Manosalvas-Paredes, Mario
(2021)
Full–scale validation and implementation guidelines of combined piezoelectric sensing and big data analytics for long-term pavement performance monitoring.
PhD thesis, University of Nottingham.
Abstract
It is generally known that pavement structures, regardless of their main structural layer, start suffering progressive damage which could be either structural, functional or a combination of both from the day they are opened to traffic loading. This progressive damage may be exacerbated by different factors such as material properties, temperature, moisture, loading, tyre pressure, axle configuration, travel speed and so on. It is also known that road infrastructure very seldom becomes redundant in a country's economy and that for some, especially developing countries, road infrastructure is considered as one of their main assets and drivers of growth. Consequently, pavement deterioration has been widely studied at both laboratory and full-scale levels but despite of the multiple efforts made by researchers around the globe, there has not been a real application nor technology that could be easily implemented in the field.
Unequivocally, it could be said that pavement engineering over the years has evolved from art to science. At the beginning, empirical relationships developed from the AASHTO Road Test, in the late 1950's, were used to predict pavement performance as a function of the layer thicknesses, layer coefficients, subgrade resilient modulus and the number of load repetitions of an standardized axle. Nowadays, those predictions are outdated and even though non-destructive testing (NDT) methods are available to compensate for those differences no-one should deny the associated high-cost and labour-intensity that NDT has. This last point has opened the door for new technologies to arise such as the one discussed here.
The objective of this research is to present a novel approach for long-term monitoring of pavement structures through the combination of piezoelectric transducers and novel condition-based interpretation methods. To that end, a laboratory campaign at the Nottingham Transportation Engineering Centre was conducted where ninety-six bending tests were made varying the induced microstrain, the loading frequency and the temperature of the chamber. Laboratory results have shown a positive and strong linear relation between the measured voltage and strain with a coefficient of determination (R^2) higher than 0.95.
The piezoelectric transducer has then been validated for the detection of bottom-up fatigue cracking through full-scale testing. Sensors were installed at the bottom of a high modulus asphalt mix (EME2) of 102 millimeters and loaded until failure. The condition-based approach, used in this research, does not rely on strain measurements and bypasses the need for any structural or finite element models. Results from the fatigue carrousel indicate a successful validation of the piezoelectric sensor in detecting damage initiation and progression in asphalt concrete pavements. Moreover, the results shown from the data analysis method, demonstrate a very early detection capability compared to classical inspection methods. This could represent a huge potential for improved pavement monitoring.
Finally, this research has tried to replicate different damage stages which are commonly seen in pavements during their design life through the combination of the reduction in the asphalt modulus and the damage extent. A total of 4,270 combinations have been generated to understand and most important to compare how distinct configurations of piezoelectric sensor would behave in a real field installation. Results have shown that 10 sensors, randomly distributed over the wheel-paths, are not sufficient to provide trustworthy information about the condition of the pavement structure and therefore should not be used. Configurations above 50 sensors started showing more promising results; nonetheless, this research could not determine which configuration outweighs the others and it is advised that machine learning techniques are brought into future analysis.
Actions (Archive Staff Only)
 |
Edit View |
Tools
Tools
![[img]](/style/images/fileicons/application_pdf.png)