Garron K. Morris

Driving energy efficiency with design optimization of a centrifugal fan housing system for variable frequency drives

Increasing demands for electrical equipment efficiency has renewed interest in improving the electrical and thermal efficiency of industrial equipment. Variable speed drives are typically air-cooled and fall under recent international standards for improved fan efficiency. This paper details the modeling, design optimization, and experimental verification approaches used to optimize blower housing designs for variable speed drives. The design of the blower housing is just as important as the blower selection. By modifying the housing dimensions, the shape and quantity of flow exiting the housing can be controlled. First, the impeller and inlet ring geometry was directly imported into Icepak from a CAD model provided by the manufacturer. Moving Reference Frame and multi-level meshing techniques were used to provide an accurate representation of the air flow. Next, a flow-pressure curve was created by varying the outlet pressure. The blower performance curve was found to follow, but consistently under-predict the empirical fan curve data given by the manufacturer. Using fan laws and a multi-objective optimization approach, a model fan speed that was 2.4% higher than the operating speed was found to make the predicted and manufacturer performance data agree with less than a 3% error. Next, a parametric model of the blower housing was created in Icepak using the tuned blower model. Four parametric variables included the distance from the impeller to the front, bottom, side, and back of the housing was chosen. A fifth variable, pressure was chosen so that the effect of outlet pressure on flow could be extracted. The blower housing was optimized using a Design of Experiments (DoE) technique where the geometry of housing was varied in a structured manner to capture expected second order behavior. The 27-run DoE was performed in Icepak and the volumetric flow through discretized portions of the outlet were recorded. The DoE data for each section of the outlet were fit to equations using a backward regression technique. A genetic algorithm-based optimization technique was used to create housing designs for two different variable frequency drives. Prototypes of the housings were constructed for each design and flow-pressure curves for three samples of each design were measured on a flow bench. The measured curves were found to agree with the predicted blower performance in each housing design to within 7%. Design curves that could be used for other housings were also generated.

Measurement of a variable frequency drive loss and efficiency using both calorimeter and powermeter

Variable frequency drive efficiency plays an important role in green energy applications. However, how to accurately measure drive loss and efficiency is still a question to the industry. This paper reviews the powermeter method, calorimeter method, and factors that affect accuracy. After a calorimeter is constructed and calibrated, a variable frequency drive is tested using both methods. Test results are compared and uncertainty analysis is given.

Predictive Reliability Models for variable frequency drives based on application profiles

A Variable Frequency Drive (VFD) is a system that can control speed, torque and position of AC motors. In process automation and motion control for various industrial and commercial applications such as cooling fans, pumps, belt conveyors, rotary kilns, and elevators, VFDs are used. Reliability of a VFD is application-dependent, as the magnitude and type of failure-governing stresses vary with each application. Understanding operating conditions and quantifying reliability under those conditions is essential for ensuring proper operation of VFD to meet the service life expectation of the customer. Analysis of field failures led to identification of failure governing stress factors. Researching technical literature on operational characteristics of VFD for various applications led to proper definition of a 24-hour worst-case use load profile. Using worst-case use load profile as baseline and by accelerating the usage rate, acceleration factor (AF usage) is determined. In addition, by using VFDs heat sink temperatures under both worst-case use load profile and accelerated test load profile and using Arrhenius Relationship, acceleration factor due to thermal stress (AF Thermal) is determined. The overall Acceleration Factor (AF) is a product of AF usage and AF thermal. A Reliability Demonstration Test (RDT) under accelerated conditions was performed to estimate Mean Time Between Failures (MTBF) and Life at which 10% of the population is expected to fail (L10) under worst case use load profile and also to establish life-stress relationship. This information aids in development of Predictive Reliability Models for VFDs for any application.

Predicting circuit board reliability for motor drives under varying application and environmental conditions

Operating and non-operating temperatures as well as power and duty cycles affect electronic component reliability. The ability to predict the reliability of a printed circuit board (PCB) is important during the early part of the design process. This paper explores the effects of using both single-value assumptions and probability distributions for environmental and operating stresses in a standards-based reliability model of a PCB by comparing the predictions with the measured field data. First, failure and shipping data for a power interface circuit board from a motor drive that has been in production for over two years was obtained. An adjustment factor for the data was developed to account for non-operational storage time. A Weibull analysis was performed on the adjusted data set and L1, L10, and MTBF reliability metrics were computed. Next, a reliability model was constructed using the bill of materials, circuit schematic, and RIAC 217Plus reliability prediction standard. The environmental and operating profiles were then defined using 217Plus defaults, worst-case, and probability distributions. Deterministic predictions were made using the 217Plus default and worst-case profiles and predicted values were compared with the field data. Stochastic predictions were made using Monte Carlo simulation techniques and also compared to field data. Predictions using the Monte Carlo technique were found to have a lowest maximum error of 22%. The discrepancy between the predictions and field for L1 and L10 reliability were reduced by using a Weibull shape parameter that matched the field data instead of the default value of one. Finally, a correction factor was developed to further improve the predictions.

Taking the (emotional) stress out of HALT

The HALT process used at Rockwell Automation takes a product, printed circuit board, or component through five test sequences: low-temperature step-stress, high temperature step-stress, random vibration step-stress, combined temperature cycling/random vibration, and rapid temperature cycling. The temperature or vibration amplitude where any anomaly occurs, as well as the time, relative to the start of the test sequence, is noted. For each anomaly that occurs, a temperature and vibration robustness factor is calculated. The robustness factor is then translated to a probability of occurrence. Additionally, the severity of each anomaly is independently evaluated by engineering and quality functions in terms of severity of the effect on warranty, operation, and safety. The independent severity assessments are then combined into a composite severity. Recommended actions for remediation of each anomaly are made based on the occurrence and severity assessments using a HALT Action Matrix. This anomaly assessment process has eliminated the emotional arguments between various functions about remediating HALT anomalies and has ultimately benefitted the customer as more anomalies are being fixed than before the HALT Action Assessment process and tool were introduced.

Operating IGBTs above rated junction temperature limits: Impacts to reliability and electrical performance

A fixture was designed and built to test power cycling reliability at maximum junction temperatures of 150°C and 175°C. A 1200V, 150A IGBT six pack module was mounted on a single cold plate and three sets of cold plates were connected to DC power supplies controlled by LabView. The IGBTs on each cold plate were cycled to give a junction-to-case temperature rise of 100°C with a maximum junction temperature of 150°C. Every 2000 elapsed power cycles, the cold plates were removed from the test setup and the junction-to-case thermal resistance and collector-emitter saturation voltage was measured. Power cycle tests were run until all IGBT modules exceed the failure criteria (5% change in collector-emitter saturation voltage or 20% change in junction-case thermal resistance). The power cycle test was repeated on a new set of IGBTs with the same temperature rise of 100°C, but at 175°C maximum. Weibull analyses of the power cycle test data resulted in shape parameters of 8.6 and 8.8 for the 150°C and 175°C maximum junction temperature data, respectively, that is indicative of rapid wear out. Inspections of the failed modules indicated wire lift-off. The B5 life at 175°C was found to be 36% lower than the B5 life at 150°C. The power cycling test result at 150°C; however, was over two times higher than the life specified by the manufacturer. High temperature reverse bias testing at 175°C was also performed to stress silicon die and packaging materials under a DC bias in order to accelerate migration of ionic contaminants and silicon die impurities. Die temperatures of 175°C caused no degradation of the key electrical static parameter in 1400 hours, which exceeded industry standard test of 1000 hours.