Maintenance and inspections of ships are traditionally based on a preventive scheme where components have been maintained according to a time schedule. This approach is based on the assumption that a component has a defined lifetime, after which its failure rate increases. However, estimates of lifetime have large uncertainties and a large percentage of failures are not age-related and are therefore not adequately addressed by preventive scheduled maintenance. We develop new approaches based on the recent availability of large arrays of sensors, which monitor condition and operation of machinery and power systems.
Sensor data are becoming available for the first time on global ship fleets, with efficient broadband connectivity to shore. We suggest new approaches to condition monitoring, which is the process of identifying changes in sensor data that are indicative of a developing anomaly or fault. In addition to using previous failure data and pattern recognition techniques to detect anomalies, we test model based approaches that exploit knowledge on the sensors and the conditions they assess. For the design of sensor monitoring systems, a key challenge is to determine the level of resolution in time and sensor density needed to have a precise dynamic picture of the actual health of the system. Borrowing strength across sensors and ships in a fleet is an important aspect, leading to increased safety of a whole fleet.
Anomaly detection via signal reconstruction based on the past
We develop a multi-sensor, multi-scale statistical approach to detect, as rapidly as possible, anomalies. Data originate from sensors covering different aspects of a ship operation (f.ex. propulsion power, speed over ground and ship motion in four degrees of freedom). The developed method employs Auto Associative Kernel Regression (AAKR) for signal reconstruction, and the Sequential Probability Ratio Test technique for anomaly detection, where different hypothesis tests looking both at mean and variance deviations have been tested. We demonstrate that our model produces good reconstructions and as long as the parameters are tuned carefully, alarms are triggered appropriately.
Models for ship operations and efficiency
Understanding ship-internal and external factors that regulate ship propulsion and fuel efficiency is important to optimise ship design and operations. Deviations between the measured ship speed and the ship speed predicted on the basis of propulsion power and other internal and external factors, is an indication of a possible anomaly, incl. hull, propeller or engine damage. We compared statistical models which allow such prediction, using also external environmental factors that clearly affect sensor data. We are able to produce very good predictions, which can be further used for anomaly detection. A similar approach was also successfully tested on an cooling system for a motor.
- Unsupervised anomaly detection in sequential categorical log data
- Surprise detection for sub systems by multivariate multiple regression
- Dimension reduction procedures for changepoint detection
- Improving the AAKR test component
- Hull condition monitoring, sensor based and virtual