Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
C H A P T E R 4 Conclusions and Suggested Research 4.1 NUCARS Model Validation The NUCARS model of the PATH PA5 car was updated with the information collected during the vehicle characterization testing. Simulations were performed using the track geometry that was measured on the PATH system in July 2013. Simulation results were compared to test data to determine whether the model accurately predicted vehicle response to track geometry input. Lateral and vertical accelerations collected in the driverâs cab were compared to the simulation results, which correctly predicted the trends and frequency content of the accelerations. There was some deviation in the magnitudes. The discrepancy may be due to the following: ⢠Dampers of the PA5 car were not measured. Manufacture damping information was used in the model. There can be some variation of actual properties and manufacturer data. In the model, it was assumed that the damper was in new condition. The vehicle had been in service, so it is likely that the dampers had some wear. ⢠Representative rail profiles were used in the simulations. Wheel/rail interaction is influenced by the shape of the wheels and rails. The difference between the actual rail profile and the representative profile can contribute to some of the deviations between test and modeling results. To determine the extent of the influence of these parameters, more simulations would have to be run varying the damping and rail profiles. The frequency content and trend were accurately predicted. The magnitude had some deviations, but was well within an acceptable range. The model was used to generate information to build the NNs for PBTG proof of concept. 4.2 On-track Tests A correlation between passenger ride quality and track geometry on the PATH system was identified. Table 11 shows a vehicle response frequency that correlated with a wavelength identified in the track geometry. Further work needs to be done to identify what track anomalies or structures correspond to the wavelengths identified. The influence of entry/exit spiral (or lack of) will need to be investigated. The roll response due to curves can affect ride quality and will need to be studied in Phase 3. Table 11. Frequency Response of Vehicle and Corresponding Wavelength in Track Location Acceleration Direction Vehicle Frequency Response (gâs) Track Geometry Wavelength Pavonia-Newport to Christopher Street Vertical 0.9 Hz 62 ft Lateral 0.76 Hz 74 ft Harrison to Journal Square Vertical 1.25 Hz 33 ft Lateral 0.45 Hz 130 ft 42
4.3 PBTG NN Development NNs were built for both the DART System and the PATH system. ⢠DART NNs predicted vehicle response to track geometry with confidence intervals of 0.1 percent and 0.25 percent based on point-by-point response and segment-based approach, respectively. ⢠The poor predictive performance on DART is due to what is known as overfitting. The track geometry had small deviations; therefore, there were no high dynamic events in the ride quality data. The data used to train the NNs was seen as patternless noise; therefore, the NNs did not recognize a trend in the data. ⢠PATH NNs predicted vehicle response to track geometry with an 81 percent confidence interval using the segment-based approach. The data collected on PATH had significant dynamic events recorded in response to wide range of track geometry deviations. These results show significant improvement compared to DART NNs, which is due to the PATH NNs recognizing trends in the data. The PATH and DART NNs show a potential for using PBTG on some transit systems: ⢠Transit systems with track structures that have significant track geometry deviations and corresponding dynamic response in the vehicle are potential candidates for using NNs to develop PBTG ⢠Transit systems with very few track geometry deviations and relatively low dynamic response are not well suited for using NNs to develop PBTG Phase 2 showed that a correlation between track geometry and vehicle performance can be established. NNs can also be used to predict the vehicle response with a high degree of confidence. More in-depth study will need to be done to establish whether what was predicted by the NNs was accurate and if the track geometry deviation can be mitigated to improve ride quality. 4.4 Recommendation for Future Research: Phase 3 The work from Phase 1 and Phase 2 shows that there is potential for PBTG to be used on transit systems to optimize maintenance and ride quality. Further work will need to be done to determine the viability of a PBTG system and establish guidelines for implementation. The following are tasks that are recommended for Phase 3: ⢠Build a complete set of NNs for PATH using the data and NUCARS model of the PA5 Car. ⢠Use PBTG developed for PATH to predict areas in track that cause adverse ride quality. Once these areas are identified, a track inspection should be done to determine if the prediction is accurate and if it is in fact a deviation that can be maintained (slab track versus ballasted track, etc.). ⢠Correct deviations in the identified track locations. Measure both track geometry and ride quality to determine if significant improvement is achieved. ⢠If significant improvement is achieved, modify the PBTG System that was developed for freight railroads to meet the transit agency needs. ⢠Write guidelines for implementation of PBTG for transit systems. 43
