arxiv: 2605.01914 · v1 · submitted 2026-05-03 · 💻 cs.LG · stat.AP

Recognition: unknown

Deep learning-based pavement performance modeling using multiple distress indicators and road work history

Lu Gao , Zhe Han , Yunshen Chen

Authors on Pith no claims yet

Pith reviewed 2026-05-10 15:38 UTC · model grok-4.3

classification 💻 cs.LG stat.AP

keywords pavement deterioration modelingdeep learningconvolutional neural networkpavement condition predictionmaintenance historyTexas DOT datamachine learning comparisonflexible pavement distress

0 comments

The pith

A CNN trained on 21 pavement distress indicators and maintenance history outperforms standard machine learning models at predicting future pavement conditions.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper sets out to demonstrate that deep neural networks can capture the complex, multi-factor process of pavement deterioration more effectively than conventional approaches. It trains convolutional neural network and long short-term memory models on eighteen years of Texas Department of Transportation data covering more than one hundred thousand pavement sections and twenty-one condition indicators such as cracking, rutting, raveling, and roughness together with work history. The central finding is that the CNN version produces more accurate forecasts of future pavement condition values. A sympathetic reader would care because better forecasts let agencies time preservation and rehabilitation activities to stretch limited budgets and reduce unexpected failures. The work treats pavement performance as a learnable sequence problem driven by observable distresses and interventions rather than as an opaque physical process.

Core claim

Using Texas DOT records spanning eighteen years, the authors show that a convolutional neural network supplied with twenty-one flexible-pavement distress indicators and the corresponding maintenance and rehabilitation history yields higher prediction accuracy for future pavement condition values than standard machine-learning baselines in case-study evaluations.

What carries the argument

Convolutional neural network that processes sequences of pavement condition indicators and work-history events to learn deterioration trajectories.

If this is right

Agencies could shift from reactive to condition-triggered maintenance schedules that extend service life while lowering total spending.
Coordinated preservation activities across road networks become feasible once future condition values are known with higher certainty.
Data-collection priorities can be adjusted toward the most predictive distress indicators once their relative contribution is quantified by the model.
Resource allocation models for pavement management systems can incorporate the CNN forecasts as direct inputs for optimization routines.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

The same indicator-plus-history framing could be tested on bridge or tunnel asset classes where comparable longitudinal inspection records exist.
Adding traffic-load or climate time series as extra input channels might further reduce residual error without changing the core architecture.
If the CNN advantage holds under cross-validation by district, transportation departments could standardize the model as a shared prediction service rather than developing separate regressions for each region.

Load-bearing premise

The twenty-one recorded distress indicators plus maintenance history are enough to represent the dominant drivers of pavement decay, so that unobserved factors do not overwhelm the learned patterns.

What would settle it

On an independent set of pavement sections from a different climate or traffic regime, the CNN model would fail to produce lower prediction error than standard regression or random-forest models when both are given the same twenty-one indicators and work history.

Figures

Figures reproduced from arXiv: 2605.01914 by Lu Gao, Yunshen Chen, Zhe Han.

**Figure 1.** Figure 1: CNN Structure The above equation shows a CNN layer by layer in a forward format. The layers are labeled as boxes. The input X1 goes through the first layer denoted as w1 , which represents a vector of parameters involved in the first layer’s calculation. The output of the first layer is X2 , which is also the input to the second layer. The final output of the network is XK. The last layer, wK, is the loss … view at source ↗

**Figure 2.** Figure 2: LSTM Block Structure in Unfolded Form 2.4 CNN-LSTM Model In this paper, we also propose a CNN-LSTM hybrid framework to combine deep feature extraction and sequence modeling together. With deep features extracted from CNN and then processed by LSTM layers, we are able to achieve much better results, which will be discussed in details in the case study section. As illustrated in [PITH_FULL_IMAGE:figures/ful… view at source ↗

**Figure 3.** Figure 3: CNN-LSTM Structure 2.5 Performance Evaluation In this case study, we used two performance metrics to measure how much forecasts deviate from observations. More specifically, we used accuracy to measure classification models and R2 score to measure regression models. Accuracy is defined as Accuracy = Number of correct predictions Total number of predictions (7) The best possible value for accuracy is 1. R2 … view at source ↗

**Figure 4.** Figure 4: Maintenance & Rehabilitation Work Map In this case study, the historical pavement management data collected between 2000 and 2018 are used to build pavement performance models. Each data point consists of 378 feature variables (21 indicators multiplied by 17 years, 20 M&R dummy variables, and 1 continuous variable indicating the time since last treatment). We developed a performance model for each of the 2… view at source ↗

**Figure 5.** Figure 5: CNN training history: R2 Score or Accuracy v.s. Epoch Figure (6) shows the plots of actual condition value and predicted value in 2018 using the the CNN model results of a randomly selected pavement section. As can be seen in the figure, the models of rut failure, rut depth, logitude cracking, raveling, flusing, ride score, left IRI, and average IRI are able to produce reasonable predictions. For indicator… view at source ↗

**Figure 6.** Figure 6: Actual Data vs. Predicted Value 4 Conclusions In this paper, we used deep learning models including CNN, LSTM, and CNN-LSTM to analyze pavement condition data for performance prediction. The results achieved R2 scores greater than 0.70 for a number of condition indicators. A total of 378 condition attributes including rutting, cracking, raveling, flushing, and roughness were used in the modeling process. M… view at source ↗

read the original abstract

The deterioration of pavement is a complex and dynamic process determined by different factors including material, environment, design, and some other unobserved variables. Accurate predictions of pavement condition can help maximize the use of available resources for pavement management agencies through better coordinated preservation and maintenance activities. This paper uses deep neural networks such as the convolutional neural network (CNN) and the long short-term memory (LSTM) to model the pavement deterioration process. In this paper, pavement condition data and maintenance and rehabilitation history collected by the Texas Department of Transportation over the past 18 years were used. Twenty-one flexible pavement condition indicators, including cracking, rutting, raveling, and roughness, collected from more than 100,000 pavement sections were included in the proposed models. Promising preliminary results were obtained. Case study results show that the proposed CNN model outperforms standard machine learning models in predicting pavement condition values.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

They applied CNN and LSTM to 18 years of Texas pavement data with 21 distress indicators plus maintenance history and report CNN outperforming standard ML, but the evaluation leaves the temporal split unaddressed.

read the letter

The main thing to know is that this paper takes a large real-world Texas DOT dataset—over 100,000 sections tracked for 18 years—and uses CNN and LSTM models on 21 flexible pavement condition indicators together with explicit maintenance and rehabilitation records. They report that the CNN produces better predictions of pavement condition than standard machine learning baselines. That specific combination of many distress metrics and work history in a deep learning setup for pavement modeling looks like a legitimate applied extension rather than something already done in the cited literature. The scale of the data and the practical framing around resource allocation for agencies are the parts that land cleanly. They avoid overclaiming and stick to showing preliminary case-study results on actual agency records. The soft spot is the evaluation. Pavement deterioration is sequential, so the reported CNN advantage only matters if the train/test split respects time order. The abstract gives no information on whether they used random splits, year-based holdouts, or forward chaining. A non-temporal split would let future observations leak into training and make every model's numbers look better without proving anything about modeling the process. They also omit architecture details, exact metrics, statistical tests, and how missing data or the different indicators were handled. This is the sort of paper that would interest transportation researchers or state DOT staff who already have condition databases and want to test modern ML on them. It is not a methods advance, but the data volume and applied focus give it enough substance for a full review. The methods and split protocol need to be checked, but that is exactly what referees are for. I would send it to peer review.

Referee Report

2 major / 1 minor

Summary. The paper proposes using convolutional neural networks (CNN) and long short-term memory (LSTM) models to predict pavement condition deterioration. It uses 21 flexible pavement distress indicators (cracking, rutting, raveling, roughness) plus maintenance and rehabilitation history from Texas DOT data spanning 18 years across more than 100,000 sections. The central empirical claim is that the CNN model outperforms standard machine learning baselines in predicting pavement condition values.

Significance. If the reported outperformance survives a leakage-free temporal validation protocol on the longitudinal data, the work would demonstrate a practical application of deep learning to multi-indicator pavement performance modeling and could support improved preservation scheduling for transportation agencies. The scale of the real-world dataset is a positive feature.

major comments (2)

[Case study results] The case study results section does not describe the train/test split procedure used for the 18-year longitudinal observations. Because pavement sections are observed repeatedly over time, a non-temporal (e.g., random) split allows future observations to leak into training, which would inflate metrics for all models and render the CNN-vs-baseline comparison uninformative.
[Case study results] No information is supplied on the CNN or LSTM architectures, hyperparameter choices, loss functions, error metrics (e.g., MAE, RMSE), handling of missing values, or statistical significance testing of the performance differences. These omissions make the outperformance claim impossible to reproduce or assess.

minor comments (1)

[Abstract] The abstract states that 'promising preliminary results were obtained' without quantifying what 'promising' means or referencing any table or figure.

Simulated Author's Rebuttal

2 responses · 0 unresolved

Thank you for the constructive feedback on our manuscript. We appreciate the emphasis on methodological rigor for longitudinal data and reproducibility. We address each major comment below and will incorporate the necessary revisions to strengthen the paper.

read point-by-point responses

Referee: [Case study results] The case study results section does not describe the train/test split procedure used for the 18-year longitudinal observations. Because pavement sections are observed repeatedly over time, a non-temporal (e.g., random) split allows future observations to leak into training, which would inflate metrics for all models and render the CNN-vs-baseline comparison uninformative.

Authors: We agree that the train/test split must be explicitly described and that a non-temporal split risks leakage in this repeated-measures setting. The original manuscript reported only preliminary results without detailing the split. In revision we will replace the random split with a temporal protocol (training on the first 12 years and testing on the subsequent 6 years, with no overlap), fully document the procedure, and re-evaluate all models under this leakage-free regime so that the CNN-versus-baseline comparison is informative. revision: yes
Referee: [Case study results] No information is supplied on the CNN or LSTM architectures, hyperparameter choices, loss functions, error metrics (e.g., MAE, RMSE), handling of missing values, or statistical significance testing of the performance differences. These omissions make the outperformance claim impossible to reproduce or assess.

Authors: We acknowledge that the lack of implementation details prevents reproduction and assessment. The revised manuscript will add a dedicated subsection that specifies: CNN architecture (layers, filters, kernel sizes, pooling), LSTM architecture (layers, hidden units), hyperparameter selection (grid search with temporal cross-validation), loss function (MSE), evaluation metrics (MAE, RMSE, and R^{2}), missing-value handling (linear interpolation for short gaps, exclusion of sections with >20 % missing), and statistical testing (paired t-tests on per-section errors with Bonferroni correction). These additions will make the outperformance claim fully reproducible and verifiable. revision: yes

Circularity Check

0 steps flagged

No circularity; empirical DL comparison is self-contained

full rationale

The paper applies standard CNN and LSTM architectures to a longitudinal dataset of 21 pavement distress indicators plus maintenance history to predict condition values, then reports that CNN outperforms baseline ML models on case-study metrics. No equations, ansatzes, or uniqueness theorems are invoked; the performance claim is a direct empirical comparison rather than a derivation that reduces to fitted inputs or self-citations by construction. The modeling chain is a conventional supervised-learning pipeline with no self-definitional steps or load-bearing prior-author results.

Axiom & Free-Parameter Ledger

1 free parameters · 2 axioms · 0 invented entities

The modeling approach depends on many unfixed neural network parameters and the domain assumption that observed indicators plus work history capture the main drivers of deterioration; no new physical entities are postulated.

free parameters (1)

neural network hyperparameters and architecture
CNN and LSTM layer counts, filter sizes, learning rates, and sequence lengths are chosen or fitted during training and directly affect the reported performance.

axioms (2)

domain assumption Pavement deterioration can be adequately captured by the 21 observed distress indicators and maintenance history
Invoked by training models on these inputs to predict condition values.
domain assumption Standard machine learning models form a fair and relevant baseline for comparison
Used to support the claim that the CNN model is superior.

pith-pipeline@v0.9.0 · 5448 in / 1361 out tokens · 56413 ms · 2026-05-10T15:38:50.533926+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

61 extracted references · 2 canonical work pages

[1]

Impacts of seasonal and annual weather variations on network-level pavement performance.Infrastructures, 4(2):27, 2019

L Gao, F Hong, and YH Ren. Impacts of seasonal and annual weather variations on network-level pavement performance.Infrastructures, 4(2):27, 2019

2019
[2]

Incorporating uncertainties into determination of flexible pavement preventive maintenance interval.Trans- portmetrica A: Transport Science, 15(1):34–54, 2019

Zhe Han, Juan Diego Porras-Alvarado, Cody Stone, and Zhanmin Zhang. Incorporating uncertainties into determination of flexible pavement preventive maintenance interval.Trans- portmetrica A: Transport Science, 15(1):34–54, 2019

2019
[3]

Stefanos S Politis, Zhanmin Zhang, Zhe Han, John J Hasenbein, and Miguel Arellano. Stochastic analysis of network-level bridge maintenance needs using latin hypercube sampling.ASCE- ASME Journal of Risk and Uncertainty in Engineering Systems, Part A: Civil Engineering, 7 (1):04020049, 2020

2020
[4]

Forensic evaluation of premature failures of texas specific pavement study–1 sections.Journal of performance of constructed facilities, 17(2):67–74, 2003

Dar-Hao Chen, John Bilyeu, Tom Scullion, Deng-Fong Lin, and Fujie Zhou. Forensic evaluation of premature failures of texas specific pavement study–1 sections.Journal of performance of constructed facilities, 17(2):67–74, 2003

2003
[5]

Interface condition influence on prediction of flexible pavement life.Journal of Civil Engineering and Management, 13(1):71–76, 2007

Hassan Ziari and Mohammad Mahdi Khabiri. Interface condition influence on prediction of flexible pavement life.Journal of Civil Engineering and Management, 13(1):71–76, 2007

2007
[6]

Development of pavement performance models for pavement management incorporating treatment type

H Xu, MY Kim, C Sabillon, L Gao, and JA Prozzi. Development of pavement performance models for pavement management incorporating treatment type. 2021

2021
[7]

Missing pavement performance data imputation using graph neural networks.Transportation research record, 2676(12):409–419, 2022

L Gao, K Yu, and P Lu. Missing pavement performance data imputation using graph neural networks.Transportation research record, 2676(12):409–419, 2022

2022
[8]

Pavement missing condition data imputation through collective learning-based graph neural networks.International Conference on Transportation and Development, 2023: 416–423, 2023

K Yu and L Gao. Pavement missing condition data imputation through collective learning-based graph neural networks.International Conference on Transportation and Development, 2023: 416–423, 2023

2023
[9]

A statewide pavement management system.Interfaces, 12(6):5–21, 1982

Kamal Golabi, Ram B Kulkarni, and George B Way. A statewide pavement management system.Interfaces, 12(6):5–21, 1982

1982
[10]

Application of markov process to pavement management systems at network level

Abbas A Butt, MY Shahin, SH Carpenter, and JV Carnahan. Application of markov process to pavement management systems at network level. InProceedings of third international conference on managing pavements, volume 2. Citeseer, 1994

1994
[11]

Number 1180

Yi Jiang, Mitsuru Saito, and Kumares C Sinha.Bridge performance prediction model using the Markov chain. Number 1180. 1988

1988
[12]

Probabilistic behavior of pavements

Kelvin CP Wang, John Zaniewski, and George Way. Probabilistic behavior of pavements. Journal of transportation Engineering, 120(3):358–375, 1994

1994
[13]

Reliability-based processing of markov chains for modeling pavement network deterioration.Transportation research record, 1524(1):203–213, 1996

Ningyuan Li, Wei-Chau Xie, and Ralph Haas. Reliability-based processing of markov chains for modeling pavement network deterioration.Transportation research record, 1524(1):203–213, 1996. 19

1996
[14]

Asemi-markovapproachformodellingassetdeterioration

MBlack, ATBrint, andJRBrailsford. Asemi-markovapproachformodellingassetdeterioration. Journal of the Operational Research Society, 56(11):1241–1249, 2005

2005
[15]

Derivation of transition proba- bility matrices for pavement deterioration modeling.Journal of Transportation Engineering, 132(2):141–161, 2006

José J Ortiz-García, Seósamh B Costello, and Martin S Snaith. Derivation of transition proba- bility matrices for pavement deterioration modeling.Journal of Transportation Engineering, 132(2):141–161, 2006

2006
[16]

Using markov process and method of moments for optimizing management strategies of pavement infrastructure

Lu Gao, Zhanmin Zhang, and Susan Louise Tighe. Using markov process and method of moments for optimizing management strategies of pavement infrastructure. InTransportation Research Board 86th Annual Meeting, number 07-3355, 2007

2007
[17]

A statistical deterioration forecasting method using hidden markov model for infrastructure management.Transportation Research Part B: Methodological, 46(4):544–561, 2012

Kiyoshi Kobayashi, Kiyoyuki Kaito, and Nam Lethanh. A statistical deterioration forecasting method using hidden markov model for infrastructure management.Transportation Research Part B: Methodological, 46(4):544–561, 2012

2012
[18]

Comparison of markov chain and semi-markov models for crack deterioration on flexible pavements.Journal of Infrastructure Systems, 19(2):186–195, 2013

Omar Thomas and John Sobanjo. Comparison of markov chain and semi-markov models for crack deterioration on flexible pavements.Journal of Infrastructure Systems, 19(2):186–195, 2013

2013
[19]

Transition probability matrices for flexible pavement deterioration models with half-year cycle time.International Journal of Civil Engineering, 16(9):1045–1056, 2018

Heriberto Pérez-Acebo, Sergiu Bejan, and Hernán Gonzalo-Orden. Transition probability matrices for flexible pavement deterioration models with half-year cycle time.International Journal of Civil Engineering, 16(9):1045–1056, 2018

2018
[20]

Estimating spatial dependence associated with deteriora- tion process of road network.Transportation Research Board 95th Annual MeetingTransportation Research Board, 2016

S Jahanbakhsh, L Gao, and Z Zhang. Estimating spatial dependence associated with deteriora- tion process of road network.Transportation Research Board 95th Annual MeetingTransportation Research Board, 2016

2016
[21]

Estimation of infrastructure distress initiation and progression models.Journal of Infrastructure Systems, 1(3):146–150, 1995

Samer Madanat, Srinivas Bulusu, and Amr Mahmoud. Estimation of infrastructure distress initiation and progression models.Journal of Infrastructure Systems, 1(3):146–150, 1995

1995
[22]

Samer Madanat and Hee Shin. Development of distress progression models using panel data sets of in-service pavements.Transportation Research Record: Journal of the Transportation Research Board, (1643):20–24, 1998

1998
[23]

Using duration models to analyze experimental pavement failure data.Transportation Research Record: Journal of the Transportation Research Board, (1699):87–94, 2000

Jorge Prozzi and Samer Madanat. Using duration models to analyze experimental pavement failure data.Transportation Research Record: Journal of the Transportation Research Board, (1699):87–94, 2000

2000
[24]

Development of pavement performance models by combining experimental and field data.Journal of Infrastructure Systems, 10(1):9–22, 2004

JA Prozzi and SM Madanat. Development of pavement performance models by combining experimental and field data.Journal of Infrastructure Systems, 10(1):9–22, 2004

2004
[25]

PhD thesis, 2005

Zheng Li.A probabilistic and adaptive approach to modeling performance of pavement infras- tructure. PhD thesis, 2005

2005
[26]

Survival analysis of fatigue cracking for flexible pavements based on long-term pavement performance data.Journal of transportation engineering, 131(8):608–616, 2005

Yuhong Wang, Kamyar C Mahboub, and Donn E Hancher. Survival analysis of fatigue cracking for flexible pavements based on long-term pavement performance data.Journal of transportation engineering, 131(8):608–616, 2005

2005
[27]

Applying method of moments to model reliability of pavements infrastructure.Journal of Transportation Engineering, 132(5):416–424, 2006

Zhanmin Zhang and Ivan Damnjanović. Applying method of moments to model reliability of pavements infrastructure.Journal of Transportation Engineering, 132(5):416–424, 2006

2006
[28]

Dynamic panel data model for predicting performance of asphalt concrete overlay.Journal of Transportation Engineering, 134 (2):86–92, 2008

Yuhong Wang, Kamyar C Mahboub, and Donn E Hancher. Dynamic panel data model for predicting performance of asphalt concrete overlay.Journal of Transportation Engineering, 134 (2):86–92, 2008. 20

2008
[29]

Empirical comparison of statistical pavement performance models.Journal of Infrastructure Systems, 14(2):138–149, 2008

Chih-Yuan Chu and Pablo L Durango-Cohen. Empirical comparison of statistical pavement performance models.Journal of Infrastructure Systems, 14(2):138–149, 2008

2008
[30]

Roughness model accounting for heterogeneity based on in-service pavement performance data.Journal of Transportation Engineering, 136(3):205–213, 2010

Feng Hong and Jorge A Prozzi. Roughness model accounting for heterogeneity based on in-service pavement performance data.Journal of Transportation Engineering, 136(3):205–213, 2010

2010
[31]

Performance modeling of infrastructure condition data with maintenance intervention.Transportation research record, 2225(1):109–116, 2011

L Gao, JP Aguiar-Moya, and Z Zhang. Performance modeling of infrastructure condition data with maintenance intervention.Transportation research record, 2225(1):109–116, 2011

2011
[32]

Pavement deterioration model incorporating unobserved heterogeneity for optimal life-cycle rehabilitation policy.Journal of Infrastructure Systems, 21 (1):04014027, 2015

Feng Hong and Jorge A Prozzi. Pavement deterioration model incorporating unobserved heterogeneity for optimal life-cycle rehabilitation policy.Journal of Infrastructure Systems, 21 (1):04014027, 2015

2015
[33]

A nested modelling approach to infrastructure performance characterisation.International Journal of Pavement Engineering, pages 1–7, 2016

Zhanmin Zhang and Lu Gao. A nested modelling approach to infrastructure performance characterisation.International Journal of Pavement Engineering, pages 1–7, 2016

2016
[34]

Develop- ment of network-level pavement deterioration curves using the linear empirical bayes approach

Antonio Pantuso, Gerardo W Flintsch, Samer W Katicha, and Giuseppe Loprencipe. Develop- ment of network-level pavement deterioration curves using the linear empirical bayes approach. International Journal of Pavement Engineering, pages 1–14, 2019

2019
[35]

Predicting roughness progression in flexible pavements using artificial neural networks

Nii O Attoh-Okine. Predicting roughness progression in flexible pavements using artificial neural networks. InTransportation Research Board Conference Proceedings, volume 1, 1994

1994
[36]

S Owusu-Ababio. Application of neural networks to modeling thick asphalt pavement per- formance.Artificial Intelligence and Mathematical Methods in Pavement and Geomechanical Systems, pages 23–30, 1998

1998
[37]

Nii O Attoh-Okine. Analysis of learning rate and momentum term in backpropagation neural network algorithm trained to predict pavement performance.Advances in Engineering Software, 30(4):291–302, 1999

1999
[38]

Pavement roughness modeling using back-propagation neural networks.Computer-Aided Civil and Infrastructure Engineering, 19 (4):295–303, 2004

Jae-Ho Choi, Teresa M Adams, and Hussain U Bahia. Pavement roughness modeling using back-propagation neural networks.Computer-Aided Civil and Infrastructure Engineering, 19 (4):295–303, 2004

2004
[39]

Prediction of onset of corrosion in concrete bridge decks using neural networks and case-based reasoning.Computer-Aided Civil and Infrastructure Engineering, 20 (2):108–117, 2005

G Morcous and Z Lounis. Prediction of onset of corrosion in concrete bridge decks using neural networks and case-based reasoning.Computer-Aided Civil and Infrastructure Engineering, 20 (2):108–117, 2005

2005
[40]

Automated diagnosis of sewer pipe defects based on machine learning approaches.Expert Systems with Applications, 35(3):1327–1337, 2008

Ming-Der Yang and Tung-Ching Su. Automated diagnosis of sewer pipe defects based on machine learning approaches.Expert Systems with Applications, 35(3):1327–1337, 2008

2008
[41]

Artificial neural network (ann) based pavement deterioration models for low volume roads in india.International Journal of Pavement Research & Technology, 5(2), 2012

Dattatraya Tukaram Thube. Artificial neural network (ann) based pavement deterioration models for low volume roads in india.International Journal of Pavement Research & Technology, 5(2), 2012

2012
[42]

Two-stage support vector classifier and recurrent neural network predictor for pavement performance modeling.Journal of Infrastructure Systems, 19(3):266–274, 2013

Nader Tabatabaee, Mojtaba Ziyadi, and Yousef Shafahi. Two-stage support vector classifier and recurrent neural network predictor for pavement performance modeling.Journal of Infrastructure Systems, 19(3):266–274, 2013

2013
[43]

Jakub Gajewski and Tomasz Sadowski. Sensitivity analysis of crack propagation in pavement bituminous layered structures using a hybrid system integrating artificial neural networks and finite element method.Computational Materials Science, 82:114–117, 2014. 21

2014
[44]

Performance models for hot mix asphalt pavements in urban roads.Construction and Building Materials, 116:281–288, 2016

Ufuk Kırbaş and Mustafa Karaşahin. Performance models for hot mix asphalt pavements in urban roads.Construction and Building Materials, 116:281–288, 2016

2016
[45]

A gradient boosting approach to understanding airport runway and taxiway pavement deterioration.International Journal of Pavement Engineering, pages 1–15, 2020

Limon Barua, Bo Zou, Mohamadhossein Noruzoliaee, and Sybil Derrible. A gradient boosting approach to understanding airport runway and taxiway pavement deterioration.International Journal of Pavement Engineering, pages 1–15, 2020

2020
[46]

Yongjun Lee, Jongwan Sun, and Minjae Lee. Development of deep learning based deterioration prediction model for the maintenance planning of highway pavement.Korean Journal of Construction Engineering and Management, 20(6):34–43, 2019

2019
[47]

Development of the road pavement deterioration model based on the deep learning method.Electronics, 9(1):3, 2020

Seunghyun Choi and Myungsik Do. Development of the road pavement deterioration model based on the deep learning method.Electronics, 9(1):3, 2020

2020
[48]

Deep learning.nature, 521(7553):436–444, 2015

Yann LeCun, Yoshua Bengio, and Geoffrey Hinton. Deep learning.nature, 521(7553):436–444, 2015

2015
[49]

Handwritten digit recognition with a back- propagation network

Yann LeCun, Bernhard E Boser, John S Denker, Donnie Henderson, Richard E Howard, Wayne E Hubbard, and Lawrence D Jackel. Handwritten digit recognition with a back- propagation network. InAdvances in neural information processing systems, pages 396–404, 1990

1990
[50]

Neural network model for a mechanism of pattern recognition unaffected by shift in position-neocognitron.IEICE Technical Report, A, 62(10):658–665, 1979

Kunihiko Fukushima. Neural network model for a mechanism of pattern recognition unaffected by shift in position-neocognitron.IEICE Technical Report, A, 62(10):658–665, 1979

1979
[51]

Gradient-based learning applied to document recognition.Proceedings of the IEEE, 86(11):2278–2324, 1998

Yann LeCun, Léon Bottou, Yoshua Bengio, and Patrick Haffner. Gradient-based learning applied to document recognition.Proceedings of the IEEE, 86(11):2278–2324, 1998

1998
[52]

Imagenet classification with deep convolutional neural networks

Alex Krizhevsky, Ilya Sutskever, and Geoffrey E Hinton. Imagenet classification with deep convolutional neural networks. InAdvances in neural information processing systems, pages 1097–1105, 2012

2012
[53]

Multi-scale convolutional neural networks for time series classification.arXiv preprint arXiv:1603.06995, 2016

Zhicheng Cui, Wenlin Chen, and Yixin Chen. Multi-scale convolutional neural networks for time series classification.arXiv preprint arXiv:1603.06995, 2016

work page arXiv 2016
[54]

Time series classification from scratch with deep neural networks: A strong baseline

Zhiguang Wang, Weizhong Yan, and Tim Oates. Time series classification from scratch with deep neural networks: A strong baseline. In2017 International joint conference on neural networks (IJCNN), pages 1578–1585. IEEE, 2017

2017
[55]

A primer on neural network models for natural language processing.Journal of Artificial Intelligence Research, 57:345–420, 2016

Yoav Goldberg. A primer on neural network models for natural language processing.Journal of Artificial Intelligence Research, 57:345–420, 2016

2016
[56]

Long short-term memory.Neural computation, 9(8): 1735–1780, 1997

Sepp Hochreiter and Jürgen Schmidhuber. Long short-term memory.Neural computation, 9(8): 1735–1780, 1997

1997
[57]

A new boosting algorithm for improved time-series forecasting with recurrent neural networks.Information Fusion, 9(1): 41–55, 2008

Mohammad Assaad, Romuald Boné, and Hubert Cardot. A new boosting algorithm for improved time-series forecasting with recurrent neural networks.Information Fusion, 9(1): 41–55, 2008

2008
[58]

Nonlinear systems identifica- tion using deep dynamic neural networks.arXiv preprint arXiv:1610.01439, 2016

Olalekan Ogunmolu, Xuejun Gu, Steve Jiang, and Nicholas Gans. Nonlinear systems identifica- tion using deep dynamic neural networks.arXiv preprint arXiv:1610.01439, 2016

work page arXiv 2016
[59]

Pavement analyst data dictionary condition summary

TxDOT. Pavement analyst data dictionary condition summary. Technical report, Texas Department of Transportation, 2018. 22

2018
[60]

Detection of pavement maintenance treatments using deep-learning network.Transportation Research Record, 2675(9):1434–1443, 2021

L Gao, Y Yu, Y Hao Ren, and P Lu. Detection of pavement maintenance treatments using deep-learning network.Transportation Research Record, 2675(9):1434–1443, 2021

2021
[61]

Evaluation and development of pavement scores, performance models and needs estimates for the txdot pavement management information system

Nasir Gharaibeh, Tom Freeman, Siamak Saliminejad, Andrew Wimsatt, Carlos Chang-Albitres, Soheil Nazarian, Imad Abdallah, Jose Weissmann, Angela Jannini Weissmann, Athanassios T Papagiannakis, et al. Evaluation and development of pavement scores, performance models and needs estimates for the txdot pavement management information system. Technical report, ...

2012