Common failures in the electronic products can be traced back to thermal-related issues. While the power of electronic components and power densities are rapidly increasing, thermal management of electronic products becomes a challenge. Since the lifespan of electronic components is shortened drastically due to the high temperature, fans are widely used to create airflow in electronic products for cooling purpose. Therefore, fan is a critical part for thermal management in electronic products, and reliability of electronic products is heavily dependent on the fan's reliability. There is a standard IPC-9591 that standardized fan's performance parameters (including reliability). However, it is still difficult to compare the fan reliability directly among different fan manufacturers because they report their fans' reliability using different parameter values in Weibull distribution, acceleration factors, and reliability tests. This paper reviews, compares, and discusses the approach of fan life expectancy calculation. It concludes that IPC-9591 recommends the most conservative values (biggest Weibull shape parameter, lowest test temperature, smallest acceleration factor) for reliability analysis. Since bearing is the most critical component in fan, its life is also discussed in the paper. Considering the very light bearing loads in fan, it is recommended that bearing life should refer to its grease life, not the fatigue life.

/pdf/an-investigation-into-fan-reliability-wu6vna43hs.pdf

An investigation into fan reliability

A counting unit counts a current supplying duration for a component which determines a life of a device A prediction unit predicts the device life based on an integrated current supplying duration integrated by the counting unit and a lifetime of the component A monitoring unit monitors an efficient operating condition of the device based on a prediction signal outputted from the life prediction unit The counting unit and the life prediction unit is provided at a short-life fan and a long-life component respectively, and the life prediction unit predicts the life of the device based on information on the short-life fan and the long-life component

Life prediction and monitoring of components

Server fans are one of the most critical parts of the data center server’s thermal system solution and a significant failure cause of many electronic systems. A fan inside the computer chassis is used to draw cooler air from outside and expel exhausted air from inside to adjust a server’s thermal range of electronic components. Fan failure can reduce cooling efficiency and result in unexpected throttling, overheating, and thermal shutdown. This paper demonstrates an AI-assisted methodology that predicts fan failure by developing a probabilistic model that correlates fan speed to workload fingerprint representing resource consumption patterns in the spatio-temporal domain. We use training data from many workloads that execute on many machines of similar configurations and record resource utilization statistics along with environmental parameters with a configurable sliding window. The algorithm evaluates the phase distribution using CPU utilization features and corresponding fan speeds to characterize the expected behavior of phases within individual phase boundaries. This allows us to explain abnormal fan speeds due to CPU’s ambient temperature, voltage, and peak hour workload demands. The paper also presents a novel simulator that helps generate artificial datasets covering various edge cases to evaluate the prediction algorithm’s performance.

AI Assisted Fan Failure Prediction using Workload Fingerprinting

A fan of a computing system is associated with a performance characteristic. It may be determined whether a component associated with the computing system is installed based on the performance characteristic.

Determine installation of components based on performance characteristics

Fan cooling is one of the most critical domains of the data center server’s thermal management solution. A fan inside a computer chassis is used to pull in cool air from outside and expel exhausted air from inside to adjust a thermal range of electronic components in a server. However, mechanical parts are prone to unexpected failures due to anomalies or routine wear and tear. Consequently, this often reduces cooling efficiency and undesired performance throttling, leading to system shutdown and service disruption. Researchers have proposed various techniques to forecast failure events in advance in the last few years to mitigate such risks and reduce server downtimes. It should be noted that the prediction of random failures of motor-driver hardware components like Fan is a known complex problem. This paper proposes an improved probabilistic prediction technique that employs a multi-class change point detection algorithm to detect phase anomalies. The phase combines the two fundamental matrices - hardware telemetry sensor data that includes fan speeds, bearing vibration, acoustic noise, and various voltages and temperatures, and a software metric - the workload fingerprint derived from the synthesis of utilization patterns of system resources - CPU, memory, I/O, and network. We further explore the use of vibration data for detecting fan anomalies. 

Online Server Fan failure prediction by Vibration Analysis augmented by Hardware and Software Telemetry

A tractable model for predicting the failure rates of electronic chassis cooled by fans with finite failure rates is developed. The model accounts for two regimes prior to electronics failure: the electronic chassis is operated at a temperature which occurs with a fan that is operational and the electronic chassis is operated at a temperature which occurs when the fan has failed. Fan failures alone do not constitute a failure of the chassis. Such a model applies where cooling fans are not actively monitored, and where their failure is noticed only during the maintenance associated with an electronic failure of the chassis, i.e., the model accounts for the fact that the electronics in general continues to operate for a period of time following a fan failure. The model directly applies to cost/benefit studies in electronics packaging, including the determination of the improvement in mean time between failures (MTBF) to be anticipated when a chassis is implemented with a cooling fan, and the determination of the performance benefit to be anticipated when fans are actively monitored and replaced upon failure. >

J.M. Hogan

Papers

The effect of fan-reliability and cooling-performance on electronic-chassis reliability