2.1 What does Data Science actually mean?

Data Science is a multi-disciplinary field whose aim is to extract knowledge from data. In some sense, it can be perceived as a modernized version of Statistics which follows the same aim. Quoting Nate Silver at the Joint Statistical Meetings of 2013, “Data Scientist is just a sexed up word for Statistician.” According to Diggle (Reference Diggle2015), Statistics is the Data Science of our modern times. In the same vein, Ley and Bordas (Reference Ley and Bordas2018) claim that Data Science is actually Statistics 2.0. What may set Data Science apart from the more “classical” Statistics (probability model building, data description, inference, and prediction) is the incorporation of machine learning (supervised, unsupervised, and reinforcement learning) and computer science besides these classical statistics. It thus consists of both top-down (a model is built that represents the information contained in the data) and bottom-up (starting point are the data and the model is then built by a computer) approaches.Footnote ¹ It is this crafty combination that makes Data Science so attractive for theoreticians and practitioners, and has caused the hype around this “recent” field. It is definitely the right discipline at the right moment bearing the right name, and therefore rightfully at the core of Data-Centric Engineering.

2.2 (Big) Data alone are not worth much!

One of the biggest misconceptions related to the Big Data phenomenon is that the sole fact of having massive data sets will solve all problems.Footnote ² Nothing is less true, because data do not speak by themselves (though this sentence is often incorrectly used). Various steps are required to obtain the best out of the data amounts: a smart data collection strategy (smart devices and wearables have opened new doors here), a thorough data cleaning (checking for outliers, handling missing data, etc.), a flawless data analysis (via statistical or machine learning methods), and ultimately a correct interpretation of the results. Besides the development of new methods to analyse data, the data analysis as such is also part of Data Science.

3.1 The assets of the application of Data Science in Structural Engineering

Over a long time, Data Science and Structural Engineering were considered as two completely separated disciplines and developed independently. In the coming years, a clear change of mentality is expected to take place in Structural Engineering. The present section summarizes the benefits and challenges of the assessment of data in Structural Engineering, more specifically for the monitoring of bridge structures. A case-study about digital inspection of an Italian bridge will highlight the advantages of Data-Centric Engineering for safety aspects of bridges in general.

Regular inspections of structures, for instance of bridges, are recognized as an appropriate tool to ensure their safety throughout their lifetime. Inspections of sensitive structural details are carried out by engineers and technicians who access the structures by cranes and/or enter the structures by special entrances to control the details from inside. Inspections consist in visual inspection and measurements with specific techniques like X-rays and ultrasonic to detect cracks inside the parent steel material and welds. Depending on the observations, engineers decide for specific maintenance actions on the structure. Maintenance actions are for example the reparation of cracked welds in steel bridges, the replacement of cracked concrete elements and corroded reinforcements. The regular inspections and maintenance costs usually represent the dominant part of the user costs of the structures. An economical structural solution stands out by low inspection and maintenance costs. Many details are in areas of the structure which are difficult to access and consequently the inspection is expensive, time consuming, and error prone. Therefore, the trend in civil engineering will develop toward digital inspections of structures, whereupon bridges are predestinated for digital inspection.

For digital inspection, sensors are applied on sensitive details of the bridge and continuously measure their loading (e.g., stresses, strains, etc.). The data of the loading are stored on computers and regularly assessed by civil engineers. The engineers determine the damage of the details and identify the required maintenance actions. The latest development in computational engineering allows for the application of sensors which continuously measure the loading of the details during the whole lifetime of the bridge. The lifetime of a bridge is defined with 100 years and one can easily imagine the amount of data which is collected over this timespan. The advantage of the method consists in the remote inspection of the bridge. The engineers no longer need to physically access the details of the bridge, which they can inspect safely from the office. This reduces the inspection costs and the measurements can be more precise.

3.2 The challenge of an interdisciplinary approach

The challenge consists in the correct and effective assessment of the Big Data to conclude on the safety of the bridge and to launch maintenance actions only if required. Since civil engineers are not experts in analyzing such data, the method is often ineffective. A general strategy to assess the measured data does currently not exist in civil engineering and the engineers apply a case-by-case approach. This leads to a situation where not all the information is extracted from the data. The risk is high that the damage of the bridge is wrongly calculated from the data (too high or too low). As a result, the safety of the bridge is wrongly assessed, or needless and uneconomic maintenance actions are undertaken. Another challenge is the hardware to store the data. Many small engineering offices are currently inadequately equipped to evaluate the data and to apply for digital inspections. A standard gear for the evaluation of data for small and medium design offices needs to be defined.

3.3 Case study: the Marchetti bridge

Let us now become more concrete and illustrate our statement via a case-study for digital inspection, the so-called Active Monitoring. This is a perfect example of Data-Centric Engineering as an interdisciplinary approach between Data Science and Structural Engineering and is considered to be a driver for the future developments of digitalization in engineering.

The Marchetti bridge is a steel arch high way bridge between Torino and Milano, Italy. The structural layout is a single arch connected to the inferior way with a span of 250 m. The deck consists of a chain beam and transverse cantilevers with a spacing of 8 m. The bridge carries two carriageways of a total width of 17.55 m per way (Chiaia et al., Reference Chiaia, Ventura, Zannini Quirini and Marasco2019). Figures 1 and 2 show the superstructure of the bridge.

Figure 1. Main beam of the Marchetti bridge (Chiaia et al., Reference Chiaia, Ventura, Zannini Quirini and Marasco2019) Springer Nature Switzerland AG 2020. This figure has been reproduced with the permission of the copyright holder, and is not included in the Creative Commons licence applied to this article. For other re-use, please contact the copyright holder.

Figure 2. Arch of the Marchetti bridge (Chiaia et al., Reference Chiaia, Ventura, Zannini Quirini and Marasco2019) Springer Nature Switzerland AG 2020. This figure has been reproduced with the permission of the copyright holder, and is not included in the Creative Commons licence applied to this article. For other re-use, please contact the copyright holder.

Active Monitoring has been developed by G. Ventura for ARCOS Engineering. It is a new paradigm in bridge monitoring. It can be considered as a novel approach to Structural Health Monitoring (SHM). Several examples of SHM can be found in the literature, see Chiaia et al. (Reference Chiaia, Ventura, Zannini Quirini and Marasco2019), Marrongelli et al. (Reference Marrongelli, Gentile, Saisi, Arêde and Costa2019), as well as Cunha et al. (Reference Cunha, Caetano, Magalhães and Moutinho2013), Koo et al. (Reference Koo, Brownjohn, List and Cole2013), and Magalhães et al. (Reference Magalhães, Cunha and Caetano2009). Active Monitoring consists in a real-time assessment of the measured data through a computational model. Sensors are placed on specific details and locations of the bridge. They measure loads, displacements, and other important parameters. The acquired data are automatically elaborated by an algorithm and sent to a finite-element model of the bridge for the real-time evaluation of its health status. A certain deviation between the measured values and the values from the numerical model is an indication of a safety degradation of the bridge. Figure 3 summarizes the philosophy of Active Monitoring.

Figure 3. Methodology of Active Monitoring (Chiaia et al., Reference Chiaia, Ventura, Zannini Quirini and Marasco2019). Springer Nature Switzerland AG 2020. This figure has been reproduced with the permission of the copyright holder, and is not included in the Creative Commons licence applied to this article. For other re-use, please contact the copyright holder.

For the Marchetti bridge, different sensors are applied on 12 monitoring points. The range of sensors covers high resolution servoinclinometers, temperature sensors, humidity sensors, triaxial accelerometers, wind pressure transducer, strain gages, and load cells. The locations are depicted in Figure 4.

Figure 4. Position of the sensors on the Marchetti bridge (Chiaia et al., Reference Chiaia, Ventura, Zannini Quirini and Marasco2019). Springer Nature Switzerland AG 2020. This figure has been reproduced with the permission of the copyright holder, and is not included in the Creative Commons licence applied to this article. For other re-use, please contact the copyright holder.

For the processing of the data, a custom software has been developed by ARCOS Engineering. The software acquires the data, calls the finite element engine and processes back the data, automatically compares the measured and calculated data and generates warning if the difference exceeds a threshold value. The finite element software runs on a high-end Microsoft windows workstation. Figure 5 shows an example of the output of the numerical model and the comparison to the measured data from the sensors installed on the bridge.

Figure 5. Semaphore in the Active Monitoring System (Chiaia et al., Reference Chiaia, Ventura, Zannini Quirini and Marasco2019). Springer Nature Switzerland AG 2020. This figure has been reproduced with the permission of the copyright holder, and is not included in the Creative Commons licence applied to this article. For other re-use, please contact the copyright holder.

Further information on Active Monitoring, which was first applied on the Marchetti bridge in 2015, can be found in Chiaia et al. (Reference Chiaia, Ventura, Zannini Quirini and Marasco2019).

The example of the Marchetti bridge in Italy highlights the benefits of Data-Centric Engineering. The automatic assessment and comparison of the data eliminates the “human factor” and allows for a quick, economical and reliable monitoring of bridges. It should be noted that Active Monitoring can still not replace the analytical capacity of an expert in structural data processing and interpretation (Chiaia et al., Reference Chiaia, Ventura, Zannini Quirini and Marasco2019). Moreover, the data to be measured on the bridge should be clearly defined and limited to the needs of the safety assessment. Too many parasitic data would exceed the capacity of the data requirement system and would falsify the results of the finite element software. Finally, the data itself is not worth anything without a proper processing and evaluation of it. The data for themselves are not important in Structural Engineering but rather the information one can gain from them; see section “(Big) Data alone are not worth much!”