3 Fault Examples

Overview | Results | Learn More

3 Fault Examples

The following fault scenarios are not unique for the used hardware, they can occur on any hardware based system. Naturally there are many more fault scenarios, contact us to learn more.

Fan Fault

The below figures show a graph of the model prediction output and actual measurements. At some point the actual measurements deviate from the model output. The left figure shows how temperature increases more than it should according to the model, which has predicted the temperature output based on load activity, fan speed, and other parameters. Examining the top right figure it is clear that the actual and predicted duty cycles are the same. This means the fault does not lie in the control signal. However, from the bottom right figure, it is seen that the fan RPM is lower than what the model evaluates it to be. This indicates that the fan is not running as it should. The reason might be increase resistance due to external conditions or the fan simply does not get the power it requires. This also explains the increase in CPU temperature.

VARYC Fault Detection Example CPU Temperature Fault

Power Fault

For the below example we have omitted the CPU prediction model to show how a faulty scenario might be overlooked without VARYC's Model Based Fault Detection System. From the left figure it is seen how temperature increases with the CPU load. This increase in temperature is perfectly normal, however the rate is too high. A conventional fault detection system would not pick up on this before the temperature gets critical, and even though we, in this example, omitted the temperature prediction model, we have another variable that we use to determine the fault. On the right figure we see how fan RPM decreases while it should increase as depicted in the model. The modeled controller duty output shows the same as the actual duty output, which means the fault is not in the control signal. Further examining the fan speed shows the fan RPM decreasing when load gets high, this means the fan does not get enough power, meaning either the power supply does not operate reliably under heavy load or the board is not able to provide enough power to the fan when the CPU is under high load.

Controller Fault Resulting in Performance Degradation

This example shows a very peculiar behavior in CPU temperature. At some point, while under max load, the CPU temperature actually begins to decrease. One might think that the fan is kicking in, but the fan is not able to reduce CPU temperature at this dramatic rate. Examining the right figures, it is seen that the fan indeed has increased to its maximum value. However that happened at around the same time as the CPU maxed out, and as mentioned before, we know from our fan models, that the fan cannot change the CPU temperature at that rate. The answer lies in the top right figure. As the temperature got into a critical state, the CPU underclocked itself to save it from potential overheating. This underclocking, results in this drastic reduction in temperature. The CPU Power State model output shows how the power state should be kept at maximum, since the computational power is needed. This means the system experiences a performance degradation without the users knowledge. VARYC's Model Based Hardware Fault Detection System lets the user know that the system is not running optimally and the critical increase in temperature might be due to faulty fan, change in ambient temperature, or change in the environment around or inside the system.