edu.cn (D. Wu). Construction and Building Materials 216 (2019) 214–226 Contents lists available at ScienceDirect Construction and Building Materials journal homepage: www.elsevier.com/locate/conbuildmat