Decision making flow

Case studies

Click one of the links below to view a case study:


The IM-SAFE project proposes a framework for the data-informed safety assessment highly interrelated with the life cycle management of the transport infrastructure. The proposed flow describes the different stages of the assessment of new and existing structures, focusing on various levels (network, system and component) and taking into account the available data during the lifespan and maintenance process of the structure. It allows to consider both input information, data collection and storage necessities, as well as the possibility to have a BIM and/or Digital Twin model to support the analysis and the infrastructures management process. The aims of the proposed flow are: i) implement in practice the data-informed safety assessment methods; ii) rationalize the decision-making process with regards to interventions; iii) provide an efficient asset management tool as further guidance on the use of data.


Figure 1 – Decision-making flow for optimized maintenance and interventions.

The assets have been subdivided and differentiated in 3 relevant classes:

  1. new structures;
  2. existing structures;
  3. existing structures after intervention.

The first step of the framework consists of a simplified assessment of the structures that are part of the same road network. This analysis is based on the review of the available relevant information (such as design documents, as-bult information incl. Birth Certificate, inspection documents, condition surveys, structural investigations, material testing outcomes, data gathered from continuous monitoring re-deign documents, Re-birth Certificate and intervention reports, etc.) and can be supported by a vulnerability analysis, aiming to identify the critical elements of each asset, taking into account that some vulnerable areas can develop only with the aging of the structure or also the occurrence of degradation processes and cannot be assessed due to their inaccessibility. The simplified analysis is required as a prerequisite to perform a classification and prioritization of the assets that are part of the same road network based on a risk analysis, which can be performed either at the network level or at the system/component level. Risk analyses aims to define the priority list for optimized maintenance and interventions policies. Based on its outcomes, structures can be assigned to one of the following classes: Low Priority, Medium Priority or High Priority. Prioritization of intervention is not required in case a single structure is being analysed.

Each of the classes defined above is characterized by a specific prioritization plan of interventions, based on the results of the preliminary assessment. “Low priority” structures undergo routine maintenance inspection plans and could change class based on the outcomes of the ordinary maintenance or eventually light monitoring. The specific maintenance strategy can be selected based on the risk analysis and considerations. A detailed investigation plan as well as a detailed assessment might be considered in case of detected or suspected anomalies. In this case, further condition survey plans can include special inspections, testing and monitoring campaigns. Structures in the “Medium priority” and “High priority” classes may be subjected to detailed assessment.

Based on the gathered information about the condition of the structure, decisions regarding the use of inspections, monitoring and testing are evaluated. In particular, with respect to the reliability levels differentiation proposed for the assessment of existing structures, actions may vary based on the target levels β_0, which is the level below which the existing structure is considered unreliable and should be upgraded, and β_up, which is the level indicating an optimum upgrade strategy while upgrading of existing structures. If the reliability is lower than the minimum accepted β_0, the outcome of detailed assessment can result directly in extraordinary maintenance or interventions. In case operational interventions (e.g. traffic limitation) are needed before any structural intervention is performed, monitoring strategies might be applied as a step to prevent undesired events before the structural upgrade. If the reliability is between β_0 and β_up , the evaluation of the most suitable monitoring strategy is suggested, based on an optimization approach to either select increased visual inspection schedules, further additional testing and/or application of periodic, frequent, or continuous monitoring. The identification of the monitoring strategy can be made based on economical and sustainability decision making processes.

In case continuous SHM systems are used, the approach to monitoring differs depending on the complexity of the structural scheme following an increasing level of approximation in the analysis approach. Monitoring of the structural performance of key structural components is suggested for simple structural schemes, whilst non-linear FE modelling coupled with model updating approaches is suggested for more complex structures. Complex structural schemes, as such, require a higher complexity of the analysis and performance monitoring, which implies numerical modelling, model updating processes and deeper analysis of the presence and localisation of damages. Regardless of the complexity of the structural system, the gathered data should feed appropriate key-performance indicators and should serve as input for structural diagnostics procedures, including the definition of thresholds, essential step to promptly identify potential anomalies or identify a sudden change of the behaviour of the structures.

Structural diagnostics procedures have to be supported by a thorough data processing phase, which allows to identify potential anomalies (anomaly detection) with respect to the standard structural behaviour expected based on both the numerical models and the past performance (statistical evaluation of the data series over time). The presence and quantification of the extent of damage in a system based on the information extracted from the measured system response may be performed using ad-hoc damage detection algorithms. In more recent years, the big-data processing is increasingly supported by machine learning and AI routines, both supervised and unsupervised. The detection of anomalies may be considered as a trigger for decisions regarding the end of service life of the structure: in case of very extreme events or exceedance of the alarm levels, the end of the service life of the structures could potentially be reached, whilst for less severe thresholds levels exceedance, the monitoring system would be a trigger for further investigation levels or, eventually, for another detailed structural assessment to evaluate if a structural intervention or an upgrade is needed.