When a metal-casting plant, producing ductile iron castings for the automotive industry, suffered catastrophic shutdowns of both of its induction melting furnaces (see The Heat is On), the consequences were significant. The furnace failures occurred within three days of each other, shortly after the plant returned to production after Christmas shutdown. After investigating the first failure, plant and electric utility personnel assumed the cause of the first failure had been found. They resumed the start-up procedure three days later, only to have an almost identical failure of their backup furnace, causing yet another shutdown.

Plant engineers were concerned that recent changes in the electric utility’s distribution system might have initiated the failures. However, the utility denied any relationship between the changes and the failures. At this point, the discussion reached an impasse. The plant shutdown continued for nearly a week, while repairs were being completed, jeopardizing the plant’s just-in-time status with its automotive customers.

Upon completion of the repairs, plant personnel had to decide whether to start up the newly repaired furnaces, thus risking another failure. If production could not start, the plant faced the expense and delay of shipping patterns to its overseas sister company to make the necessary castings. The Thursday after the second failure, plant managers realized that if the facility could not restart by Sunday, the patterns would be shipped overseas and customers notified of the pending delay. The prospect shook the company all the way up to its CEO. At this point — with both furnaces shut down and only three days before the deadline — our team was called in to analyze the situation.

Testing pointed to switching transients associated with the metal casting plant’s 12kV power factor correction capacitor banks as initiating the failures (click here to see Fig. 1). These capacitors had been in service for more than a year prior to the failures, but had been operating in manual mode. Just before the Christmas shutdown, the capacitor bank control was changed from manual to automatic. This meant that the capacitors were switching periodically, whereas they had previously remained online all the time. Failures were linked with transients created by switching operations.

The engineering team recommended the capacitor banks be turned off and locked out. The plant returned to full production before the Sunday deadline, and the plant engineer kept the only key to the capacitor lock on the dresser in his home!

Although the immediate crisis had been averted, the full extent of the problem remained unclear. In addition, the plant would need the capacitors online by summer for voltage support and power factor correction, so the bank could not remain de-energized indefinitely. The team needed to conduct additional analysis that was broad in scope, including taking a more in-depth look at transients modeling, fuse testing, bus bar finite-element modeling, harmonics measurements, and simulation.

On-site assessment

Additional on-site testing showed that high-frequency transients were initiated each time the capacitors switched (Fig. 1). The transients were large enough to cause false turn-on and possible failure of the induction furnace silicon-controlled rectifiers (SCRs). The transients did not indicate problems with the vacuum switches; the problems were caused by the particular configuration of the electrical distribution system. At the end of the on-site testing, the engineering team formed a theory about the failures:

  • Capacitor switching transients caused false turn-on of induction furnace SCRs and a short circuit through the converter’s DC bus.
  • Fast-acting fuses blew to clear the short circuit.
  • The fuse blowing event “chopped” current in the supplying transformer, creating huge transient voltages across the bus bars.
  • Transient voltages were sufficient to jump the air gap between the bus bars and cabinet.
  • Deficiencies in fuse coordination allowed the resulting short circuit to burn until the bus bars separated.

However, additional engineering analysis, fuse testing, and computer modeling were required to test each phase of the theory.

Fuse testing at high-power laboratory

Fuses from the metal casting plant were obtained to test their response to high-current short circuits. The fuses were tested at 39,525 peak amperes at the high-power test lab. The results clearly eliminated the fast-acting fuses as the cause of the bus failures. Voltage developed across the fuses was limited to only about 1,500V (click here to see Fig. 2), much less than the 20,000V required to jump the air gap between bus bars and the furnace cabinet. This led the engineering team to abandon the initial current chop theory.

Computer modeling of transients

At this point, the engineering team suspected that the plant’s electrical system was amplifying transients due to its unusual response characteristics. The only way to investigate this theory, short of turning capacitors on and off with the induction furnaces energized and at risk of failure, was to model the power system’s response to high-frequency events. The model included both the plant’s 12kV capacitors and the electric utility’s four separate 34kV capacitors, assessing the effects of all the possible combinations of the seven banks.

Modeling showed that capacitor switching transients were severe enough to cause false turn-on, or even failure, of the furnace SCRs. Further analysis showed that the transients could be reduced by converting the capacitors to harmonic filters with the addition of line reactors. In fact, the capacitors had originally been selected to accommodate line reactors in a harmonic filter arrangement.

Measurements of power system harmonics

The process of converting AC power to DC and back to AC generates harmonic currents. Power factor correction capacitors do not produce harmonics but can create conditions that worsen harmonics on the power system. During the on-site assessment, our power system engineers learned that circuit monitors were located on the capacitor circuits and were storing harmonic information. This data revealed that voltage distortion at the plant was slightly higher than the IEEE 519 recommended level of 5%. Although harmonics had not contributed to the failures, voltage distortion levels were high enough to warrant further action. Converting the existing banks to harmonic filters solved both the switching problem and reduced excessive harmonic distortion levels as well.

Computer modeling of bus bar movement

The location of the bus bar failures indicated that current had either jumped the air gap between the bus bars and cabinet, the bus bar structure had moved during the fault and touched the cabinet, or the bus bars from different phases had touched during the event. Fuse testing eliminated the first theory. The bus bars were then modeled to determine if they moved or touched.

Fault currents flowing through bus bars or wires create large mechanical forces that act to repel conductors on different phases. The analysis work first required that a finite-element computer model of the bus bars be created. This model was then analyzed to determine the amount of mechanical force generated during various kinds of faults. Once the mechanical forces were determined, the amount of deflection, or movement, of the bus bars could be predicted.

Figure 3 (click here to see Fig. 3) shows the results of the modeling. Clearly, mechanical forces generated by the short circuit current were sufficient to cause bus bars to touch. Once the bus bars touched, the fault current flowed until bus bars burned in two.

Lessons learned

The final results of the engineering team’s extensive analysis revealed a simple cause of the failures: Bus bars were inadequately braced to withstand the forces generated during a short circuit (click here to see Fig. 4). The short circuits resulted from unintended SCR conduction due to capacitor switching transients, but the short circuits would have caused little damage if the bus bars had been adequately braced.

Our power system engineers inspected the induction furnace repairs and found that the furnace manufacturer had installed additional bus bracing after the first failures, indicating that perhaps the vendor suspected the real problem from the beginning.

In addition to the bus bracing, it was recommended that the bus bars be equipped with transient voltage surge suppression devices. These devices limit the magnitude of voltage transients and provide additional protection.

Finally, the power factor correction capacitors are being converted to harmonic filters. This conversion helps to limit the magnitude of voltage transients, and removes some of the harmonic currents produced by the furnaces.

The benefits realized from this engineering analysis were significant. First, the cause of catastrophic failures that had cost $750,000 in repair and restart for the facility were identified and eliminated. The just-in-time status was preserved, saving millions in potential lost sales. Not only was the electric utility exonerated, but the plant and utility gained a better understanding of the intricacies of their shared power system characteristics.

Ray is the director of Schneider Engineering Services based in Raleigh, N.C. He can be reached at Larry.ray@schneider-electric.com.


SIDEBAR: The Heat Is On

The furnaces at this plant melt scrap iron, attaining a temperature of 2,600°F. They operate by inducing currents in the scrap by rapidly varying the magnetic field around the metal. The highly fluctuating field is produced from a constant frequency AC source through use of large electronic power converters. These converters change 60-Hz AC voltage and current to DC and then to 200-Hz AC through silicon-controlled rectifiers (SCRs).

The induction furnaces were served by twin 12kV to 575V transformers. The furnaces each had an electrical capacity of 8MW. AC power was delivered from transformers to the SCRs through large copper bus bars, which experienced the catastrophic failures shown in the Photo.

The plant operated three 12kV capacitor banks for power factor control. These banks were turned on or off with vacuum switches. Testing and analysis showed that the vacuum switches initiated voltage transients each time the capacitors were turned on — and sometimes when the banks were turned off. Vacuum switches are commonly used on capacitor banks and other loads that require connection or disconnection at high power levels.