Fault identification on reefer systems can be a time consuming and expensive process, even for a skilled operator. Therefore, automatic accurate identification of system failures can help reduce cost of service and maintenance on reefer systems. The benefits include shorter service time as well as it reduces replacement of non-faulty components during manual fault identification.
One approach to automatic faults identification is Kalman Filtering, which have been around since the 30s. One of the first applications were guidance of the Apollo spacecraft which led us to the moon. Today it is widely used in a whole range of technologies which include several sensor inputs to a system, for estimation of hidden states, measurement noise rejection and fault detection.
Lodam believes that if a Kalman filter is applied to a reefer system it could increase the operational time window, and aid in reducing the number of faulty diagnostics reports (e.g. devices sent back that are working properly) as well as keeping the system running, if one or more sensor stops working.
Refrigeration system uptime improvement by:
• Precise and early sensor fault diagnosis.
• Sensor substitution and estimation.
• Shutting down affected sub-systems.
• Possible cost savings through cheaper sensors which uses Kalman filtering to improve measurement quality.
• Estimate unobservable states, i.e. measurements without an actual sensor.
The Kalman Filter is based on a model of the system, which uses state measurements as input, and outputs estimates of the measurements. The correlation between these estimates, are then correlated with the actual measurements, and an updated estimate is calculated, considering the model as well as the sensor-reading. The output, is a filtered estimate of the sensor measurement, as well as a matrix that can be examined to determine if a sensor needs replacement or is drifting from its operating point. The main advantage is that the system can operate with missing sensory information, as the filter automatically disregards missing sensors, by reducing the influence this has on the estimate. This is done by modifying the correlation matrix described above.
Further information Please contact Kresten Kjær Sørensen (Lodam employee) by email: firstname.lastname@example.org or by phone: +45 73 42 37 37.