3.2 Deterioration due to Running
Running of the target engine gives changes to state variables related to the soundness of each element, even Through the running, under normal conditions. To reflect the deterioration, deterioration occurs. soundness is defined, which indicates margins to the limits. The soundness index sij is a scalar value, where (sij = 1) means the jth state variable of the ith element is normal and (sij = 0) means the jth state variable of the ith element is failed. The integrated soundness index si of the ith element is also calculated from sij's. The default deterioration can be defined as a relation like constant rate decreasing which is determined as a reflection of a statistical analysis, e.g., MTBF. A threshold to be considered as abnormal should be determined according to the nature and the importance of the element.
If the direct monitoring of the deterioration is possible, the soundness index can be estimated with related state variables. Most of the soundness indices need to be estimated through simulation with corresponding relations.
4. SOUNDNESS INDEX
4.1 Classification of Deterioration
Deterioration is classified into three groups; the first is a linear change like wear of a material surface, the second, a logarithmic change like fatigue, the third, a sudden fail without apparent deterioration. A preventive maintenance can't be effective on the third group. Effective monitoring for the initiation of failure and countermeasures incase of the occurrence of the failure are essential for that group.
For the elements whose deterioration is classified into the first, wear type, group, monitoring the extent of the wear in depths and the rate of change of them. The soundness index can be defined as a normalized wear depth and its rate, thresholds for these indices are decided from mechanical viewpoints.
4.2 Accumulation of Damages in Elements
For the elements whose deterioration is classified into the second, fatigue type, group, the soundness can't be observed directly. The estimation of the accumulation of damages from stresses is essential. Continuous monitoring of the related state variables like stresses gives a good estimation of the accumulation of the damages. In these cases, failures are quite probabilistic and countermeasures in case of the failure are also essential.
5. DIAGNOSIS AND MAINTENANCE SCHEDULING
5.1 Monitoring of Soundness of Elements
A regular monitoring gathers temperatures, pressures and flow rates at the typical points. Human operator gathers much complicated information such as sound noises, vibrations, leakage of water and oil. The former sensor based data can be easily stored in the structural and functional model, while the latter subjective information is difficult to be added to the model. A recent hand-held computer and a voice-memo technique enable to gather these data. Still the handling of the subjective information remained in difficulty. They can be used in the analysis of the expert engineer.
We also propose a multi-agent monitoring system for a large and complex plant. It consists of multiple independent agents who monitors each assigned part of the operation and maintenance database like the structural and functional model of the marine engine. One agent monitors present states of the engine, the other, historical changes, for example. This system can be realized where onboard LAN and several PC's are equipped.
5.2 Inspections and Repairs
The structural and functional model is a combination of actual acquisition of running data and estimation through relations. A regular inspection gives information about present states of such estimated ones. In that case, the corresponding states variables are replaced. If some significant difference is recognized, simulations are performed to identify causes of the difference. It may be necessary to modify the relation itself.
The structural and functional model of the target engine contains soundness indices of each element. As thresholds are determined according to the nature and the importance of the element, margins in time can be estimated for each element. Timings of repairs are planned based on these margins.
5.3 Estimation of Present Status and Future Changes
As mentioned above, most of the state variables of engine elements can't be observed directly. It is necessary to estimate the present status of the engine. Historical data of the state variables are accumulated in the structural and functional model. From the past to the present, a simulation is performed to estimate the present status. Using the same procedure, we can get the estimation of the future status on the engine.
6. HUMAN INTERFACE FOR OPERATION
It is important for the support system to be easy understanding and handling. Voice communication is essential among human operators. Recent progress in speech recognition and voice synthesis enables such support system even in a PC [2] . Under a remote monitoring and control condition, it is important to give on-site information to remote operators. In addition to a conventional control panel, a 3D-VR display and sensory feedback control devices are effective. These functions are also provided on PC's.
7. CONCLUSION
Maintenance work for marine engines is one of the most essential and hard works onboard. We focused on the accumulation of the monitoring data of the engine and estimation of the states not to be monitored directly. We propose a structural and functional model for the target engine that has engine elements and relations among them. An initial and default model is constructed during a design work and a manufacturing work. The default model can be improved through replacement of the element structure and the relations step by step.
The inverse simulation from the present to the past gives information about causes of the failures. The simulation from the past to the present gives estimation of the present status of the elements not to be monitored directly. The simulation from the present to the future gives information about possible danger caused by failures. Based on the information about the present and future status of the engine, timings of repairs can be decided to make a maintenance-scheduling plan.
Human interface of the support system is also essential for the actual use. Application of the speech communication and 3D-VR expressions are powerful techniques for the interface. Remote operations are inevitable for a large and complex plant. Sensory feed back control devices give on-site information in case of the remote control operation. All these functions can be provided on PC's now.
We expect the coming engine operation works including maintenance work to be comfortable through co-operations between onboard operators, monitoring agents and land support experts.
ACKNOWLEDGEMENT
The basic idea of this study is based on the discussion with and questionnaires to expert marine engineers. We are grateful to these expert engineers from shipping companies, engine manufacturers and related organizations.
REFERENCES
[1] Numano et a1.: Preventive Control of Marine Machinery by Macroscopic Simulation, Technical Papers of ISME KOBE'90, D-3-17.II, 1990.
