Statistically Predicting Electrical Arcing
UL engineers are using predictive modeling to help quantify risk and advance standards related to electrical arcing. We showcase a broad range of updates of arc-related research initiatives.
WHY STATISTICALLY PREDICTING ELECTRICAL ARCING MATTERS
Electrical arcing can create significant health and safety risks for home occupants and firefighters. According to a statistical report by the National Fire Protection Association (NFPA), an estimated 44,800 home structure fires reported to U.S. fire departments in 2009 involved some type of electrical failure or malfunction as a factor contributing to ignition. These fires resulted in 472 civilian deaths, 1,500 civilian injuries, and $1.6 billion in direct property damage.1 Approximately 57 percent of these fires originate from wiring and related equipment.2 UL is working in the area to prevent loss of life and property. We are using our data to quantify risk, inform code and help shape standards.
Electrical Fire Safety continues to be a concern, and arcing is a common cause of home building fires. Even if circuit breakers function as they are supposed to, arcing can still occur.
From 2005-2009, home electrical fires represented 13 percent of total home structure fires, 17 percent of associated civilian deaths, 11 percent of associated civilian injuries, and 21 percent of associated direct property damage.3 Home electrical wiring can be damaged through a number of common occurrences during installation or after. These include over-stapling, crushing, bending, penetration by screws and nails and damage caused by rodents and insects. Elevated temperatures and humidity can also adversely affect cabling over time, which can lead to arcing faults and ignition of combustibles in proximity.
For these important reasons UL is investing its expertise in helping better understand and prevent fires and damage due to electrical arcing, which we define as a luminous charge of electricity across an insulating medium, usually accompanied by the partial volatilization of the electrode.4
WHAT DID UL DO?
UL has the long-term goal of advancing fundamental knowledge related to electrical arcing by developing research, new approaches and standards. Recently, we have focused on applying statistical methods to our work. Some key areas of this New Science include:
- The use of parametric data from large sets of arcing data to build statistical models that characterize arcing behavior.
- Identification of the lognormal probability distribution as the key statistical model for peak current and other arcing parameters.
- Identification of the importance of total arc energy release as a primary indicator of material ignition.
- Characterization of expected arcing event lengths based on test conditions.
Figure 1. Distribution of Peak Current for Carbonized Path Arcing. Peak Current Is Normalized With Respect to the Short-Circuit Current
PARAMETERIZING AND STATISTICAL ANALYSIS OF ARCING DATA
UL has pioneered a key approach involving the use of parameterized data extracted from large (>100,000) sets of arcing data, which is then analyzed using a variety of statistical methods. UL’s analyses have shown that arcing events typically follow statistically understandable patterns, which can be modeled and used for predicting probabilistic outcomes. This information has been particularly helpful in predicting the distribution of real-world systems, such as the arcing current expected in a residential branch circuit with known length and wire size.5
The approach involves breaking down arcing events into small units (e.g., a single arcing half-cycle when analyzing residential AC arcing), then extracting single numerical parameters from each of these units. These parameters are then indexed to a variety of environmental test conditions. This data set is then analyzed to gain insight into arcing behavior through the use of statistical tools.6
This method has been applied to the analysis of peak arcing current, strike and stop phase angle and voltage, and arcing energy. For example, a review of arcing peak current has shown that the probability distribution function (PDF) of these data is described by a lognormal function, with the median value generally at 75 percent to 80 percent of short-circuit current. Importantly, this relationship is found to be true regardless of the magnitude of short-circuit current (whether it is arcing occurring 5A or 500A, series or parallel arcing, carbonized path or point contact arcing).
ARC ENERGY DISTRIBUTION AND IGNITION THRESHOLD
UL determined that the amount of energy needed to ignite non-metallic (NM) cable was characterized assuming that “ignition” was achieved if any observable discoloration or charring was present on the NM cable. Data determined that the threshold for igniting NM cable through arcing is approximately 2kJ.7
Figure 2 . Cumulative CDF Fits From (left), Comparing Data From Samples Showing Ignition to Those That Did Not. Numbers Show Coordinates for 5 Percent Probability of Ignition and Point Where the CDFs Cross
Importantly, the use of energy release in predicting material ignition was applied in setting pass/fail criteria for UL Standard 1699B, where it was determined through experiments that there is a 5 percent probability of burning through a 1/16 inch PV connector when 750 J is released.8 This demonstrates another value of our New Science work. We are able to use data to determine and set objective test criteria, enabling standards to be based on science.
UL research indicates that both DC and AC systems show that power is not a significant variable with respect to ignition in most arcing situations of short duration, suggesting that thermal loss through convection or conduction is not significant for the arcing events under consideration.
ARCING EVENT LENGTH
UL found that the total length of an arcing event is dependent on the arcing current and type and gauge of cable used.9 The arcing event tends to be shorter at higher currents than at lower currents. Arcing events also tend to be longer when a solid conductor is used, in comparison to standard conductors, which result in shorter arcing events.10 While these observations may have been noted before, the UL work has quantified the effects of stranding and arc current on arcing event length.
UL has demonstrated through research and analysis that the probability of arcing continuing to occur at a given point in time after the arcing event starts can be characterized by an exponential decay function.
Figure 3. Plot of the Probability of Arcing With Respect to Time and Short-Circuit Current. Time Zero Equals the Start of the Arcing Event
UNIQUE CHARACTERISTICS OF CARBONIZED PATH AND POINT CONTACT ARCING
UL research in 2010 suggested that carbonized path and point contact arcing differentiate significantly only through a change in strike voltage. The carbonized path shows strike voltages over a given threshold value, while point contact evidences voltages at all levels.11 Subsequent research confirms this previous finding and found that the method of arcing itself can also have an influence on behavior.12 For example, the distribution of strike angle differs between the arcing methods described in UL Standard 1699, Sections 40.3 and 40.4. The strike angle distribution in Section 40.3 shows a higher median value (approximately 90 degrees) than the methods described in Section 40.4 (approximately 40 degrees). These two methods also show different arc event lengths, with Section 40.3 resulting in shorter arcing events than Section 40.4. In addition, UL research found a series contact resistance that tends to lower the peak current values for point contact arcing relative to carbonized path arcing, which is on the order of 30 mΩ.13
UL is continuing to research and analyze electrical arcing. Several areas under additional review include quantifying the effect of arc gap and peak current, improved arc ignition modeling, spectral analysis and glow connection phenomena. These topics will likely be shared in future New Science journals.
UL’s commitment to Fire Safety is long-standing, and while arcing has historically been associated with fires and is not new, UL’s statistical approach is. Our innovative methodology has enabled us to predict arcing behavior, expected energy release, fire incidence and other areas where we are able to better prevent arcing incidences and better protect residents and firefighters. We have developed standardized tools and analysis methods, and have accumulated knowledge that enables us to problem solve more effectively and efficiently. Additionally, our expertise and research-based knowledge have allowed us to collaborate with the National Electrical Manufacturers Association (NEMA) to offer a free online program designed to teach effective and safe ways to install and troubleshoot issues with arc fault circuit interrupters (AFCIs). UL is also working with the National Electric Code Panel, IEEE and NFPA, using science to further advance understanding and prevention of electrical arcing.