Methods for Hull and Propeller Performance Monitoring (Part 3)

Antti Solonen, Data Scientist

Part 2 of this blog series introduced the basic philosophy behind our Propulsion Power Decomposition (PPD) model. Here, we dive deeper into how the model can be used for performance tracking, and how we solve the common issues faced with STW log data. 

Modelling Performance Changes over time

The most straightforward way to apply the PPD model for performance monitoring is to apply the ISO-19030 framework; use the model as the speed-power reference. The benefit of this is that the model contains a more comprehensive set of descriptions of different resistance components, removing, to some extent, the need for filtering the data to the reference conditions.

However, we go one step further, and include the temporal changes in the vessel performance (e.g., changes in the frictional resistance coefficient) explicitly in the PPD model as a time varying extra resistance term. This extra resistance term is continuously learned from data using Bayesian dynamical state space estimation techniques. In this way, the model is able to keep itself “up to date” with the current hull and propeller condition.

To demonstrate the effect of the dynamically changing extra resistance term, we plot an example of the PPD model residuals (predicted power subtracted from the observed power) with and without the extra resistance term below in Fig. 5. Note how the systematic long-time trends in the residuals disappear after adding the extra resistance term.

Power model residual

Fig. 5. Power model residuals with and without the extra resistance term.

Combining this kind of modeling with our Virtual Speed Log technology to alleviate STW log issues (discussed further on) allows us to raise the accuracy of the hull performance tracking to a new level. A comparison of our approach to the ISO-19030 method is given below in Fig. 6. One can clearly see how the noise level is reduced, which allows for more detailed assessment of individual hull treatment effects.

ISO vs Eniram

Fig. 6. Top: ISO-19030 performance values compared to the ones computed via the next-generation Eniram model. Bottom: estimated change in the hydrodynamic resistance coefficient due to fouling.

Tracking Calm-sea Power

Another pleasant consequence of integrating the hull and propeller performance model tightly with the PPD model is that we can, at any time, simulate the vessel at any condition with or without the extra resistance component included. For example, we can estimate how the calm-sea power (power consumption at calm sea conditions) at some selected speed level evolves in time. An example of such a time series is given in Fig. 7.

Calm sea power comp

Fig. 7. Calm-sea power consumption estimate at STW=18kn. Vertical lines are different hull treatments.

This approach also enables, for instance, vessel comparisons, since we can normalize the vessels to the same “operating point” by removing the external resistance effects and the effects due to the varying speed profiles. Such comparisons are hard to do with an ISO-19030 type of approach, which present fouling as a relative difference to a reference. In addition to vessel comparisons, one can derive interesting “fleet-level” performance indicators, such as the total calm-sea propulsion power at some selected speed over a collection of vessels. Example illustrations of vessel comparisons and such “fleet-level” performance tracking approaches are given in Fig 8.

Power18

Fig. 8. Left: calm-sea power at STW=18kn time series compared for 5 vessels (note the dry-dockings for vessels 3 and 4). Right: the sum of calm-sea powers over the 5 vessels.

Virtual Speed Log

Speed through water (STW) is perhaps the most important variable when looking at the energy efficiency of a vessel. At the same time, it is generally known that the onboard STW logs are one of the least reliable sources of data. We estimate that around 20% of vessels experience severe STW log issues, and less severe problems are obviously encountered even more frequently.

There are two main issues with STW logs. First, they are often poorly calibrated, which can be seen as a constant offset between SOG and STW over a long period of time. In addition, the noise level of the speed logs is often high, and sometimes the logs suffer from sudden large errors. Since the speed-power relationship is roughly cubic, any speed-power – based performance monitoring method that relies on STW logs (such as ISO-19030) is very sensitive to these errors.

Our solution to the STW log problem is the Virtual Speed Log technology, which combines various data sources (hydrodynamic modeling using RPM and torque, SOG, STW and current hindcasts) via a dynamical Bayesian estimation framework to provide an accurate and robust estimate of STW. Figure 9 below illustrates the approach. The details are skipped here, and left as a topic for another blog post, stay tuned!

Virtual stw chart

Fig. 9. Eniram Virtual Speed Log combines data from different sources in a Bayesian dynamical estimation machinery. The resulting STW estimate fixes two common issues in STW logs.

How large is the effect of a poor speed log on performance monitoring results? Below in Fig. 10 we give two examples of ISO-19030 results calculated with STW log and Virtual Speed Log. The first example (Fig 10a) compares results in the case when the STW log is relatively good (no major calibration issues). The results look similar, as expected. However, when the STW log is poor, the situation is different (Fig. 10b). In this case, the STW log has major calibration problems that change over time, and the results change significantly; the STW log based analysis indicates that the performance of the vessel is gradually improving over time, whereas in reality the performance has not changed much at all. This example illustrates that using a faulty speed log in such long term performance tracking can give misleading results. We use the Virtual Speed Log as the basis of our new performance monitoring solution.

Speed log comp

Fig. 10. ISO-19030 results computed with STW log (left column) and Virtual Speed Log (right column) in two cases: a) good quality speed log and b) speed log with calibration issues. Color indicates sea area.

Summary and Conclusions

As a summary ISO-19030 approach is a good start to track hull performance based on typically available data. However, there are a number of challenges related to the approach: extensive filtering of input data restricts the number of usable data points; limited normalization decreases accuracy of results; and utilization of measured STW data may result in misleading conclusions.

Based on our practical experience and extensive research Eniram has developed next generation approach to tracking performance alleviating afore mentioned problems: instead of filtering, our method uses advanced statistical modelling to enable more extensive normalization leading to higher yield of available data points and superior accuracy; and instead of utilization of problematic STW log data we use our patent pending Virtual Speed Log technology. This unique approach also enables tracking both relative and absolute calm sea performance.

Eniram Hull & propeller performance online application is now available for vessels equipped with Eniram Platform. We can help you track hydrodynamic performance to manage hull fouling and evaluate impact of investments. Contact us today for a demo or more information:

Demo request  Datasheet