top of page

Digitalization & Winter Maintenance Optimization (EN)

Vienna, January 2024:




ASFINAG Winter Maintenance Requirements:

„An efficient winter maintenance requires optimization of clearing and salting strategies for all types of weather events with actual effects on road condition, safety, and current costs, taking into account technical and legal boundary conditions“


Scientific Fundamentals Road Winter Maintenance:

Winter road maintenance ensures accessibility and road safety during winter weather events. The skid resistance of the road surface and thus road safety change depending on (dry, wet, frost, snow, and ice), the winter road maintenance, and the resulting road conditions. The key to efficient winter road maintenance is, therefore, to prepare everything in good time to clear the road efficiently at the right time and to optimize the dosage of selected deicers. To address these challenges starting in 2009 the winter maintenance research in Austria was reorganized. A key component of this was the winter road maintenance committee of the Austrian Research Association for Roads, Rail, and Transport (FSV) as a platform consisting of representatives from ASFINAG, the federal states, the Vienna University of Technology, Hoffmann Consult, and the industry. Relevant research projects in Austria have been the Optimization of prewetted salting (2011), Selection criteria for efficient deicing agents (2015) as well as Models for clearing, deicing, and residual salt (2019). In the framework of the FFG-Project WINTERLIFE (2021) on behalf of ÖBB-INFRA, all known gritting agents and additives were also comprehensively analyzed physically and chemically. The development and formalization of a holistic model based on this were carried out in the R&D Projects WinterFIT I-IV (2009 - 2023) for ASFINAG.


Simulation, Validation, and Optimization in Winter Maintenance:

With the development of the winter road maintenance model in the R&D Projects WinterFIT I-IV, the formalized models for precise road condition forecasts and holistic optimization of winter maintenance for the highway operator ASFINAG have been created over the last 10-15 years. The model combines road condition and traffic data, sensor data, and weather forecasts and calculates hourly forecasts for 248 road weather sections and 136 gritting routes in real time. Together with the deployment data of the maintenance vehicles, the change in skid resistance without and with clearing and salting operations is calculated. Over the past year, we have worked on validating the road grip forecasts, forecast and maintenance benchmarking, parameter tuning, and optimizing the deployment strategies. Together with ASFINAG, we are currently working on rolling out the road condition forecasts and treatment recommendations across the entire ASFINAG network in 44 motorway maintenance depots on 2,250 km of motorways and motorways in Austria, including an evaluation of the results (Figure 1).


Figure 1: Road weather sections and depots of ASFINAG on highways in Austria (2023)

Software tool Winter Maintenance:

„The Software tool is based on several research projects with ASFINAG, Federal Regions, the Ministry of Transport, TU Wien, and Hoffmann Consult under the lead of Assoc. Adjunct Prof. Markus Hoffmann covers all aspects from prediction to recommendations and benchmarking“


Overview key functions software tool:

The software tool for ASFINAG's winter road maintenance is based on a structured map interface and network graphs with assigned sensors, road weather sections, and maintenance routes. The existing data (network, sensors, road sections, etc.) and dynamic data (operational journeys, road temperature, events, road conditions) can be displayed under the maps tab. The sensor data tab shows all sensors and their status and allows to edit the sensor network and visualise the measurement data (function/error analysis). Under the forecast models tab, the models for air temperature, pavement temperature, and relative humidity (cluster analysis + neural networks + regression = ensemble) can be trained and analyzed using the measurement data. The hoarfrost forecast model, which is based on comprehensive field measurements with a specially developed sensor in combination with TAWES weather stations, is also included here.

The maintenance models tab primarily contains the physical models such as freezing curves, freezing and melting models, influences of traffic/wind, clearing performance, as well as road conditions and development of skid resistance according to type/amount of precipitation. The cost models for de-icing agents and brine, personnel, and maintenance vehicles as well as the settings and boundary conditions for modelling the clearing and salting strategies can also be found here. Due to the large number of parameters, automatic documentation of changes is also implemented. The road condition tab allows access to the traffic data and condition data for the entire network as an important starting point for the calculations with updates as new data becomes available.

The winter maintenance tab covers the calculation of the maintenance strategies, which can be carried out both automatically and manually without/with consideration of previous operations for each road weather section, route, or depot. The calculation is always carried out hourly for the next +12 hours based on the weather forecast and sensor data, which allows for continuous adjustment and optimization of the deployment strategies. To take account of deviations, particularly in the precipitation forecasts and actual precipitation, it is also possible to forensically calculate the maintenance strategies based on the actual measured values instead of the forecasts. This is particularly important for the validation of the calculations through skid resistance measurements and the investigation of accident events, as deviations occur between forecasts and events as well as between treatment recommendations and actual maintenance.

The last tab benchmarks comprises the standardized analysis of the forecast models, weather events, road conditions, and winter road maintenance. The analysis of the forecast models includes a statistical evaluation of the forecast quality and deviations from measurements. The evaluation of weather events is based on a decision tree for classifying weather events (snow, freezing wet, rime, rain, dry) by days and hours. This functionality can be used to analyze defined periods or winter seasons and create weather profiles for routes, regions, and motorway maintenance depots. The road condition function can be used to analyze the development of the road condition without and with gritting operations (actual/recommended) or the safety gain achieved. Finally, the evaluation of winter road maintenance allows for analysis of the maintenance operations according to weather event type, costs, and effects of selected strategies and thus a comparison of the efforts and results in winter maintenance.

Generation of maintenance strategies:

Based on the objective of achieving safe road conditions with the least necessary costs, all possible treatment strategies in the solution space must be identified, and the resulting costs and effects calculated. In the winter maintenance model, salting strategies with and without plowing are generated for individual weather events based on the combination of circulation time (less than 3h), salt quantity (0 - 40 g/m2), and brine content (FS0 = dry to FS100 = brine only) for the forecast period of +12h. Due to combinatorics, the number of possible strategies increases exponentially with a larger number of possible application rates and brine fractions and, in particular, shorter circulation times (min 3h on highways, min 5h on regional roads in Austria). Figure 2 shows the calculation of the possible combinations according to type (quantity/brine), number of applications, and sequence.

Figure 2: Combination brine types, application rates, and treatment sequences in the model

Figure 3 shows the calculation of the possible start times with the consecutive sequences of deployments with a temporal resolution of 1h (discrete) in the winter maintenance model. Depending on the type, number, and turnaround time, this results in the total number of possible strategies and calculation time. As the strategies in the winter road maintenance model have to be recalculated every hour for the entire ASFINAG network for all 248 road weather sections and 136 maintenance routes, a calculation time of more than 15-20 minutes is unfavorable. Therefore, a reasonable limitation of the possible strategies is necessary using comparative calculations, to avoid excluding favorable to optimal solutions.


Figure 3: Timing and temporal resolution of treatments in the prediction horizon in the model

The results of these comparative calculations with a powerful desktop computer for different numbers, types, and circulation times are shown in Figure 4. If, for example, a circulation time of max. 3h is adhered to by following the winter maintenance guidelines in Austria for highways (RVS 12.04.12) and a maximum of 2 treatment runs with FS30 or FS50 at 10 g/m2 or 20 g/m2 are permitted (e.g. frost), this results in 769 possible strategies. If, on the other hand, the circulation time is shortened to 2 hours, with up to 6 applications in 12 hours being permitted and typical brine proportions (FS0, FS30, FS50, FS70, FS100) and application rates (5, 10, 15, 20 g/m2), this results in over 500 million strategies. Irrespective of a possible optimization of the algorithms, an attempt was therefore made to limit the number of possible strategies to those that are practically relevant (max. 30-50k strategies, computing time < 2-3s) to avoid excessive computing time.

Figure 4: Strategies and computing time for different brine types, rates and turnaround times in the model

Optimization of winter maintenance strategies:

„In contrast to ranking, optimization requires maximization or minimization of an objective function f(xi) under the given boundary conditions. For ASFINAG's winter maintenance, the improvement of road grip (a proxy for safety) was chosen as the objective function“


Costs, Impact & Optimization Functions:

The costs of each winter maintenance strategy are made up of the costs for salt and brine, the vehicle costs (variable/fixed), and the personnel costs (variable/fixed). The positive impact or benefit is calculated via the improvement in skid resistance (frictional resistance) without or with a maintenance strategy, whereby a skid resistance > 0.4 is generally unobjectionable for reasons of driving safety, provided that the permitted speed is maintained. On this basis, different optimization targets - with different optimum strategies - can now be formulated:

  • #1 Minimisation of costs with boundary condition µ ≥ 0.4 (selected)

  • #2 Maximisation of the lowest skid resistance value µmin

  • #3 Minimisation of weighted time with skid resistance reduction µweighted

  • #4 Maximisation of the average skid resistance value ص

  • #5 Maximisation of the benefit/cost ratio


Optimization for hoarfrost events:

The condensation of water vapor to form ice on surfaces with temperatures below the freezing point is called hoarfrost. Hoarfrost events are usually to be expected if there are plus degrees during the day with cooling below freezing point at night. The onset of hoarfrost events is usually between 22:00 and 01:00 in the night with a continuous increase in the amount of ice, reaching a typical maximum of between 50 - 150 g/m2 of ice between 07:00 and 09:00 in the morning. The reason for a low skid resistance at very low amounts of ice compared to snowfall events is that the ice forms on the upper road surface at the typical contact points of the tires. However, the low amount of ice has the advantage that even small amounts of salt at the right time are sufficient to prevent slipperiness. The prerequisite for applying an optimal strategy is precise forecasts of air and road temperature, relative humidity, and hoarfrost events.

The strategy optimization for hoarfrost events is shown as an example of a large frost event in Figure 5. In the example the relative humidity is consistently high, the temperature drops below freezing point from around 9 p.m., and hoarfrost gradually forms, which reduces grip and leads to slipperiness from around 2 to 3 a.m., which lasts until around 10-11 a.m. without winter maintenance. In the case of small and medium events, on the other hand, recognizable slipperiness can be expected between 3 to 5 a.m., especially in exposed areas (e.g. bridges). What are now the most important findings from practical observations and simulations:

  • Any treatment run too early before 7-8 p.m. is not very effective, as a large proportion of the salt is lost due to traffic and wind before the event

  • Preventive application (approx. 10-15g FS50 or FS70) between 10-11 p.m. is usually sufficient for small and medium-sized events (70-80% of cases)

  • For large events (>130 g/m2) as shown here, the intervention time remains the same, but the salting rate should be increased slightly

  • If there are no precise forecasts, the best strategy is a preventive treatment (approx. 10g/m2 FS50 or FS70) around 10-11 p.m. and a second treatment at 3-4 a.m. if necessary

  • If the event is missed and reports of the event are issued around 4.00 a.m. with treatment at 5.00 a.m., even with 20 g/m2 it will take around 1 hour until there is no more ice


Figure 5: Treatment optimization for a large hoarfrost event with safety gains in the model

Optimization for snowfall events:

When analyzing snowfall events, the combination of snowfall amount, event duration, and temperature (LT, FBT) is crucial. In the case of small and medium snowfall events with an amount of 1-3 cm of snow in a treatment cycle and temperatures around the freezing point, it is possible to keep the roads free of snow and ice continuously (Figure 6). However, if temperatures fall below -2°C or the amount of snowfall increases, the thawing effect of salt is no longer sufficient. It should also be noted that it also snows after the maintenance vehicle during the event, causing the road to "close" again. This effect occurs more slowly on roads with heavy traffic, as the waste heat from the traffic can increase the temperature by 1.0 to 1.5 °C, some of the snow falls onto the vehicle roofs, and the thawing effect is somewhat accelerated. Shortening the plowing and salting cycles makes it possible to control some of these events, but cannot prevent slipperiness in all cases, as the salt needs time to thaw and the residual salt is largely cleared away with the fresh snow in the next cycle. What are the most important findings from practical observations and simulations of snowfall events?

  • Any treatment run > 3-4 hours before the event is not effective, but it is important to carry out an initial preventive treatment as close as possible to the start of the event

  • Preventive treatments, especially on dry roads, always require an increased proportion of brine (FS50 or higher) to minimize salt losses

  • For small and medium events and temperatures around freezing point, the road can usually be kept free of snow and ice with correspondingly short circulation times

  • For larger events or lower temperatures, this is generally not possible; excessive salting during the event is both expensive and of little additional benefit

  • In the case of major events, short circulation times are essential to ensure accessibility with typical application rates of 10 g/m2 FS30 or FS50 (formation of separating film)

  • Although this does not prevent the road from "closing" (snow on the surface) again during the event, it does significantly improve the plowing results

  • At the end of the snowfall event, the remaining snow can then be removed by appropriate plowing and salting until sufficient skid resistance is achieved


Figure 6: Treatment optimization for a medium snowfall event with safety gains in the model

User-Platform Weather 2.0:

„The Weather 2.0 platform summarises all information relevant to ASFINAG's winter maintenance, such as weather forecasts, sensor data, webcams, treatment runs, and recommendations from the model, in a web-based form laid out for the personnel“


Software - Architecture Winter Maintenance Tool and user-platform Weather 2.0:

In general, network operators such as ASFINAG are faced with the challenge of either relying on solutions from individual providers or developing their solutions. Solutions "out of the box" are sometimes available for common tasks and often offer good basic functionality. As soon as the tasks become more complex, specific requirements exist or systematic solutions are not yet available, existing solutions must be adapted or new solutions developed. The same applies when data security and availability become more important and cannot be guaranteed with existing solutions. Based on such considerations and the optimal utilization of existing data and applications, ASFINAG has decided to develop its solution for winter road maintenance together with TIETO and Hoffmann Consult. Data from ASFINAG systems, weather forecasts from a service provider (UBIMET/ZAMG), and analyses from third parties are fed into the Weather 2.0 system as a front end for users. In addition to a standardized appearance, even when changing service providers, the main advantage lies in data security and the possibility of ongoing adaptation and further development.

The software tool developed by Hoffmann - Consult for ASFINAG is an example of a third-party analysis being presented in Weather 2.0 with a focus on the end user. The winter maintenance model in the software tool has a modular structure, which means that the individual models and parameters can be validated, adapted, or expanded at any time. The software tool itself runs as a service in the background on the ASFINAG servers, obtains its data from various databases via defined interfaces, and delivers the analyses, forecasts, and treatment recommendations (time, plowing yes/no, salting rate) directly in Wetter 2.0 for the maintenance personnel. This offers the advantage of smooth, independent operation of the systems. In Weather 2.0 as the front end, only the information required for winter maintenance planning such as forecasts (temperatures, humidity, and hoarfrost) with weather warnings and treatment recommendations for the next +12 hours are displayed from the maintenance model. Depending on the desired information, the precipitation forecasts, sensor readings, camera images, or ongoing emergency runs can be shown or hidden in Weather 2.0 (Figure 7). Beyond the scope of the provided information here, the maintenance model offers a variety of additional options for systematic analysis and improvement, which are discussed elsewhere.


Figure 7: Weather 2.0 user interface with a weather forecast, webcams, sensor data, and treatment recommendations

For a more detailed representation of the weather development in Weather 2.0, meteograms of the individual road weather sections with forecasts for +30h, +72h, or +10 days can be selected (Figure 8). In addition to the usual meteorological parameters, the Meteograms also contain weather and ice warnings, whereby the long-term forecasts for the preparation and adjustment of duty rosters, the short-term forecasts for preparedness, and the nowcast for the next +12 hours form the basis for winter maintenance. This approach has proved its worth in practice as it allows us to address both the uncertainties in the forecast of events further in the future and the immediate requirements to be optimally taken into account. Furthermore, there is a comprehensive archive search to be able to retrace past events and operations as comprehensively as possible.


Figure 8: Weather 2.0 user interface with meteogram and event warning in the road weather section

Another key feature of the Weather 2.0 user interface is the optimized view for mobile phones for maintenance personnel. Figure 9 shows some screenshots of selected events with an overview map (precipitation), a brief overview and treatment recommendations of a maintenance depot or route, access to webcams (road conditions), and the data of a sensor location (GMA). In full operation, in the event of weather changes or when traveling, low-threshold access via smartphones to the latest winter service information is essential for maintenance personnel to be able to react quickly. In addition, mobile access also enables convenient control and situation briefings as well as coordination with the coordinators and the personnel, regardless of whether they are in the maintenance depot, on duty, or at home.


Figure 9: Mobile Weather 2.0 with an overview, recommendations, webcams, and sensor data with predictions

Conclusions and Outlook:

Winter road maintenance research in Austria has gradually developed since 2009 based on the systematic investigation of requirements and observations from practice. The winter maintenance model was developed in several research projects with ASFINAG, the federal states, BMK, TU Wien, and Hoffmann Consult under the leadership of Assoc. Adjunct Prof. Hoffmann. The software tool winter maintenance of ASFINAG was developed by Hoffmann Consult and is based on this winter maintenance model. It covers all relevant aspects from sensor data to road condition forecasts, treatment recommendations, and benchmarking. It allows the optimization of the treatment strategies for weather events snow, hoarfrost, and freezing wetness with their effect on the road condition and respective costs. The information relevant to ASFINAG's winter maintenance, such as weather forecasts, sensor data, webcams, current treatment runs, and recommendations, is summarised in the web-based Weather 2.0 portal clearly for the maintenance personnel.

Due to the complexity of the influencing factors in general, models can only provide an incomplete picture of reality. However, if the predictions and results made by the models are sufficiently close to reality, it is possible to compare results and strategies, also validating improvements compared to other models. The software tool developed for implementing the winter road maintenance model is unique in its type and functionality worldwide. Nevertheless, there are uncertainties in several areas, and simplifications have been made in some models in favor of consistent functionality. For future development, existing gaps must therefore be closed, the models improved and the software tool further optimized through the interplay of practical experience and continuous improvement.

However, in the foreseeable future, the final decision regarding the implementation of treatment recommendations in winter maintenance practice will lie with trained and motivated personnel. With the support of modern tools such as the winter maintenance software tool and the Weather 2.0 platform, this should be even more successful in the future. The best possible road conditions in wintry conditions thanks to a modern winter maintenance service is an important prerequisite for a high level of road safety. Overall, however, a reduction in the risk of accidents and a high level of road safety can only be achieved with the cooperation of as many road users as possible and a driving style adapted to the conditions. This is particularly important because the vast majority of accidents are primarily caused by driving behavior and not the condition of the road surface.


Acknowledgements:

Our special thanks go to ASFINAG Service GmbH with the managing directors Mrs. Tamara Christ and Mr. Heimo Meier-Farkas, the project manager Mr. Georg Steyrer, Mr. Szilard Polyanyi, Mrs. Alice Mahr-Saverschel, Mr. Heimo Berghold as well as the many employees of the motorway maintenance depots and the project partners for their advice, feedback, and support with measurements and the application and testing of the solutions developed. We would also like to thank all the members of the Winter Maintenance Committee, headed by Peter Nutz, for their support and encouragement. Special thanks also go to our team Mr. Valentin Donev (since 06/2023 ASFINAG), Mr. Alexander Haberl, and Mr. Markus Hoffmann, whose commitment made these developments and their implementation possible in the first place.


Selected References:

HOFFMANN, M. & NUTZ, P. & BLAB, R. (2011); Forschungsbericht Optimierung der Feuchtsalzstreuung; Forschungsbericht für Ämter der Landesregierungen, ASFINAG, BMVIT; 2011; 213 S.

HOFFMANN, M. & NUTZ, P. & BLAB, R. (2012); Dynamic modeling of winter maintenance strategies and their impact on skid resistance; Vortrag: TRA2012 - Transport Research Arena 2012, Athen; 23.04.2012 - 26.04.2012; Procedia - Social and Behavioral Sciences / Elsevier, Volume 48 (2012), ISSN: 1877-0428; S. 682 – 691; DOI: 10.1016/j.sbspro.2012.06.1046

HOFFMANN, M. & NUTZ, P. & BLAB, R. (2014); Holistic winter maintenance model; XIVth International Winter Road Congress; Andorra la Vella, 4-7 February 2014; PIARC; Beitrag & Vortrag; ISBN 978-99920-0-773-0 Tagungsband

HOFFMANN, M. & DONEV, V. & KANN, A. & HADZIMUSTAFIC, J. & MAIER-FARKAS, H. & STEYRER, G. & BERGHOLD, H. (2016); WINTERFIT II – Echtzeit - Prognosemodellen für den Winterdienst für Fahrbahntemperatur- & Reifentwicklung mit Potenzialanalyse und Wirtschaftlichkeitsuntersuchung für den optimierten Winterdienst; Forschungsbericht für die ASFINAG; Wien; 187 Seiten

HOFFMANN, M. & DONEV, V. & MAIER-FARKAS, H. & STEYRER, G. & BERGHOLD, H. & KANN, A. & HADZIMUSTAFIC, J. (2018); Implementing a dynamic winter maintenance management with real-time measurements and high-resolution weather nowcasts; Peer reviewed paper XVth PIARC International Winter Road Congress, Danzig; 14 S

HOFFMANN, M. & DONEV, V. (2022); Hoarfrost in Winter Maintenance – Measurement, Prediction & Control; eingeladener Fachvortrag International Winter Maintenance Talks 12th – 13th of October; Tulln, Austria

HOFFMANN, M. & DONEV, V. & HABERL, A. (2023); ASFINAG Griffigkeitsvalidierung, Forschungsbericht 104 Seiten, unveröffentlicht

Comments


bottom of page