Condition monitoring and fault detection in steam production system

Abstract

A ship’s steam generation system is a critical part of vessel operation, with its efficiency and reliability being in key role in energy efficiency and operating costs. Traditionally, maintenance of these systems is performed at predetermined intervals. However, the condition of components may deteriorate sooner than expected, potentially resulting in equipment malfunction or failure before the scheduled maintenance. Condition monitoring and predictive fault detection aim to identify the deteriorating condition of components at an early stage, enabling cost effective scheduling and planning of maintenance actions and spare part procurements.

The objective of this thesis was to investigate suitable condition monitoring and fault detection methods for the ship’s steam generation system. In the marine environment, both customer requirements and the added value by condition monitoring relative to the required investment must be considered. The literature review examined general components and equipment used in steam generation systems supplied by Alfa Laval. Furthermore, the review covered condition monitoring methods commonly applied in industrial facilities through both theoretical background and practical examples.

The practical part of the thesis evaluated the applicability of different condition and fault detection solutions for the ship’s steam system. The solutions were assessed based on comprehensive criteria including implementation cost, expected benefits, retrofitting feasibility, and suitability for different operating models. The evaluation considered both commercially available solutions, provided by component manufacturers and software suppliers utilizing model-based monitoring tools, as well as solutions that could be developed in-house. The assessment was based on the fact that the existing automation system already collects a wide range of process measurements for process control purposes. The study also examined how future measurement implementation should consider the requirements of condition monitoring in addition to process control, allowing the measurements to serve multiple purposes as efficiently as possible.

Based on the findings, a phased implementation plan was developed to enable the step-by-step adoption of condition monitoring solutions with different levels of complexity and investments. In the first phase, control signals generated by the existing automation system would be combined with other available process measurements. In the studied system, monitoring was limited to process outputs, meaning that gradually developing faults may remain undetected if the control system and actuators are able to compensate for their effects. This limitation could be addressed cost-effectively, to selected components, such as variable-speed pumps, using the existing automation and control system. In the second phase, component models would be developed for applications such as offline performance analysis. In the final phase, these models would also be utilized for real-time condition monitoring once the model development and validation processes have been demonstrated to work reliably. Compared with the first-phase solutions, implementing model-based methods or investing to similar commercially available software, was considered to require significantly greater financial investment and technical expertise. In addition, the thesis proposes several additional measurements that would improve the reliability of condition assessment for system components. For example, adding differential pressure measurement to the cooling wa-ter line of the surplus steam condenser would enable more reliable assessment of conden-ser condition and performance, allowing maintenance activities to be scheduled according to actual need.