Comparison of multiple monitoring techniques for the testing of a scale model timber Warren truss

The benefits of SHM within civil engineering include extending lifespan, contributing to infrastructure management, and increasing occupant safety (Chang et al. 2003). Long-term sensing capabilities, temperature and environmental interference, data acquisition, sensor placement, among other numerous factors must be considered during the design of a SHM program (Hannan et al. 2018). A robust SHM program is capable of performing three forms of monitoring: knowledge monitoring to assess the structure and determine structural behaviour, control monioring to include periodic testing, and emergency monitoring to provide an early warning system. An efficient SHM program for Hangar 1 must satisfy the three monitoring forms. As such, the program must be capable of providing data remotely, continuosuly, and in the long term. Furthermore, in terms of emergency monitoring, the program must incorporate an alarm system capable of communicating the level of risk to the occupants of Hangar 1 in real time. Lastly, the program must be nonintrusive and minimize disruptions in the day-to-day operations that occur within Hangar 1. With the minimum SHM criteria now finalized, the measurands required to meet these robust conditions were determined. These measurands were determined to be deflections, structural behaviour (stress/strain, global and localized), in-situ moisture properties, and internal as well as external environmental effects/conditions.

SHM programs have been designed and implemented on DND Warren trusses (Locklin et al. 2017). Conventional electric strain gauges, discrete monitoring using fibre optic sensors, vibration and wave-propagation based sensors, and temperature sensors were utilized. However, high data volumes and environmental conditions limited the effectiveness of this SHM endeavour. Moreover, the program did not include deflection instruments or humidity sensors, the latter being essential for timber monitoring. Locklin et al. (2017) also noted the importance of creating a more refined system of sensors thus underlining the importance of determining an optimized SHM program prior to in-situ implementation. Accordingly, the implementation of a more robust and redundant system, combined with a monitoring regime to capture internal and external environmental factors, will lead to the determination of an optimized SHM method. Lastly, previous monitoring endeavours were not capable of collecting data remotely, in real time, or communicating with the user. As such, this monitoring plan must incorporate these capabilities to be efficient for emergency monitoring.

The type of instruments utilized in a SHM project is dependent on the nature of the project including the type of structure, material, and loading arrangements, among others. In addition, instrument selection is determined by the requirements of the actual SHM program (i.e., research questions that need to be answered). Several forms of instrumentation were considered during the laboratory component of the overall research program. The instrumentation as part of the monitoring program can be divided into the monitoring of deflections, strains/stresses, and environmental conditions.

Deflection or displacement monitoring is a significant portion of SHM, particularly for serviceability requirements. Linear variable differential transducers (LVDT) have been used extensively for timber monitoring including the in-situ, full-scale load testing of historic timber systems (Foo and Seckin 1988; Branco et al. 2017). Selected limitations include temperature sensitivity, limited range capability, and the requirement of a stable reference system (Moreu et al. 2015), which is a particular issue for the monitoring of long-span structures. The automatic total station (ATS) has been successfully employed in numerous static SHM applications (Merkle and Myers 2004; Psimoulis and Stiros 2013). The autonomous nature of the ATS makes it an ideal instrument for remote, continuous monitoring. The ATS is limited by line of sight and is restricted to determining the behaviour at discrete locations. Conversely, light detection and ranging (LiDAR) can accurately provide distributed global deflection data within a structure (Bertolini et al. 2015). The requirement of extensive postprocessing and the inherent nonautonomous nature of LiDAR limit its effectiveness for emergency monitoring. Other challenges include line-of-sight requirements and large data files.

Electric strain gauges (ESG) are the most traditional technique and a proven method for measuring strain at discrete locations, for both laboratory and field SHM projects, and have been used for the testing and monitoring of timber truss structures (Lanata 2015; Locklin et al. 2017; Helmer-Smith et al. 2021). The reliability of these conventional gauges for general timber monitoring is demonstrated by the field applications described by Palma and Steiger (2020) and Cavalli and Togni (2014). Various types of strain gauges have also been extensively used in steel truss field monitoring programs (Bhalla et al. 2009; Park et al. 2013; Moreu et al. 2017). Highlighted limitations include long-term sensor creep and drift. As such, it is necessary to consider a compensation or reference gauge throughout the long-term program. The correct installation of a strain gauge is nontrivial and requires attention to detail in terms of alignment, type of gauge, epoxy, connectivity, and placement. One inherent limitation of the ESG is that they are restricted to local data points, thus, it is important to optimize the placement of sensors. In addition, this limitation may lead to strain measurements that are misleading or nonrepresentative of the true timber properties due to the high variability and nonhomogeneity of timber (Palma and Steiger 2020). Conversely, fiber optic sensors (FOS) are an emerging form of instrumentation due to their capabilities of providing local, quasi-distributed, or fully distributed strain data. Quasi-distributed fibre bragg grating (FBG) sensors were used by Locklin et al. (2017) and proved more stable than the conventional strain gauges. However, large data files obtained from the FBG sensors were highlighted as a major limitation. Similarly, FOS have been respectively utilized for the quasi-distributed and distributed strain monitoring of historical buildings (Whelan et al. 2002; Bastianini et al. 2005). Distributed FOS have proven to be a successful instrument in laboratory monitoring endeavours (Kwon et al. 1998; Van Der Kooi and Hoult 2018). A steel model truss was instrumented with distributed FOS and loaded under two loading conditions to (i) determine FOS capabilities in capturing local strains and (ii) identify effects of different joint conditions. It was determined that the distributed sensors were successful in capturing complex axial and bending strains as multiple FOS were installed on different faces (top and bottom) of the same member. The effects of different joint conditions, namely fixity in connections, on the overall structure was also captured by the FOS (Van Der Kooi and Hoult 2018). Distributed FOS have also been extensively used as a practical field-monitoring application in recent years. Specific to truss monitoring, distributed FOS accurately measured bending moments along the top and bottom chords of an in-service steel truss railway bridge (Van Der Kooi et al. 2018). In addition, the strain-sensing system captured the general behaviour of members as well as the effects of connections, such as gusset plates. FOS applications are well documented; the reader is referred to Bado and Casas (2021) for a complete review of recent civil engineering applications. A practical issue that affects the performance of FOS is the installation procedure including bonding, adhesive types, and sensor placement issues. Overall, both strain measuring technologies are proven methods in both laboratory and field monitoring applications. Notably, FOS may compliment discrete sensors by providing a more general and complete insight into the behaviour of the structure (Barker et al. 2019). Moreover, the distributed capabilities of the FOS capture general behaviour that may not be observed by conventional, discrete, and local instruments such as strain gauges (Van Der Kooi and Hoult 2018).

Lastly, humidity, temperature, and moisture content are all environmental factors that influence the equilibrium moisture content and subsequent shrinking and swelling of timber members (CSA 2015). Temperature sensors have been extensively coupled with other SHM instruments for validation purposes (Tennyson et al. 2001; Locklin et al. 2017). Moisture sensors and humidity sensors have also been incorporated into past timber monitoring programs (Dietsch et al. 2014; Björngrim et al. 2016). Overall, the inclusion of environmental sensors allows for a correlation between environmental conditions and the corresponding building response.

The laboratory stage provided a means to determine the capabilities and limitations of each instrumentation method. Similarly, this research allowed for a comprehensive comparison of instruments in terms of their effectiveness in monitoring timber elements. As such, within the context of an options analysis, a weighted matrix was utilized to compare the various forms of monitoring. Several factors were considered to determine the effectiveness of the instruments including: accuracy; autonomy; real-time, long-term, and continuous capabilities; installation process; site-specific limitations; operational ability on-site; alarm system compatibility; ease of use; training requirements; and overall cost.

A one-quarter scale Warren truss was designed, constructed, monitored, and tested within the Royal Military College (RMC) Civil Engineering structures laboratory as shown in Fig. 3. The truss was constructed with SPF No. 2 2 inch × 4 inch (38 mm × 89 mm) members, placed to model the in-situ truss, with the notable exception of the variable-sized web members in Hangar 1 (see Fig. 1). The connections were composed of half-inch bolts and half-inch malleable iron washers. The assumed orthotropic properties of SPF No. 2 used for the truss and associated with the numerical model are outlined in Table 1.

The instrumentation outlined in Table 2 was utilized, compared, and evaluated throughout the experimental procedure and further employed as part of the in-situ SHM portion of the overall research program. As this paper focusses on the laboratory portion, the reader is referred to Helmer-Smith et al. (2021) for information regarding the in-situ employment of instrumentation within Hangar 1 for local point-load testing. Instrument locations were selected based on preliminary modeling results and several redundant instruments were placed for comparison purposes. As such, the instrumentation could be refined and optimized in terms of the number of sensors as data files were considerable. The multiple forms of SHM instrumentation are shown in Fig. 4. The locations of selected instruments are detailed in Fig. 5.

A QUANTUM X data acquisition system (DAQ) was used throughout the experimental procedure for the conventional instrumentation utilized as part of this investigation. The software employed for the monitoring procedure was Catman^®AP and was compatible with all monitoring instruments except the ATS, LiDAR, and the FOS, as these required specific (proprietary) post-processing software programs. As shown, conventional strain gauges (SG) were installed on several Warren truss members. Multiple gauges were installed on the front (F) and back (B) of the same member (i.e., SG 1(F) and SG 14(B) installed on member L0U1) to determine bending effects. Moreover, the average of the two strain results (given that the gauges are installed front and back) provides the pure axial strain. Employing multiple gauges is also a redundant measure in the case of damaged or unreliable gauges. Finally, the multiple gauges allow for an averaged and more realistic result due to the high variability of timber properties. FOS were employed to capture the internal mechanisms along the entire member, as data were not restricted to localized, discrete points. This provision of distributed strain data is a unique capability of distributed optical sensing techniques. In this experimental procedure, the Luna ODiSI-B (Optical Distributed Sensor Interrogator) was considered. In comparison with other commercially available optical strain sensing technologies, the ODiSI-B can provide an array of strain measurements along a specimen of study. The technology uses a standard, low-cost telecom optical fibre as the transducer and can resolve strain at a spatial resolution (gauge length) as fine as 0.65 mm along a fibre upwards of 40 m in total length (Forbes et al. 2018). Accordingly, the ODiSI-B solution was selected to investigate the strain distribution and potential inhomogeneities across truss members. Limited information was found on using this FOS technology with timber structures. Several installation procedures were trialed. These included, (i) using two adhesives types to bond the FOS: loctite acrylic adhesive and five-minute epoxy resin and (ii) testing two variations of the protective layering along the FOS: nylon protective coating layer (900 μm approximate outer diameter) and acrylate coating (250 μm approximate outer diameter).

Five distributed point-load tests (DPLT), with intermediate columns at varying locations, were conducted throughout the experimental procedure as shown in Fig. 6. Intermediate columns were not utilized for Test 1 and Test 5. Test 2 represents the location of the in-situ intermediate columns in Hangar 1. Varying locations of the intermediate columns were considered for Test 3 and Test 4. The DPLT were conducted using weight plates that were manually placed incrementally at each vertical strut along the truss. The total loading on each vertical strut was 0.7 kN, representing a total load of 4.2 kN. The instruments utilizing Catman^®AP collected data continuously at a frequency of 1 Hz. The ATS, LiDAR, and FOS data were collected at each load increment.

The laboratory stage developed an understanding of the key capabilities and limitations of each instrument. This understanding is crucial in optimizing the SHM program and determining a suitable method for employing the monitoring instruments. Table 3 summarizes the capabilities and limitations of each respective instrument as observed during the laboratory stage. In addition, the error or data variation associated with each instrument is outlined and the requirement for control points or reference instruments is noted. The error in each instrument was determined by variation in control point measurements (ATS and LiDAR) or data noise at zero loading (LVDT, ESG, FOS). Table 3 can be used to analyze and compare SHM instruments for various monitoring projects.

This section summarizes selected results based on the data obtained from the various instruments and monitoring methods employed in this laboratory research program. The data are discussed within the framework of their reliably and accuracy in determining the behaviour of the Warren truss system for each of the testing scenarios.

An objective within this research was to identify and capture the general behaviour of the Warren truss. In doing so, critical monitoring locations and members of interest were determined to optimize the in-situ SHM program.

An initial comparison between the measured FOS strain data on the wooden truss members and that from previous experiments investigating steel supporting members (Forbes et al. 2018) determined the strain measurements to be much less consistent. Two types of protective coatings were trialed along the bottom chord: a nylon protective layer and an acrylate coating. The two optical fibre strain profiles from Test 1 are compared in Fig. 15. As shown, both provide similar strain profiles; however, there is a high level of localized noise. It was hypothesized that this high noise level was a result of inconsistent bonding between the wood face and the fibre due to the inhomogeneity of wood. Thus, warranting further bonding development and testing for FOS installations on wood. Specifically, more testing should be conducted, within a laboratory setting, to determine optimal adhesive types, coating/fibre types, surface preparation procedures, and application/installation techniques for FOS applications on timber members.

Due to the relatively large data files obtained along a single sensor (i.e., thousands of strain measurement points), it would be necessary to restrict the data acquisition rate of the FOS without implementation of local post-processing (e.g., change detection). Such development is not prohibitive using the ODiSI-B, which is controlled by a personal computer, but was not investigated within this research effort. Accordingly, the commercial off-the-shelf FOS solution is less effective as an emergency monitoring technique. The FOS proved to provide reliable distributed strain data along the entire truss during testing, as opposed to the local strain data provided by the strain gauges. Within the scope of this research, FOS was determined to be most effective for control/knowledge monitoring, measuring a strain profile of the truss that is not practically feasible with discrete gauges. In this sense, the FOS was utilized based on its spatial resolution rather than its temporal resolution.

To effectively compare monitoring methods, with an aim to optimize the in-situ monitoring program, an option analysis was conducted. Each respective instrument/method was assigned a numerical value from 1 (lowest) to 5 (highest) for each given parameter. The parameters and respective metrics and (or) criteria that were used to assess the various instrumentation methods included:

Based on the experimental data and lessons learned, the respective instruments were compared using an options analysis based on the abovementioned criteria. The purpose of this analysis is to compare all factors associated with the effective selection of monitoring instruments. The parameters are considered based on the experimental data obtained in terms of accuracy and identifying truss behaviour within the laboratory environment. The lessons learned throughout the experimental procedure were considered and parameters were assigned based on the expected capabilities and limitations in-situ or in real-world installation. For example, the installation process was not a limitation throughout the laboratory endeavour. However, it must be considered greatly for the design of the in-situ program. As such, this metric is applied to the design of the in-situ monitoring project, instead of the laboratory project. Values (1–5) are assigned for the respective instrument for each parameter as displayed in Table 4.

The three primary forms of monitoring include control, knowledge, and emergency monitoring. For this study, control and knowledge monitoring are combined into the same criteria. To determine the optimal instrument for each monitoring form, the above matrix is weighted in terms of what parameters are more important for each respective monitoring case. The weighted percentages are based on how each parameter will affect the efficiency of the monitoring project. For example, autonomy or real-time capabilities are more important and thus weighted much higher for emergency monitoring compared with control. The assigned weights for each case are described in Table 5. The assigned values from Table 4 are then multiplied by the respective weighting and summed as shown in Table 6.

Based on the weighted matrix, in terms of deflection monitoring instruments, the ATS is the optimal instrument for emergency monitoring. Specifically, its autonomy, real-time capabilities, simple installation process, and built-in alarm system capabilities significantly outweigh the other deflection instruments. Moreover, the ATS is still operational during short-term power outages due to its long-term battery capabilities, which is advantageous during emergency events such as snow or ice storms. Conversely, the inherent nonautonomous nature of the LiDAR and extensive post-processing makes it unsuitable for emergency monitoring. However, the LiDAR is suitable for control/knowledge monitoring (and historical record keeping with respect to condition) mainly because it can determine deflections along the entire truss, is only restricted by line of sight, and is easy to set up. The ATS is also deemed suitable for control/knowledge monitoring due its simple installation process, accuracy and reliability, and because site limitations can be avoided. Although the ATS is restricted to local points, multiple targets are easily installed to increase the amount of data points available and provide control points. The LVDT is suitable for control/knowledge monitoring in-situ; however, the requirement for a stable reference system is still a major limitation for long-span trusses. Finally, both the ATS and LiDAR can provide transverse, longitudinal, and vertical deflection data. This can provide much insight regarding the 3-D behaviour of the trusses. Conversely, the LVDT is solely restricted to vertical deflection data at local points. In this regard, the LVDT is more suitable as a redundant instrument to validate ATS or LiDAR results.

In terms of strain monitoring, it was determined that the conventional gauges were more suitable for emergency monitoring compared with the FOS. Mainly, it reliably and accurately preformed in real time, continuously, and in the long term. In addition, the associated software was compatible with alarm systems and was inherently autonomous. Due to the extremely large data files obtained during continuous monitoring, the FOS was not deemed effective for emergency monitoring. However, the FOS proved to provide reliable distributed strain data along the entire truss during testing as opposed to the local strain data provided by the strain gauges. As such, it was determined that the FOS is reliable for control/knowledge monitoring as it can provide reliable information that is not necessarily feasible by the conventional gauges. In particular, the FOS provided valuable data in terms of bending stress as it provided an indication into the behaviour of the global truss and truss connections. In addition, the installation process was significantly easier compared with the strain gauges relative to the information being obtained.