MENSUSMONITOR – Algorithm implementation for detecting live load events and assessment of structural effects on bridges

The handling and treatment of data from automatic monitoring systems installed in civil infrastructures is by itself a hard and time consuming task in the search of structural behavior knowledge. This paper presents the implementation of an algorithm for detecting live load events and assessment of structural effects on monitored bridges. The events detection algorithm offer a set of results in form of histograms events that characterize several variables distributions like the number of events detected, the amplitudes of observed parameters, or even the level of load applied to the structure. The proposed algorithm is implemented in the MENSUSMONITOR software specifically devoted to the structural health monitoring. Grounded in the know-how gathered by the research unit FEUP-LABEST in the recent years, MENSUSMONITOR proposes to give faster and efficient answers to the processing and interpretation of records collected by the structural monitoring systems deployed on bridges. of experimental results obtained in materials and structures of Civil Engineering. This reflection results from the experience acquired in the last years by FEUP-LABEST, a R&D; unit of the Faculty of Engineering University of Porto. In this R&D; unit a group of researchers have been developing methodologies and tools aiming the implementation of the structural monitoring as a practical tool to support the structures safety evaluation, thus assuring a valid path towards a better comprehension of civil infrastructures real behavior. The problems discussion and their possible resolutions were the basis to create a common platform MENSUSMONITOR. The database consulting with the intention to extract a superior knowledge is a hard task, with significant time consuming, if no specific numerical tools are used. The potential knowledge given by the monitoring systems is high, but it is also important to reach that knowledge in real time for an effective management and surveillance of civil infrastructures (Sousa et al., 2008a). 3 MENSUSMONITOR – THE APPLICATION TOOL MENSUSMONITOR is a specific application guided to the SHM cause, whose functionalities are intended to give useful help in the consulting, treatment and knowledge extraction from the collected data. Its mainframe is implemented in LabVIEW code (Beyon, 2001), with the attachment of a set of modules that can be developed in the same language code of the mainframe or in C++ or MATLAB (Sousa et al., 2008a, b). Regarding its conception, MENSUSMONITOR is an application that have a main program that manage: (i) the input of data; (ii) the functioning of all numerical tools that transform that input data in other relevant information, and; (iii) the output data in form of data files that can be text files, graph images or summaries html pages. All these modules are dependent between them in a sequential manner but independent in what concerns to the source code (Sousa et al., 2008b). At the present, MENSUSMONITOR offer a set of numerical tools which are highlighted: (i) “mensus_samplingrate_change”: readings sample reduction or increasing of an experimental result; (ii) “mensus_filter”: data filter by using a filter library; (iii) “mensus_fitting”: curve fitting of the theoretical expression of Eurocode 2 (Bebby, 2005) to experimental results; (iv) “mensus_correlation”: identification of correlations between two experimental results and the corresponding correlation equation (Sousa et al., 2008b). MENSUSMONITOR has a user guide to help the user in its use, namely in the correct use of the numerical tools. This was one of the noble aspects in the development of the MENSUSMONITOR, because compromised the numerical tools developers to think in the user perspective beyond the programmer perspective. Figure 1 illustrates the functional structure of MENSUSMONITOR. To extract useful information from databases created by monitoring systems that acquire with high sampling rates, a new numerical tool was implemented and attached to the MENSUSMONITOR. The collected data with those high frequencies allow to characterize the dynamic behavior of the monitored structure, and identifies also the actions applied to the structure, namely live loads. An appropriate selection of the sensors positioning, as well as the sensors type, gives the possibility to have a weight balance that identifies traffic events which occur in the monitored structure. The adopted designation for the numerical tool was “mensus_traffic” considering the relevance that this tool has for the traffic characterization on the monitored structure. Figure 1. Functioning structure of MENSUSMONITOR. 4 “MENSUS_TRAFFIC” – EVENTS DETECTION ALGORITHM 4.1 Problem formulation As it was referred, some of the actual monitoring systems used in civil infrastructures allow high sampling rates. Based in those databases, typically fed with readings acquired above a 5 Hz frequency, it is possible to evaluate some indicators related with the traffic that passes over the monitored structure, namely: (i) daily traffic volume; (ii) velocities and moving directions; (iii) load levels due to the traffic events detected. 4.2 Methodology and numerical implementation The adopted methodology was based in the following five main steps: (i) pre-treatment of the experimental results; (ii) detection of the extreme values; (iii) calculation of the respective velocities and moving directions; (iv) determination of the load levels; (v) plot of histogram results. These five main steps embody the “mensus_traffic” tool, which was developed in LabVIEW and C++ and afterwards included in the mainframe of the MENSUSMONITOR. 4.2.1 Pre-treatment of data Considering the database to study, the first step to be performed is to split that database in smaller sets. This procedure is controlled by the user by defining how large those smaller databases are. For each one, a procedure is made to extract only the information related with the traffic events, by subtracting a tendency line and data filtering. 4.2.2 Extreme values detection The pre-treated data is now used to identify the local extreme values. This identification is made by considering a sensibility value for which the search method is initialized (Fig. 2). Figure 2. Local extreme values identification. Therefore, the local extreme values identification of a time series Y(t,y) of ‘n’ samples is performed by the following algorithm: Matrix ‘Y’; //time series (ti,yi) with dimension nx2 Matrix ‘Y_sub’; //time series part (ti,yi) with dimension m(Yo then add (ti,yi) to ‘Y_sub’ matrix; } //-------------------------//cicle over ‘Y_sub’ Matrix For i=1 unti m(’max’ { ‘max’=yi; } else { Add ‘max’ to ‘Y_ext’ Matrix in index position j; j=j+1; } } At the end of this calculation step, a matrix Y(ti,yi extreme) is created with j extreme values identified. It is relevant to refer the importance of the sensibility value, Yo, which makes possible to identify the extreme values and allows the user to obtain different results by setting the desired value for this parameter. 4.2.3 Velocity calculation and moving direction identification With the extreme values identified, the achievement of velocities and moving directions is the next calculation step. For this propose, the user chooses two time series from the time series database related with the traffic events that are intended to be studied, as well as the physical distance between the two sensors from which the data were collected (Fig. 3). Figure 3. Time interval identification between the local extreme values of two sensors readings. The calculation procedure pairs the time instants where the extreme values occur. To control the procedure robustness, the user can pre-define: (i) the minimum time interval allowed between two consecutive traffic events for each register; (ii) the maximum time interval allowed for the two extreme values identified by the two sensors for a traffic event. These two parameters can be easily changed as a function of the expected traffic. The velocity calculation is performed by Eq. 1, where the signal result indicates the moving direction.