My title https://doi.org/10.14311/APP.2022.36.0149 Acta Polytechnica CTU Proceedings 36:149–160, 2022 © 2022 The Author(s). Licensed under a CC-BY 4.0 licence Published by the Czech Technical University in Prague ON THE EFFECTS OF MODIFIED LANE CONFIGURATION DUE TO CONSTRUCTION WORKS ON BRIDGE LOADING Marcel Nowak∗, Oliver Fischer Technical University of Munich, Chair of Concrete and Masonry Structures, Theresienstrasse 90/N6, 80333 Munich, Germany ∗ corresponding author: marcel.nowak@tum.de Abstract. Construction works are frequently recurring events within highway networks, temporarily altering the characteristics of the traffic flow and hence of the bridge loading process. Consequently, highway bridge superstructures usually carrying unidirectional traffic might have to face modified configurations, with possibly altered lane widths and transverse lane positions, or even an increased number of lanes and the occurrence of bidirectional traffic flow. While usually covering only short periods compared to the average service life of a bridge, such scenarios might represent extreme loading situations, potentially dominating the traffic loading process. Therefore, their consideration and accurate modeling in the course of traffic load assessment and development of object-specific traffic load models are of utmost importance. The presented study analyzes data from bridge monitoring along federal highway A92 in Southern Germany. The data covers standard and different modified lane configurations due to construction works. Besides analyzing the effects on traffic flow characteristics and resulting occurrence rates of multiple truck presence on the bridge, the consequences for load effects and their extreme value behavior are also evaluated. The findings support verification of related code background works and can serve as base and guidance for future research efforts on bridge load modeling in this context. Keywords: Bridge loading, modified lane configuration, monitoring, multiple truck presence, transverse distribution factor. 1. Introduction Throughout the regular service life of a highway bridge, occasional construction works along the road might lead to a temporary modification of traffic lane config- uration on the roadway, with possibly altered number of lanes, lane widths, and transverse lane positions. These modifications potentially cause changes in the characteristics of the traffic flow and the loading of the bridge, i.e. the quantity and position of truck vehicles simultaneously present on the bridge. Therefore, such scenarios might represent extreme loading situations, potentially dominating the bridge loading process. Consequently, the possibility of modified lane con- figurations on a bridge needs to be considered for code calibration of traffic load models. In the back- ground works for the German national annex to EN 1991-2 and the reassessment guideline for existing road bridges (Nachrechnungsrichtlinie), this scenario is accounted for [1, 2]. A modified lane configura- tion with bidirectional traffic on four traffic lanes is considered, with a ratio of 20 % of truck vehicles per direction traveling on the respective fast lane, allowing for the occurrence of truck overtaking events for each driving direction. Additional investigations of altered bidirectional four-lane configurations are performed in [3]. Based on a (more realistic) reduced width of the respective fast lane, the assumption is made that truck overtaking is not possible due to spatial con- straints. The assumed absence of truck vehicles on the fast lane (no overtaking trucks) reduces the level of bridge loading. Investigations within the scope of mentioned background works are performed based on numerical simulations and for a limited set of bridge structures. The resulting load levels are evaluated based on extrapolated characteristic values with a re- turn period of 1000 years (even though the occurrence period of modified lane configurations is considerably shorter compared to the standard lane configuration). Measurements from an actual structure to compare and verify the effects of modified lane configuration on traffic flow characteristics and bridge loading are unknown to the author. Within this context, the presented study analyzes monitoring data from a highway bridge along federal highway A92 in Southern Germany. The monitoring data covers - besides standard two-lane configuration - two periods with different modified lane configura- tions due to constructions works along the considered highway section: a modified unidirectional two-lane configuration and a bidirectional four-lane configura- tion. Direct measurement of the resulting load effects with strain gauges and additional traffic monitoring with a laser scanner for vehicle detection, counting, and classification are performed. The monitoring data allows for an in-depth comparison of the modified lane configurations with the standard case. Besides ana- lyzing the effects on traffic flow characteristics and resulting occurrence rates of multiple truck presence 149 https://doi.org/10.14311/APP.2022.36.0149 https://creativecommons.org/licenses/by/4.0/ https://www.cvut.cz/en Marcel Nowak, Oliver Fischer Acta Polytechnica CTU Proceedings on the bridge, consequences for load effects and their extreme value behavior are also evaluated. The find- ings from this study support verification of related code background works and can serve as base and guidance for future research efforts on bridge load modeling accounting for scenarios of modified lane configurations. 2. Methods 2.1. Signal Processing The raw signal of strain sensors is composed of differ- ent parts, mainly differing in their rate of change: • load effects due to temperature and other non- permanent constraining loads (low rate of change, usually several minutes or longer) • road traffic (static) load effects (medium rate of change, for structure and sensor locations in this work up to a few seconds) • dynamic effects due to excitation by passing vehicles and interaction with structure (high rate of change, fractions of second) • measurement noise (wide range of rates of change, possibly over entire frequency spectrum) Signal processing is necessary to extract the relevant parts of the raw signal resulting from road traffic, i.e., passing vehicles’ (static) load. For this purpose, the procedure from [4, 5] is adopted within the scope of this work, performing signal decomposition exploiting the different rates of change of the single signal com- ponents. For this purpose, mean filtering (excluding parts of the signal with a low rate of change) and low pass frequency filtering (excluding dynamic effects and most parts of measurement noise) are applied. Mean filtering is done based on block-wise deter- mined and linearly interpolated mean values. The choice of block size constitutes a compromise between sufficiently detailed representation of the signal parts with a low rate of change and ‘robustness’ against the influence of the signal part due to traffic loading on the resulting block mean values. For low pass frequency filtering, a Butterworth filter of seventh order is ap- plied. The cutoff frequency fcutoff is chosen based on the natural frequencies of the monitored structure. As the focus is on analyzing the load effects result- ing from the static load of passing vehicles, the value should be sufficiently low to exclude all significant parts of the signal due to dynamic effects. However, it is also essential to choose the value for fcutoff not too low to avoid significant alteration of relevant parts of the signal. Depending on the velocity of a passing vehicle, the signal part due to its static load results in low-frequency contents of the signal that could be affected by the low pass filter [5]. It is acknowledged by the authors that there are more sophisticated and accurate methods for signal processing and decomposition to extract the compo- nent representing static traffic load effects. However, the achieved accuracy of the previously described pro- cedure is considered sufficient for the objectives of this work. 2.2. Mulitple Truck Presence According to [6], a bridge loading event (BLE) is de- fined as the presence for a continuous period of time of at least one truck on the influence area of the load effect of interest, meaning two successive events are separated by a time gap with no truck present on the influence area. In other words, a BLE consists of a sequence of vehicles consecutively arriving at a bridge (in the same or different traffic lanes and driving di- rections) with at least one vehicle being present on the structure at each instant of time during the event. In the context of multiple truck presence evaluation, BLEs can be described by following characteristic values based on the number of contributing truck vehicles: • total number of truck vehicles forming part of the corresponding BLE [7]. • number of truck vehicles present on the bridge for a specific instant of time during a BLE. • number of truck vehicles contributing to the max- imum value of a specific load effect during a BLE [8]. While the first two values solely depend on the traffic flow and the length of the considered bridge structure (response independent characterization), the third value is specific to a certain load effect, and hence depending on the structural response (for the same BLE, this value can vary for different load effects). BLEs can be further categorized into following types, based on their transverse location and rela- tive driving direction on the bridge: • in-lane event, with multiple truck presence on the bridge within the same traffic lane • overtaking event, with multiple truck presence on the bridge parallel on adjacent traffic lanes in uni- directional traffic • meeting event, with multiple truck presence on the bridge parallel on adjacent traffic lanes in bidirec- tional traffic Arbitrary combinations of the different event types are also possible. For the scope of this work, multiple truck pres- ence is evaluated based on monitoring data from both structural and traffic monitoring. Data from traffic monitoring provides detailed information on the traf- fic flow over the bridge, with timestamps for every passing vehicle indicating its entrance to and exit from the measurement section. Knowing the bridge length, the location of the measurement section along the bridge, and the velocity of each vehicle, timestamps for entrance to and exit from the bridge deck can be obtained. Based on this data, detection of BLEs 150 vol. 36/2022 On the Effects of Modified Lane Configuration and their (response independent) characterization is straightforward. Data from structural monitoring represents the structural response (e.g. strains) at selected sensor locations due to the traffic flow over the bridge. These load effect time histories contain no direct informa- tion on the causative load impact. However, data fusion from multiple sensors also allows for this data evaluation towards multiple truck presence. In a first step, the processed signal (extracted component due to static traffic loads) is analyzed to identify single BLEs. In approximative accordance with the pre- vious definition, BLEs are identified as continuous sections of the traffic component of the strain signal exceeding a predefined threshold (continuous presence of vehicles on bridge causing significant structural response). For each identified BLE, number and rela- tive position of trucks on the bridge contributing to the maximum load effect εk,max measured at sensor k (response-sensor) at time tεk,max is estimated based on transverse distribution factors (TDFs) fT DF,k|i. These factors describe the relative share of structural response εi(t = tεk,max ) for a representative set of n selected sensors (TDF-sensors) at same longitudinal but different transverse positions and indicate the transverse position of passing vehicles: fT DF,k|i = εi(t = tεk,max ) n∑ j=1 εj(t = tεk,max ) . (1) This concept is based on the assumption of linear elastic structural behavior under usual traffic loading. In this case, the transverse distribution of structural response depends only on the transverse position of passing vehicles and not on their total weight [9]. The knowledge of this (response dependent) characteri- zation of BLEs is essential for subsequent extreme value analysis (EVA), as different types of BLEs rep- resent different loading processes to be accounted for separately. 2.3. Extreme Value Analysis For EVA, two different methods are applied based on results from previous works [10, 11], the block maxima (BM) method with fit of generalized extreme value (GEV) distribution and the level crossing counting (LCC) method with fit of Rice’s formula. By applying the BM method, data from load effect time histories is blocked into intervals of a chosen size. The maxima in each interval are determined, generat- ing a series of BM. A GEV distribution can then be fitted to this series [12]. By differentiating the load ef- fects according to the underlying BLE types, separate BM series can be obtained. A different GEV distribu- tion is fitted to each of these series, jointly composing a composite distribution model for the BM series from a mixture of different BLE types. Each BLE type con- sidered for the composite distribution model must occur at least once per block [6]. Statistical inference is performed applying maximum likelihood estimation to obtain the parameters of the fitted GEV distribu- tions. The estimation is performed by minimizing the respective negative log-likelihood expressions for the BM data. For the LCC method, Rice’s formula is fitted to the upper tail of an outcrossing rate histogram. Perform- ing LCC, the number of times is counted at which positive values are crossed upwardly in a load effect time history. By normalizing the resulting level cross- ing histogram to the time history length, the outcross- ing rate histogram is obtained, representing for each level the mean rate of its crossing during the reference period. Rice’s formula describes the mean rate of up- crossing for a certain level during a reference period and is fitted to the outcrossing rate histogram. Rice’s formula is fitted only to the significant tail regions, and the proper choice of the starting point for Rice’s formula fit is crucial to the method. It should be as low as possible to ensure sufficient representative- ness for statistical extrapolation but not too low to provide still a reasonable approximation of the sig- nificant tail region. An optimal starting point can be identified by evaluating the goodness of fit using a modified Kolmogorov test. Statistical inference is performed applying the least square method to obtain the parameters of the fitted Rice’s formula. [13] 3. Bridge Structure and Monitoring 3.1. Bridge Structure 29/1 Bridge structure 29/1 is an overpass of federal highway A92 between the cities of Deggendorf (DEG) and Munich (MUC) in the Southern part of Germany, crossing an outlet ditch in five spans of uniform length of about 16 m. The bridge comprises two separate, identical superstructures, each carrying the roadway for one driving direction with a standard configuration of two traffic lanes plus emergency lane. The cross- section of each superstructure consists of five precast prestressed concrete t-beam girders of 69 cm height with a cast-in-situ slab of 23 cm height on top. The precast girders lie on elastomer bearings and cross each a single span without a rigid connection between them from one span to the next. Only the in-situ slab on top is continuous over all five spans. This design reduces the superstructure’s stiffness at the piers, making each span act approximately as an independent single-span system by itself. The total width of one superstructure is 15.35 m, the width of the roadway is 12.0 m (refer also to Figures 1 and 2). 3.2. Alternating Lane Configurations Federal highway A92 can be considered a minor section within Germany’s highway network, with moderate traffic volumes. Throughout the year 2019, construc- tion works for renovation of the roadway are carried 151 Marcel Nowak, Oliver Fischer Acta Polytechnica CTU Proceedings girder 1 span 1 span 2 span 3 span 4 span 5 dir. MUC girder 2 girder 3 girder 4 girder 5 012345 16.00 79.90 16.00 15.9516.0015.95 Figure 1. Top view of bridge structure 29/1. gi rd er 1 gi rd er 2 gi rd er 3 gi rd er 4 gi rd er 5 1.12 1.12 10 40 10 12.00 15.30 1.65 1.65 1.42 2.86 2.86 2.86 2.86 1.42 14.30 2 3 6 9 Figure 2. Section of bridge structure 29/1. dir. MUC lane 1 dir. MUC lane 2 3.75 75 3.753.75 emergency lane dir. MUC lane 1 dir. MUC lane 2 3.50 6.002.25 25 closed to normal traffic (occasional construc�on traffic) dir. MUC lane 1 dir. MUC lane 2 dir. DEG lane 2 dir. DEG lane 1 3.50 3.502.502.25 25 Figure 3. Lane configurations LCstan (top), LCmod,uni2 (middle), and LCmod,bi4 (bottom). 152 vol. 36/2022 On the Effects of Modified Lane Configuration 012345 4-1 4-2 4-3 4-4 4-5 5-2 5-4 3-2 3-4 2-2 2-4 1-2 1-4 4-1T 4-3T 4-28 4-27 4-18 4-17 4-22 4-21 4-12 4-11 4-24 4-23 4-14 4-13 4-26 4-25 4-16 4-15 8.00 8.00 7.988.007.98 7.988.008.007.98 2.80 2.802.40 Figure 4. Overview of sensor layout of structural monitoring at bridge structure 29/1. 80 1.26 80 1.00 86 1.00 1.00 86 1.00 80 1.26 80 thermo sensorstrain gauge (local sensor)strain gauge (global sensor) 2.86 2.86 2.86 2.86 Figure 5. Sensor locations within bridge section of span no. 4 (top) and remaining spans (bottom). out in the section of the highway where bridge struc- ture 29/1 is located, leading to alternating lane con- figurations on the bridge. Besides the standard config- uration of unidirectional traffic with two traffic lanes plus emergency lane per superstructure (LCstan), a modified unidirectional two-lane configuration with re- duced roadway width (LCmod,uni2) and a bidirectional four-lane configuration on a single superstructure with closure of the second superstructure (LCmod,bi4) occur throughout the construction works (Figure 3). Even though a part of the roadway is closed for regular traf- fic in configuration LCmod,uni2, construction vehicles pass the bridge in this part For expected multiple truck presence on the bridge, the occurrence of in-lane events is excluded from the scope of this work. Due to the short span length of 16 m and the approximate independence of each span from the rest of the structure (refer to Section 3.1), the influence line for the considered load effects (strain at bottom of girder at mid-span) is of restricted length. Hence, the contribution of an additional vehicle in the same lane is not significant. In this case, the char- acteristics of the loading process for in-lane events are similar to those of single vehicle events (SVEs). Furthermore, overtaking events for LCmod,uni2 and LCmod,bi4 are likely not to occur due to spatial con- straints by the reduced roadway width, making it almost impossible for two truck vehicles passing the bridge parallel in usual velocities. Therefore, overtak- ing events are expected only for LCstan. Additionally, multiple truck presence is expected for LCmod,bi4 by meeting events in the bidirectional traffic. Besides oc- currence due to regular highway traffic, the occasional passing of construction vehicles on the closed part of the roadway in configuration LCmod,uni2 also poses a source of multiple truck presence on the bridge. 3.3. Structural Monitoring The sensors for structural monitoring are installed on the northern superstructure, which carries the traffic with driving direction towards MUC. The monitoring concept comprises 31 sensors (29 strain gauges, two thermal sensors). The sensor layout of the strain sensors consists basically of two components: sensors for global structural response at bottom of webs of t-beams at mid-span, and sensors for local structural response at bottom of flange of t-beams at quarter- span (refer to Figures 4 and 5). The layout was designed for the realization of brigde weigh-in-motion (BWIM). The BWIM application and evaluation of related results are, however, beyond the scope of this work. In the following, the focus is on strain data from three global sensors in span no. 4 (sensors 4-2, 4-3, and 4-4). For the signal processing (Section 2.1), the block size for the mean filtering is chosen to 10 min. The cutoff frequency fcutoff for the low pass frequency filtering is chosen based on natural frequencies of the moni- tored structure. The eigenvalues of the structure are determined using operational modal analysis (OMA), and are about 6.7 Hz and 7.7 Hz for the first two rele- vant eigenmodes [14]. For further analysis, fcutoff is chosen as 6.0 Hz. The TDF threshold values for response dependent characterization of BLEs identified from structural monitoring data are derived from proof load testing of the bridge. Multiple runs with three different proof load vehicles with known weights and dimensions are 153 Marcel Nowak, Oliver Fischer Acta Polytechnica CTU Proceedings 0 2 4 6 8 10 12 transverse position on roadway (m) 0.0 0.2 0.4 0.6 0.8 1.0 f T D F ,k |4 2 L C st a n d ir M UC , l an e 1 L C st a n d ir M UC , l an e 2 L C m od ,u n i2 / bi 4 d ir M UC , l an e 1 L C m od ,u n i2 / bi 4 d ir M UC , l an e 2 L C m od ,b i4 d ir DE G, la ne 2 L C m od ,b i4 d ir DE G, la ne 1 proof load vehicles (fictitious) regular truck vehicles Figure 6. TDFs fT DF,k|4-2 from proof load testing and evaluation of regular truck vehicles with reference influence line (RIL). vehicle axle weight (t) spacing (m) crane 1 12.0/12.0/12.0/12.0 1.65/1.90/1.75 crane 2 11.6/11.5/12.6/12.5 1.65/2.05/1.65 truck 7.9/8.7/9.4/6.1 1.35/3.00/1.80 Table 1. Parameters of proof load vehicles. performed (refer to Table 1). Based on the RILs derived from the strain data of the proof load events [5], additional fictitious load patterns representing truck vehicles from regular traffic are investigated (refer to Table 2). Theoretically, TDFs can be evaluated for any set of TDF-sensors at same longitudinal but different trans- verse positions, according to Equation (1). For the scope of this work, the set comprising sensors 4-2 and 4-4 proved to be a proper choice for TDF-sensors able to differentiate clearly between the several transverse position in different traffic lanes for different lane configurations (refer to Figure 6 for TDFs fT DF,k|4-2, values for fT DF,k|4-4 complement to 1). For all con- sidered response-sensors k ∈ [4-2, 4-3, 4-4], instants tεk,max are nearly identical, and therefore fT DF,k|i result in equal values. 3.4. Traffic Monitoring Besides the previously described structural monitor- ing, an additional traffic monitoring unit is installed. This unit consists of a laser scanner for vehicle de- tection, counting, and detailed classification and a webcam providing image data of the traffic flow on the bridge for comparison and verification purposes. The monitoring unit is placed on top of a telescop- ing mast right next to the northern superstructure in span no. 4, about 8 m above the top of roadway (see Figure 7). Due to lateral installation of the monitor- ing unit, this height is necessary to avoid shadowing ID axle weight (t) spacing (m) 11 9.0/11.0 4.5 12 6.8/7.6/5.5 4.3/1.3 1111 10.3/14.9/7.6/7.2 4.9/6.5/5.0 1211 8.4/10.3/6.4/7.8/7.1 4.6/1.3/5.2/4.6 112 12.2/12.4/7.6/7.8 3.7/6.6/1.3 113 8.3/11.2/6.8/6.8/6.8 3.7/5.6/1.3/1.3 14 12.0 (all) 2.6/1.7/1.7/1.7 222 12.0 (all) 1.9/2.9/1.7/2.6/1.7 111 6.0/8.3/10.8 5.9/6.0 11 7.2/12.8 5.9 Table 2. Parameters of regular truck vehicles (data adopted from [1, 4], ID corresponds to sequence of axle groups with number of axles per group). effects and provide the best possible quality of vehicle detection and classification. Figure 7. Traffic monitoring unit next to bridge. The laser scanner used in the measurement unit is the profiling system TIC501 from SICK AG. Its 2D light detection and ranging (LIDAR) sensor works with multiple echo technology, making it robust against atmospheric exposure. It detects the travel 154 vol. 36/2022 On the Effects of Modified Lane Configuration lane of the vehicles and scans passing vehicles up to 100 times per second, creating a 3D point cloud of the vehicles’ silhouette serving as a base for further classification. The profiling system can cover up to four traffic lanes. Furthermore, vehicle dimensions and velocity are recorded, as well as the timestamp of entrance to and exit from the scanning section allowing for determination of inter-vehicle gaps and reconstruction of actual traffic flow. [15] 4. Results from Monitoring Data Analysis For the scope of this work, complete daily data sets for the different lane configurations for the following periods are available: • LCstand: 22.11.2019 until 18.12.2019 • LCmod,uni2: 25.10.2019 until 18.11.2019 • LCmod,bi4: 11.10.2019 until 23.10.2019 For further analysis, only data from working days within the mentioned periods is considered. 4.1. Multiple Truck Presence on the Bridge As previously stated in Section 3.2, the possibility of overtaking events in the same driving direction for LCmod,uni2 and LCmod,bi4 can be excluded due to spatial constraints by the reduced roadway width. Analysis of the traffic monitoring data for the different lane configurations confirms this statement (Figure 8). -101 0 101 102 103 < 16 m LCstan, dir. MUC < 16 m LCmod, uni2, dir. MUC -101 0 101 102 103 -101 0 101 102 103 < 16 m LCmod, bi4, dir. MUC -101 0 101 102 103 < 16 m LCmod, bi4, dir. DEG distance to preceding vehicle (m) di st an ce to fo llo w in g ve hi cle (m ) Figure 8. Spatial gaps of vehicles on fast lane to adjacent vehicles in slow lane, for different lane con- figurations. For each truck traveling on the fast lane, the spatial gap to its preceding (distance from front of vehicle in fast lane to end of vehicle in slow lane) and following LCstan LCmod, uni2 LCmod, bi4 0.0 0.2 0.4 0.6 0.8 1.0 f T D F ,k |4 2 MUC-L1 MPE-PE MUC-L2 0.50 0.25 MUC-L1&2 MPE-ME(CV) DEG-L1&2(CV) 0.65 0.35 MUC-L1&2 MPE-ME DEG-L1&2 0.65 0.35 proof load vehicles (fictitious) regular truck vehicles Figure 9. Definition of threshold values for MPE identification for different lane configurations. (distance from end of vehicle in fast lane to front of vehicle in slow lane) truck on the slow lane is determined, based on timestamps and velocities from the laser scanner data record. If the vehicle in the fast lane is overtaking another vehicle on the slow lane while crossing the bridge, either of these spatial gaps takes a small value (threshold value of 16 m corresponding to span width of bridge structure). The results in Figure 8 show that in general, there are only very few vehicles traveling in the left lane for LCmod,uni2 and LCmod,bi4 being hardly part of an overtaking event. For LCstan, it can be seen, however, that significantly more vehicles travel on the fast lane and that the vast majority of them are overtaking other vehicles. Based on the evaluation of TDFs for proof load testings and fictitious regular truck vehicles (see Sec- tion 3.3), and explanations regarding possible types of multiple presence events (MPEs) for the different lane configurations (see Section 3.2), appropriate threshold values for fT DF,k|4-2 are defined to allow for MPE identification (refer to Figure 9). For the different lane configurations, following classifications of BLEs are specified: • LCstan: SVEs on lane 1 or lane 2 (MUC-L1 and MUC-L2 ), MPEs as overtaking events (MPE-OE) • LCmod,uni2: SVEs on direction MUC (MUC-L1&2 ), SVEs due to possible construction vehicles on direc- tion DEG (DEG-L1&2 ), MPEs as meeting events with regular traffic on direction MUC and possi- ble construction vehicles on direction DEG (MPE- ME(CV)) • LCmod,bi4: SVEs on direction MUC and direc- tion DEG (MUC-L1&2 and DEG-L1&2 ), MPEs as meeting events (MPE-ME) Due to spatial constraints for LCmod,uni2 and LCmod,bi4, and the resulting decreased probability of trucks travelling in the fast lane, no differentiation is made for these lane configurations for vehicles trav- elling either in lane 1 or lane 2 within the same driving direction. 155 Marcel Nowak, Oliver Fischer Acta Polytechnica CTU Proceedings 0 50 100 ε 4 2 ·1 0 6 LCstan threshold for multiple presence events all bridge loading events block maxima events (block size 6h) LCmod, uni2 LCmod, bi4 0 50 100 ε 4 3 ·1 0 6 0.0 0.2 0.4 0.6 0.8 1.0 0 50 100 ε 4 4 ·1 0 6 0.0 0.2 0.4 0.6 0.8 1.0 fTDF, k|4 2 0.0 0.2 0.4 0.6 0.8 1.0 Figure 10. TDFs fT DF,k|4-2 evaluated for all BLEs identified from structural monitoring data of working days (BM events with blocks size 6 h highlighted), for different lane configurations and considered response-sensors. MUC-L1 MPE-OE MUC-L2 0 50 100 MUC-L1 MPE-OE MUC-L2 traffic monitoring structural monitoring (4-2) structural monitoring (4-3) structural monitoring (4-4) MUC-L1&2 MPE-ME DEG-L1&2 0 50 100 MUC-L1&2 MPE-ME DEG-L1&2 m ea n ra te p er h ou r Figure 11. Mean hourly occurrence rates of different BLE types for lane configurations LCstan and LCmod,bi4, evaluated for working days. The TDF evaluation for the three different lane configurations is shown in Figure 10. Based on the available data from structural monitoring, relevant BLEs are identified and classified according to the previously defined criteria for TDFs. The results gen- erally confirm the choice of TDF-thresholds for MPE identification. For all three lane configurations, dense accumulations of data points can be observed beyond the thresholds, representing the respective SVEs dom- inating (in the sense of occurrence rates) the bridge loading process in all cases. MPE occurrence in sig- nificant numbers is observed for lane configurations LCstan (overtaking events) and LCmod,bi4 (meeting events). For lane configuration LCmod,uni2, the vast majority of BLEs are SVEs on driving direction MUC. However, due to the occasional passing of construction vehicles on the closed part of the roadway, small num- bers of SVEs on driving direction DEG and MPEs (meeting events) can be observed. For further quantification of MPE occurrence, the mean hourly occurrence rate of SVEs and multiple truck presence on the bridge is evaluated and com- pared for lane configurations LCstan and LCmod,bi4 (Figure 11). While slightly differing in absolute num- bers, the occurrence rates obtained from traffic moni- toring (based on data from laser scanner) and struc- tural monitoring (based on data from strain sensors) show similar tendencies. It can be observed that MPE occurrence rates are significantly higher for bidirec- tional traffic (meeting events) than for unidirectional traffic (overtaking events), but still relatively small compared to occurrence rates of SVEs. 156 vol. 36/2022 On the Effects of Modified Lane Configuration 0 2 4 k = 4 2 LCstan BM mixed BM for MUC-L1 / MUC-L1&2 BM for MPE-PE / MPE-ME BM for MUC-L2 / DEG-L1&2 LCmod, uni2 LCmod, bi4 0 2 4 k = 4 3 0 20 40 60 80 100 0 2 4 k = 4 4 0 20 40 60 80 100 εk · 10−6 0 20 40 60 80 100 − lo g( − lo g( C D F )) Figure 12. BM data (block size 6 h) of working days for considered sensors and different lane configurations; with and without differentiation of different BLE types. 4.2. Load Effects and Extreme Value Behavior For analysis of the resulting load effects and the im- pact of multiple truck presence, BM data (block size 6 h) is obtained from the structural monitoring data for the different lane configurations, with and without differentiation of different BLE types (refer to Fig- ure 12). The BM data reveals only for a few cases actual significance of MPEs, e.g. for sensor 4-4 in LCstan. In most cases, the bridge loading process is dominated by SVEs, primarily by one SVE type alone (e.g. MUC-L1 for sensor 4-2 in LCmod,stan) or sporadi- cally by a mixture of two SVE types (e.g. MUC-L1&2 and DEG-L1&2 for sensor 4-3 in LCmod,bi4). These observations are supported by illustrations in Fig- ure 10 for the highlighted BM data. Regarding lane configuration LCmod,uni2 it can be observed that the regular traffic (MUC-L1&2) dominates the structural response of sensors 4-2 and 4-3, while data from sensor 4-4 reveals the dominant influence of the occasional construction vehicles passing the bridge on the closed part of the roadway. Concerning the absolute load level, data shows only slight variation with chang- ing lane configurations for sensor 4-2. In contrast to that, sensor 4-3 experiences its most severe loading for LCstan, whereas sensor 4-4 faces the heaviest loading for LCmod,bi4. This has mainly to do with the change of transverse position of the traffic lanes on the road- way relative to the single sensor locations, as for all three sensors, the data shows no significant impact of MPE for lane configurations leading to the most adverse loading. Regarding the extreme value behavior, extrapolated values with a return period of 1000 years are deter- mined applying the different methods mentioned in Section 2.3. Due to the limited amount of available method ε4−2 ε4−3 ε4−4 LCstan BM-mix 146.9 144.6 74.3 BM-comp 146.9 154.7 89.1 LCC 158.9 156.1 121.6 LCmod,uni2 (incl. CV) BM-mix 110.8 75.1 342.2 BM-comp − − − LCC 130.3 92.4 114.7 LCmod,uni2 (excl. CV) BM-mix 110.8 75.3 33.1 BM-comp 110.8 75.3 33.1 LCC 130.3 92.5 42.3 LCmod,bi4 BM-mix 112.0 83.4 111.5 BM-comp 111.6 90.3 111.5 LCC 145.2 122.2 149.4 Table 3. Extreme value estimates with different methods for 1000 years return period. data for the BM method, a choice of a usual block size of 1 d is not feasible. However, a block size chosen too small (e.g. 1 h) might not fulfill the requirement of statistical independence of the different BM events necessary for application of BM method for EVA. As a compromise, a block size of 6 h is chosen for the scope of this work. For lane configuration LCmod,uni2, extrapolation is performed for both cases consider- ing and neglecting BLEs resulting from construction vehicles on the closed part of the roadway. The ex- trapolation results are shown in Table 3. It can be observed that both BM methods with (BM-comp) and without (BM-mix) differentiation of different BLE types lead to quite similar results in many cases. The reason for this is that the bridge loading process is dominated by a single BLE type in most cases, as previously explained. Both BM meth- 157 Marcel Nowak, Oliver Fischer Acta Polytechnica CTU Proceedings ods face problems for LCmod,uni2 when considering BLEs due to occasional construction vehicles. Due to their irregular occurrence on the bridge, the re- lated BLE types do not occur in every block, making application of method BM-comp unfeasible. More- over, structural response at sensor 4-4 is dominated by two different BLE types in this lane configuration, causing ill-conditioning for the method BM-mix. In general, the most robust results seem to be provided by LCC method, which makes more use of the data by considering each loading event, and not just a few per block. A tendency of slightly higher extrapolated values compared to BM methods can be observed. Al- together, the previously identified tendencies for the measured load effects are approximately reflected in the results of EVA again. Also, the importance of BLE due to occasional construction vehicles for sensor 4-4 in LCmod,uni2 is clearly shown. However, in parts significant variation of results between the different methods of EVA can be observed. 4.3. Discussion of Results Analysis of traffic flow characteristics and resulting occurrence rates of multiple truck presence on the bridge show that the modified lane configurations have a contrastive impact. On the one hand side, reduced available roadway width leads to a disappearance of overtaking events compared to the standard lane configuration. However, in the case of bidirectional traffic on the structure, additional multiple presence events occur in the form of meeting events. The mean occurrence rates of meeting events in bidirectional traffic are significantly higher than those of overtaking events in unconstrained unidirectional traffic. This can be explained by the different volumes of available truck vehicles potentially involved in the respective MPEs. While in unconstrained unidirectional traffic only relatively few truck vehicles travel on the left lane and possibly are part of an overtaking event, a significantly larger number of truck vehicles passes the bridge in the opposite direction in bidirectional traffic that potentially cause a meeting event. The correla- tion effect of vehicles involved in overtaking events – if a truck travels on the left lane, it is very likely part of an overtaking event instead of crossing the bridge alone (data analysis shows that 65 % to 75 % of trucks traveling in the fast lane are involved in an overtaking event) – usually would lead to a higher occurrence rate. However, this effect is outweighed in this case by the significant increase in the volume of possibly involved truck vehicles in meeting events (even though here the vehicles are entirely uncorrelated). Regarding the resulting load effects, it is found that modifications of the lane configuration on the bridge noticeably alter the characteristics of the bridge load- ing process. However, the related change in occurrence rates of multiple presence events has only for a few cases of considered load effects and lane configurations significant impact on the extreme value behavior. Due to the distinct biaxial structural behavior of the consid- ered bridge structure in combination with the limited span width compared to the usual length of regular truck vehicles, it is rather the change of transverse position of the traffic lanes on the roadway playing a more dominant role compared to multiple truck presence. Therefore, the loading process for the differ- ent lane configurations is dominated by single vehicle events in most scenarios. For bridges with different di- mensions and structural systems (uniaxial rather than biaxial structural behavior of superstructure), oppo- site characteristics are expected, with multiple truck presence playing a more dominant role in the loading process. A particular case is observed in the study for the modified unidirectional two-lane configuration, with occasional bridge crossings of construction ve- hicles on that part of the roadway closed to regular traffic. Even though their occurrence is sparse, they dominate the loading process for specific load effects. Concerning the results of extreme value analysis, in parts significant variation of results between different extrapolation methods can be observed. This raises the question of accuracy of the different extrapolation methods and their appropriateness for application in this study, especially in the context of the rela- tively small extent of available monitoring data and the limited degree of knowledge on the actual load impact leading to the measured structural response (knowledge on underlying loading processes). The LCC method with fit of Rice’s formula makes the best use of the available data by considering each loading event and seems to provide the most robust results over the range of investigated load effects and lane configurations. However, the evaluation of the accu- racy of the individual extrapolation results is beyond the scope of this study. 5. Conclusions The presented study analyzes data from bridge mon- itoring along federal highway A92 in Southern Ger- many. The data covers standard and different modi- fied lane configurations due to construction works. Regarding the impact on multiple truck presence on the bridge, the results from the study confirm the basic assumption from code background works that a realistic, reduced width of the fast lane in modified lane configurations effectively prevents truck vehicles from overtaking. Analysis of the monitoring data proves an almost complete disappearance of overtaking events for both modified configurations in this study. Furthermore, it is found that the mean occurrence rates of meeting events in bidirectional traffic are significantly higher than those of overtaking events in unconstrained unidirectional traffic. This observation can be considered representative for typical highway traffic characterized by small ratios of truck vehicles traveling on the left lane compared to the right lane (for unconstrained unidirectional traffic) and similar 158 vol. 36/2022 On the Effects of Modified Lane Configuration truck traffic volumes for both driving directions (for bidirectional traffic). How the change of lane configuration alters the bridge loading process heavily depends on its govern- ing aspects, which are determined by the character- istics of the bridge structure itself. For the bridge in this study (distinct biaxial structural behavior, short span width), single vehicle events play the dominant role in most scenarios. The loading processes for the considered load effects are governed rather by the change of transverse position of the traffic lanes on the roadway than by multiple truck presence on the bridge. Therefore, it is crucial for traffic load model- ing to accurately represent the single truck vehicles’ parameters (i.e. axle weights and spacings) and their transverse position on the roadway. In the particular case of a lane configuration with partial closure of the roadway, it is either essential to consider possible construction vehicles in this part or to ensure that the closed part of the roadway is, in fact, free of any type of vehicle (e.g. in case of partial roadway closure as a constraining measure to reduce the level of traffic loading on the bridge). Extreme value analysis proves to be challenging due to the relatively small extent of available data and limited knowledge on underlying loading processes. Application of level crossing counting method with fit of Rice’s formula and classification of bridge loading events based on transverse distribution factors towards composite distribution models for block maxima aim to make the best use of the available data. Never- theless, results from different extrapolation methods show considerable variation, and straightforward iden- tification of the most accurate and reliable results is not feasible. Therefore, statistical extrapolation is to be performed with great care. Further research efforts beyond the scope of this paper are necessary, i.e. towards adopting techniques allowing for com- paring the accuracy of the results from the different extrapolation methods based on a given data set of limited extent. Further future research efforts should focus on extending investigations on the impact of modified lane configuration on the bridge loading process to a broader range of different types of bridge structures. Especially for bridge superstructures with a rather uniaxial than biaxial structural behavior, multiple truck presence is expected to play a more dominant role in the loading process. Moreover, a more realistic consideration of scenarios with modified lane config- urations could exploit further potential in the scope of object-specific traffic load models. By quantifying the expected occurrence periods of different modified lane configurations throughout regular bridge lifetime, and consistent consideration in bridge load modeling besides the standard configuration based on findings from this study, a more accurate estimate of extreme bridge loading throughout the service life of a bridge structure can be obtained. Acknowledgements The research this work is based on is funded by Die Auto- bahn GmbH des Bundes, branch Southern Bavaria. References [1] U. Freundt, S. Böning. Teil 1: DIN-FB 101: Einwirkungen auf Brücken. In Anpassung von DIN-Fachberichten Brücken an Eurocodes, Berichte der Bundesanstalt für Straßenwesen: Brücken- und Ingenieurbau Heft B 77, pp. 13–104. Bundesanstalt für Straßenwesen, Bergisch Gladbach, 2011. [2] U. Freundt, S. Böning. Verkehrslastmodelle für die Nachrechnung von Straßenbrücken im Bestand. Berichte der Bundesanstalt für Straßenwesen: Brücken- und Ingenieurbau Heft B 82. Bundesanstalt für Straßenwesen, Bergisch Gladbach, 2011. [3] U. Freundt, S. Böning. Einfluss der veränderten Verkehrsführung bei Ertüchtigungsmaßnahmen auf die Bauwerksbeanspruchungen. Berichte der Bundesanstalt für Straßenwesen: Brücken- und Ingenieurbau Heft B 97. Bundesanstalt für Straßenwesen, Bergisch Gladbach, 2013. [4] M. Nowak, O. Fischer, A. Müller. Realitätsnahe Verkehrslastansätze für die Nachrechnung der Gänstorbrücke über die Donau. Beton- und Stahlbetonbau 115(2):91–105, 2019. https://doi.org/10.1002/best.201900060. [5] M. Nowak, O. Fischer. Estimation of truck weights based on strain measurements from tendons of a post-tensioned concrete bridge. In Life-Cycle Civil Engineering: Innovation, Theory and Practice, pp. 1667–1676. CRC Press, 2021. https://doi.org/10.1201/9780429343292-227. [6] C. C. Caprani, E. J. OBrien, G. J. McLachlan. Characteristic traffic load effects from a mixture of loading events on short to medium span bridges. Structural Safety 30(5):394–404, 2008. https://doi.org/10.1016/j.strusafe.2006.11.006. [7] M. Nowak, F. Schmidt, O. Fischer. Comparison of measured and simulated traffic loading based on bwim data from the millau viaduct. In International Probabilistic Workshop IPW 2020. 2020. https://doi.org/10.1007/978-3-030-73616-3_16. [8] C. C. Caprani. Probabilistic Analysis of Highway Bridge Traffic Loading. Ph.D. thesis, National University of Ireland, University College Dublin, School of Architecture, Landscape and Civil Engineering, 2005. [9] A. Žnidarič. Influence of number and quality of weigh-in-motion data on evaluation of load effects on bridges: doctoral thesis. Ph.D. thesis, Univerza v Ljubljani, Fakulteta za gradbeništvo in geodezijo, 2017. [10] M. Nowak, D. Straub, O. Fischer. Statistical extrapolation for extreme traffic load effect estimation on bridges. In 14th International Probabilistic Workshop, pp. 135–153. Springer International Publishing, 2016. https://doi.org/10.1007/978-3-319-47886-9_10. [11] M. Nowak, D. Straub, O. Fischer. Extended extrapolation methods for robust estimates of extreme traffic load effects on bridges. In Proc. 6th International Symposium on Life-Cycle Civil Engineering (IALCCE). 2018. https://doi.org/10.1201/9781315228914. 159 https://doi.org/10.1002/best.201900060 https://doi.org/10.1201/9780429343292-227 https://doi.org/10.1016/j.strusafe.2006.11.006 https://doi.org/10.1007/978-3-030-73616-3_16 https://doi.org/10.1007/978-3-319-47886-9_10 https://doi.org/10.1201/9781315228914 Marcel Nowak, Oliver Fischer Acta Polytechnica CTU Proceedings [12] S. Coles. An Introduction to Statistical Modeling of Extreme Values. Springer London, 2001. https://doi.org/10.1007/978-1-4471-3675-0. [13] C. Cremona. Optimal extrapolation of traffic load effects. Structural Safety 23(1):31–46, 2001. https://doi.org/10.1016/s0167-4730(00)00024-2. [14] L. R. Casas. Evaluation of Dynamic Bridge Excitation byPassing Vehicles Based on Modal Contribution Analysis. Master’s thesis, Technische Universität München, Lehrstuhl für Massivbau, 2020. [15] SICK AG. Profiling-Systeme - TIC102, TIC501, TIC502: Zuverlässige Fahrzeugklassifizierung über mehrere Spuren, 2017. 160 https://doi.org/10.1007/978-1-4471-3675-0 https://doi.org/10.1016/s0167-4730(00)00024-2 Acta Polytechnica CTU Proceedings 36:149–160, 2022 1 Introduction 2 Methods 2.1 Signal Processing 2.2 Mulitple Truck Presence 2.3 Extreme Value Analysis 3 Bridge Structure and Monitoring 3.1 Bridge Structure 29/1 3.2 Alternating Lane Configurations 3.3 Structural Monitoring 3.4 Traffic Monitoring 4 Results from Monitoring Data Analysis 4.1 Multiple Truck Presence on the Bridge 4.2 Load Effects and Extreme Value Behavior 4.3 Discussion of Results 5 Conclusions Acknowledgements References