Maintenance is considered a strong contributor in every manufacturing firm’s objectives as it holds key role in the cost minimisation, production efficiency and quality, profitability, competitiveness and customer satisfaction. For ship-owner companies, maintenance safeguards the smooth, safe and uninterrupted operation of all of the ship’s components, modules and systems. Moreover, it aims at mitigating the risk of failure of any proper operation of any of these ship’s parts that could result to significant damages and propagating effects to various interconnected systems.
The BigDataOcean platform aims at improving the business situation of FOINIKAS by providing the means for the design and implementation of efficient maintenance strategies that will be based upon the concept of Condition Based Maintenance (CBM). CBM is maintenance strategy that is both proactive and predictive. In CBM for every asset the actual condition is continuously assessed by monitoring the conditions of the components of the asset through equipment-installed sensors and their comparison with expected performance or average values. The scope of CBM is to provide a dynamic preventive schedule enriched with dynamic pattern recognition in the monitoring data and it is influenced by the recent development of new predictive maintenance approaches such as the prediction of Remaining Useful Life (RUL) or the Remaining Life Distribution (RLD).
These maintenance strategies will facilitate the reduction of the efforts and resources spent in unplanned repairs through the provision of estimates of probabilities regarding potential breakdowns, as well as through the performance of maintenance very close to breakdown time, maximising RUL, efficiency and profit. Hence, in the Fault Prediction and Predictive Maintenance scenario we are focusing on providing useful insights related to the unplanned repairs and equipment changes that arise outside the scheduled maintenance programme and are characterised as mechanical accessory faults. It aims at facilitating the design and implementation of better and more efficient maintenance strategies based upon the concept of Condition Based Maintenance, which will ensure the preservation of the level of availability and reliability of the vessel ship’s components, as well as the desired level of performance and quality of the ship’s components. The effective and efficient fault detection and predictive maintenance is translated in profitability optimisation for FOINIKAS as it leads to overall maintenance costs reduction and risk mitigation. Furthermore, these strategies will enable FOINIKAS on being proactive rather than reactive, towards controlling unpredicted damages and/ or mechanical failures, in order to avoid unnecessary costs.
Towards this end, the scenario aims at: (a) identifying the supplies in machine room equipment of each ship that are needed to be stocked or changed and (b) identifying the root causes of the equipment malfunction prior to the estimated Time To Live by the equipment manufacturers (if possible). The pillars of the design of a smart and more efficient risk maintenance management strategy are the estimation of the failure mode probability and the associating failure mode criticality. These estimations will provide insights for the design of an optimised inventory management based upon more accurate equipment fault detection, and optimised proactive equipment maintenance very close to the actual time-to-live of the equipment. To achieve this, data analytics techniques will be applied towards the analysis of:
- Data from the Defects Reporting Database of FOINIKAS that incorporates all communication and details about defects in ship components and equipment along their status with metadata such as the date created, due date, vessel, description and category of the defect, the list of action taken for this defect, etc.
- Data from the Planned Maintenance System of FOINIKAS that incorporates all the information about the maintenance tasks performed in intervals according to manufacturers and class requirements with metadata such as the code of the task, the component under examination, the description of the scheduled task, the status, the start date, the due date, etc.
The execution of the described scenario is composed by the following three test cases:
- P1SC1_1 – Representation model design: In this first test case the model representing the dependencies between the components is defined and the criticality of each component in terms of failure effects (i.e. the level of impact by the breaking down of the component with regards to individual operation and the propagating effects to interconnected systems) is manually declared. The execution of the test case is illustrated in the following figure.
- P1SC1_2 – Predictive diagnostics estimates extraction: In this test case, the predictive diagnosis estimates for the failure cause probability regarding the vessel’s components under investigation are produced. In this step, the proactive actions that can be taken for the design of a proactive maintenance strategy are extracted. The execution of the test case is illustrated in the following figure.
- P1SC1_3 – Predictive prognostics estimates extraction: In this test case, the predictive prognostics estimates for the failure effect probability regarding the vessel’s components under investigation are produced. In this step, the proactive actions that can be taken in order to prevent or be better prepared against components failures through the design of a proactive maintenance strategy are extracted. The execution of the test case is illustrated in the following figure.
In accordance with the demonstrators implementation plan, the first phase of all three test cases (P1SC1_1, P1SC1_2 and P1SC1_3) was implemented and integrated with the BigDataOcean platform. The described test cases were executed and assessed under real circumstances in order to provide the initial valuable feedback and evaluation of both the service that was developed for the test cases execution, as well as the interaction with the BigDataOcean platform. The initial results indicate that the envisaged business situation is already improved. Valuable knowledge was produced, even in this initial phase, with regards to the design and implementation of better and more efficient maintenance strategies based on the estimation of the failure mode probability and the associating failure mode criticality. Furthermore, the failure cause probability and the failure effect probability enables the design of an optimised inventory management and optimised proactive equipment maintenance.
This valuable feedback and evaluation that was collected and is further analysed in order to extract useful insights for the second phase of the implementation of all three test cases. As indicated also in the presented results, there is room for improvement in the implementation of the test cases in the second phase in both the perceived usefulness and the perceived ease of use. Hence, in the second phase a series of adjustments and refinements will be introduced towards the better addressing of the stakeholder’s requirements and needs, as well as the added value that is offered from the developed service for this scenario.