Prognostics-Based Product Warranties Yan Ning, Peter Sandborn, and Michael Pecht Center for Advanced Life Cycle Engineering (CALCE) University of Maryland College Park, MD 20742 USA
[email protected] Abstract—Prognostics and health management (PHM) is an enabling discipline consisting of technologies and methods to assess the reliability of a product in its actual life cycle conditions and evaluate its remaining useful life. This paper presents the application of PHM methods as a proactive and predictive means to enable new warranty approaches. Warranty service is typically performed after the occurrence of failure. However, with the implementation of the predictive and proactive warranty strategies enabled by PHM, companies can make decisions for warranty returns up front. The prognostics-based warranty models developed in this paper include part-based warranty return, lifetime warranty, and customized extended warranty, and a case study of warranty validation and user abuse detection is also provided. The results of this work can increase the competitiveness of businesses by reshaping their warranty policies, improving their maintenance practices under warranty, and reducing their warranty costs. Keywords-PHM; prognostics and health management; warranty; prognostics-based warranty; part-based warranty return; lifetime warranty; extended warranty; warranty validation
I.
INTRODUCTION
A warranty is a manufacturer or seller’s assurance to a buyer that a product is or shall be as represented [1]. When the product does not perform as expected, the manufacturer or seller must provide a remedy, such as repair or replacement, to fulfill the contract of the warranty. A warranty is part of the product purchase, whereas an extended warranty is entered into voluntarily and purchased separately [2]. An extended warranty provides additional protection beyond what the “base warranty” offers on the product. Warranties usually do not cover problems caused by “acts of God,” abuse, misuse, malicious destruction, ordinary wear-out, failure to follow directions, or improper maintenance. Increasing global competition has driven manufacturers to seek competitive advantages, and warranties are used as a marketing strategy. A warranty provides evidence that the manufacturer cares about its customers, indicates its willingness to solve possible problems, and ensures the dependability of its products [3]. Marketers use longer warranty coverage to convey the seller’s confidence in the product. For example, Hyundai started to offer a 10year/100,000-mile powertrain warranty in the U.S. in 1999 to differentiate itself from its competitors, and its U.S. market share increased from 1% to 4.6% by 2010, in part due to the warranty change [4][5].
Despite their advantages, warranties can be expensive for manufacturers and their suppliers. According to Warranty Week, the top 20 U.S.-based warranty providers paid over $25.5 billion in warranty claims in 2009 [6]. Automotive manufacturers and their suppliers spent $11.3 billion on warranty claims in the U.S. in 2009 [7], and the total cost of warranties for computer and related high-technology US companies is about $8 billion per year [8]. In order to decrease these costs, companies are searching for ways to design or test for reliability. Warranty costs are calculated based on warranty period – the length of the warranty coverage in terms of age (e.g., time) or usage (e.g., mileage) – and warranty policies. The expected warranty cost increases as the length of the warranty period increases, and different warranty policies result in different warranty costs. The most commonly used warranty policies are free replacement warranty (FRW) and pro rata warranty (PRW), as well as a combination of the two [9][10]. Under an FRW, the manufacturer agrees to provide a replacement or repair at no cost to the consumer if a product fails prior to the end of a specified warranty period. The costs of PRWs are shared between the customer and the manufacturer; the amount a customer needs to pay is usually equivalent to the amount of service received up to the time of the failure. Warranty policies can be either renewing (for example, the warranty period resets after a repair or replacement) or non-renewing (the warranty terms do not change during the warranty period) [1]. For costly maintainable systems, the warranty is a type of maintenance agreement. For example, the standard warranty for a wind turbine usually covers the first two or five years of operations and 100% of the cost of parts and labor to replace failed components. The maintenance agreement from the OEM with the original purchase of the product is regarded as a warranty in this paper. Manufacturers need to estimate the future cost of servicing warranty claims and hold a warranty reserve fund for potential liabilities. Failure to understand the warranty cost can lead to a financial disaster. In practice, warranty cost analysis uses historical warranty data on similar products and data from warranty claims captured at an early stage in the product lifetime. However, the historical data are often not representative of products with new technologies and usage conditions, due to uncertainties and interactions between system hardware and software. Early-stage warranty data may not capture pending problems caused by a changing failure rate with time [11]. Other problems with warranty claims data include delays between failure occurrences and the time when
claims are made [12], as well as underestimation as a result of some claims not being documented [11].
maintenance and repair tasks can be performed during lowwind periods of the year to meet the required availability.
Prognostics and health management (PHM) is an enabling discipline consisting of technologies and methods to assess the reliability of a product in its actual life cycle conditions to determine the advent of failures and mitigate system risks [13]. PHM allows systems and products to move away from unscheduled maintenance (reactive repair) after failure or scheduled maintenance based on population statistics to a predictive maintenance approach based on the performance and failure characteristics of each individual system. This allows manufacturers and warrantors to rewrite their warranty policies and contract structures. Prognostics-based warranties can reduce warranty costs, solve problems with warranties, and create new warranty strategies for businesses.
For middle-level maintainable and repairable products, such as cars and household appliances, prognostics-based warranties allow the customer to schedule a warranty service prior to a catastrophic failure to mitigate the system degradation and assure customers of safety and convenience by eliminating sudden product breakdowns. In this way, a prognostics-based warranty is able to increase customer satisfaction and promote customer loyalty.
Section 2 of this paper describes the proactive and predictive warranty paradigm that PHM enables. Section 3 presents three prognostics-based warranty approaches. In Section 4, a case study is provided for warranty validation and user abuse detection based on health monitoring. Section 5 discusses potential challenges to implementing prognosticsbased warranties and provides recommendations for overcoming those challenges. Section 6 concludes the paper. II.
PROACTIVE AND PREDICTIVE WARRANTY PARADIGM BASED ON PHM
Most warranty service is performed after the occurrence of field failures. PHM can shift the warranty philosophy from a reactive to a proactive and predictive warranty paradigm. With the implementation of PHM, warranty claims and service can be made and conducted when anomalies are detected (proactive process), or when future (but not imminent) failures are predicted (predictive process). The proactive warranty paradigm uses real-time interpretation and feedback during operation to control system degradation and avoid losses in system functionality. The predictive warranty process estimates the remaining useful life (RUL) to conduct maintenance and logistics planning for a system or a fleet to reach optimized warranty decisions. For costly maintainable systems, such as aircrafts and production lines, a proactive warranty can be achieved by conducting proactive maintenance when anomalies are detected, instead of reactive repair after catastrophic failures. Anomaly detection provides warrantors with an opportunity to conduct repair tasks based on real-time health status to prevent further system deterioration, field failures, and system downtime. It has been estimated that $35 billion could be saved per year in the U.S. alone if maintenance based on real-time health assessment were employed [14]. A predictive warranty paradigm can use RUL to prioritize maintenance resources and plan maintenance schedules, preventing both unscheduled maintenance and unnecessary time-based scheduled maintenance, and helping obtain the maximum availability from a system or a fleet. For example, a wind farm can conduct maintenance on multiple turbines at one time based on the predicted RUL in order to reduce the cost of cranes and other equipment per repair. Additionally, most
Prognostics-based warranties also allow companies to make active decisions for warranty returns. In the case of a warranty claim, a manufacturer often has the option of either replacing the failed item with a new one or repairing it by means of perfect (overhaul), imperfect, or minimal repair [15][16]. Different warranty services result in different costs and hazard rates of future failures. Replacement enables a product to be in “as good as new” condition. Minimal repair makes insignificant improvements, and the condition after the repair is “as bad as old;” it does not change the hazard rate of the system. Major repair restores a substantial portion of a product, and the failure rate falls somewhere in between “as good as new” and “as bad as old.” Based on the estimation of system health status and prediction of RUL, the warrantor can choose an optimum warranty service action, achieving active health management of the system and reducing the warranty cost. For example, if a part in a car engine is predicted to fail in two weeks, the dealer may replace it with a new one to avoid a repetitive failure of the same part within the warranty period. If the vehicle is running at the end of its warranty period, the dealer may repair the item or replace it with a used part and attempt to make it survive through the warranty period. If PHM is incorporated into a non-maintainable product that is usually replaced after failure (e.g., a cell phone or a tire), a customer can prepare prior to the occurrence of failure to reduce losses from the product breaking down and reduce the waiting time for a replacement item. The preparation may include contacting the manufacturer for warranty return, backing up data from the device, and invoking other resources to temporarily replace the product, if necessary. To the manufacturer, the early information about product performance will allow them to prepare warranty service resources (e.g., items to replace the returned product), improve new product design, and make further business decisions. The cost of the PHM system could be a concern for manufacturers regarding prognostics-based warranties. A prognostics-based warranty can use the data from sensors and the computing power that already exists in many products to obtain the health condition and RUL for warranty management. For instance, a mid-level car has more than 50 sensors, and luxury cars often have more than 100 sensors [17]. A new car may have as many as 50 separate microprocessors, and some luxury cars have up to 90 microprocessors [18]. There is little need to change the physical structure of the system or install extra hardware to apply lifetime warranties.
III.
PROGNOSTICS-BASED WARRANTY APPROACHES
In this section, we present three prognostics-based warranty approaches that are more marketable and/or less expensive for manufacturers and warranty providers than traditional warranties. The mathematical models for these warranty approaches will be developed in future work. A. Part-Based Warranty Return In a warranty service, the cost can be unnecessarily increased because the technicians take a long time to identify the faulty part, do not replace the right part, or do not repair it correctly the first time. These factors can cause potentially small claims to become more complicated. Based on a survey of 98 participants from automotive OEMs and part suppliers, in the course of warranty repair, technicians often repair more than is necessary – they replace more than one part because it's easier to do [19]. As Fig. 1 details, 49% of the respondents said the average claim only replaced one part. Another 21% said that the typical claim included two replaced parts on average. The remaining 30% said that they replaced three, four, five, or even more parts per claim.
PHM collects real-time environmental and operational data, which allow technicians to understand field usage conditions. Therefore, it is more likely to duplicate the field failure by reproducing the usage conditions in diagnostic testing. Moreover, it is possible, through data-driven PHM approaches, to notice sudden changes in monitoring parameters [22] and provide online fault detection and failure prediction. These will reduce NFFs in warranty returns. When a fault is detected or a failure is observed, the problematic parameters and parts can be located and isolated from the system. Then, efforts can be focused on the problematic parts for diagnostics. PHM will permit the examination of root causes of failure by combining performance data, load history, and physics of failure (PoF) models. The implementation of PHM into a product will reduce diagnostic testing efforts and accelerate the understanding of failures and root causes.
Figure 1. Number of parts changed in one warranty claim in the automotive industry (source [19]).
One reason that a technician might replace more than one part is that a failure is classified as no fault found (NFF). NFF occurs when a failure is reported but cannot be verified, replicated at will, or attributed to a specific failure site, mode, and mechanism [20][21]. Devices that are reported as NFF during initial troubleshooting attempts are often returned with the same NFF symptoms or a permanent mode of failure. Technicians tend to replace everything that may have caused the problem in order to avoid sequential failures and save testing time; however, this often leads to good units being removed. The NFF behaviors could be due to intermittent failures, or parts that do not test well outside the system – once the part has been removed from the system, it no longer interacts in the same way with all the other parts in the system. With the implementation of PHM, a part-based warranty return method can be used to eliminate NFFs in warranty returns and expedite the root cause analysis of failures. The process of part-based warranty return is shown in Fig. 2.
Figure 2. Process of part-based warranty return.
Another benefit of the part-based warranty return is to prevent a system from degrading when an anomaly is detected. With PHM, an alarm can be triggered that will be used to shut down the system for warranty service. An anomaly can be a failing part or component or an unexpected environmental or operational condition. By shutting down the system, only the defective parts or components will need to be repaired or replaced under warranty; the rest of the system will be protected and system degradation will be minimized. The operator can choose whether to automatically or manually close down the system when an anomaly is identified. For example, if PHM detects that the weight of laundry is greater than the expected operational load of a washing machine, the machine may automatically refuse to run in order to prevent the system from degrading due to over-
stress. In the case of an essential system, such as an airplane, the operator may take manual actions, since a catastrophic accident might occur if the plane’s operating system were to shut down automatically without notifying the operator. Part-based warranty return data indicate specific defective parts and unreliable components. This information can be used to determine the responsibilities of suppliers. Based on the defective rate of a part, the manufacturer and the corresponding supplier can assume shared responsibility for the warranty costs. Over time, the PHM data provide information that can help a manufacturer evaluate its suppliers’ reliability capabilities. A capable supplier offers parts or components with fewer defects, and hence, the manufacturer’s product will involve fewer warranty claims and less warranty costs. B. Lifetime Warranty Most warrantors cannot bear the financial burden of longterm warranties. However, customers expect a long warranty and even warranty throughout a product’s useful life. A product carrying a longer warranty period provides better assurance to customers. Ford sees longer warranties as a competitive advantage for the company, and believes that customers look at warranties when they make purchasing decisions [23]. Incorporating PHM into products will make manufacturers confident about prolonging their warranty periods due to the proactive and predictive warranty service. PHM has the potential to enable a lifetime warranty to assure customers, differentiate a manufacturer from its competitors, and minimize warranty costs at the same time. With the implementation of such a warranty, the manufacturer will be able to provide customers with a near zero failure system. “Lifetime” is defined in different ways by warranty providers and researchers. For example, General Motors regards “lifetime” as 100,000 miles in its Lifetime Warranty on Opel and Vauxhall cars in Europe [24]. Chrysler’s Lifetime Powertrain Limited Warranty expires when the original purchaser sells the vehicle [25]. Chattopadhyay and Rahman [26] described different lifetime definitions in their study, including definitions based on technical/physical life, technological life, commercial/economic life, and ownership/social life. Manufacturers can choose a lifetime definition based on their product design and customer expectations. For example, customers may expect a laptop computer to last for at least 3 years. If the lifetime is defined as a system’s physical life, the end of the lifetime is not obvious. For example, when a vehicle breaks down, it can almost always be restored by repairing or replacing some parts or sub-systems. The health and performance status from PHM can be used to determine the physical life of a product. For instance, the end of physical life of a vehicle can be defined as when the following three conditions are met: 1) the vehicle is unreliable and unsafe to drive; 2) major rehabilitation is required to restore the vehicle’s reliability; and 3) the cost to restore the vehicle is uneconomic. An uneconomic condition is defined as follows:
rehabilitation cost > × vehicle salvage value
where is a positive number based on the production cost and marketing price, and vehicle salvage value is the residual value of the unreliable car, including the value of good parts that can be reused for other cars and the value of materials that can be remanufactured and recycled. When the rehabilitation cost of a vehicle is larger than the vehicle salvage value multiplied by the factor , the vehicle’s lifetime ends. Fig. 3 shows the process of implementing a prognosticsbased lifetime warranty. The PHM system monitors the health status of the product continuously. When an environmental or operational load exceeds a specified threshold, the user is warned to change the operating condition. If the system or a component is degrading or failing, warranty service will be scheduled based on the RUL estimation. When the system is within its lifetime, the technician can pull out the PHM data, conduct a thorough failure progression analysis, and determine the root causes. If necessary, diagnostics can be done on the faulty structure. The result of data analysis and diagnostics will provide maintenance recommendations for the technician to repair or replace faulty parts. Thus, failures will be under control with the prognostics-based warranty. PHM system
Health Monitoring
PHM system
Alarm Warning
Owner/ operator
Scheduling Warranty Service
End of life
Changing Operation Conditions
Yes
Operator
End of Warranty
No
Technician
Analyzing PHM Data and Providing Necessary Diagnostics on the Faulty Structure
Technician
Repair or Replacement Failure UnderControl System
Figure 3. The process of implementing lifetime warranty based on PHM.
The prognostics-based lifetime warranty also enables the OEM to have access to the usage data of the system over its entire lifetime, and therefore, the manufacturer has knowledge of the failure modes, mechanisms, and causes for the system throughout the lifetime. Based on this information, the manufacturer can improve its design and manufacturing processes to make the system more reliable. With the usage data and lifetime warranty data, the manufacturer will also be able to better understand what customers care about and how they behave. For example, vehicle speed sensor data can show how drivers accelerate or brake, and sensors can be designed to measure the torque that drivers exert on the steering wheel when making a turn. The customer behavior data and system
reliability data can be correlated to demography and gender information to better satisfy the manufacturer’s diverse clientele by designing and providing features for different groups. This will help the OEM retain customer loyalty, increasing sales and market share. A concern for manufacturers regarding lifetime warranties could be the cost for the long-term coverage. As discussed in the previous sections, with the implementation of PHM, there are many ways in which warranty costs can be minimized. For example, a failing component can be changed when it is worn beyond the manufacturer’s recommended tolerance, to prevent other parts or the whole system from deteriorating. PHM’s indication of a fault location will decrease diagnostics costs by reducing the amount of test equipment and labor hours. PHM is capable of eliminating NFFs and preventing good parts from being changed. Time-based scheduled maintenance and inspection can be extended based on RUL prediction. The logistics cost for warranty service will be reduced with PHM. By determining the responsibility of part suppliers, some parts can be guaranteed by the suppliers under the lifetime warranty. C. Customized Extended Warranty Manufacturers, insurance companies, and third party administrators are providing consumers with extended warranties for a prolonged period at an extra cost to reduce consumers’ risk of product failures after the base warranty coverage. For example, the designed life of a wind turbine is 20 years; the maintenance cost will be a large expenditure to the owner after the OEM’s two- or five- year warranty. According to the U.S. Department of Transportation, the average life of a vehicle is over 13 years (145,000 miles) [27], but auto manufacturers offer much shorter warranties than that. Thus, there is a huge market for extended warranty providers. According to Warranty Week’s estimation, the revenue of 109 extended vehicle warranty providers in the U.S. in 2010 was over $11 billion at the contract sales level [28]. The pricing of an extended warranty depends on the original price of the product and the period and coverage of the service contract. However, different products have different levels of reliability when they exit the manufacturer’s base warranty, which will result in different costs to the extended warranty provider. An alternative strategy to the unified pricing method is to price the extended warranty according to the health status and RUL of the product through PHM tools. If the product is well maintained and in good condition, the customer will spend less for an extended warranty; low price is good motivation for an extended warranty purchase. If the product has been used extensively (e.g., a car with a mileage of 150,000 miles at the end of its five-year OEM warranty), it is reasonable for the contractor to charge more for offering an extended warranty. This will reduce the risk for both parties. At the end of the base warranty period, a reliability report can be generated through the PHM system. The reliability status of the products can be categorized into several levels. Based on the reliability levels, the service provider is able to estimate warranty costs for different policies and periods. For example, the reliability status of a product is categorized into six levels, zero through five. Level 5 indicates excellent
reliability of the product, while Level 1 represents the worst condition where an extended warranty is still worthwhile. If the product condition is even worse, at level 0, an extended warranty will not be profitable to the provider (or not worthwhile to the customer). The warranty provider may refuse to offer a service contract. The reliability evaluation is used when the service provider and the customer negotiate the price of the extended warranty. The warranty provider can then offer customized extended warranties where each user’s warranty is unique. Customers can choose different warranty periods, for example, from one year to five years for a car at the end of a five-year base warranty, and coverage of different parts, such as the car motor, electronics, or body. The extended warranty provider can make tailored warranty terms for different customers. As different customers may need different warranty coverage for their products, the customized warranty will interest more customers and expand the market for extended warranties. This customization strategy may be applied to base warranties as well. Based on industry practices, the product characteristics, and the requirements of the customer, the warranty policy can be FRW, PRW, or a combination of both, and the warranty period and coverage can be the result of a negotiation between the manufacturer and the customer. Considering that different customers use the same product differently, a customized warranty can be based on the accumulated environmental and operational stresses in the life cycle, rather than age (measured in time) or usage (such as mileage for vehicles), as in conventional warranties. For example, the warranty of an electric vehicle (EV) battery can expire when the thermal, electrical, and chemical stress accumulation exceeds the specified value; the stress accumulation can be acquired through health monitoring. The customized warranty will require new methods and models to predict accurate warranty costs for product pricing. IV.
CASE STUDY: WARRANTY VALIDATION BASED ON HEALTH MORNITORING
Warranty validation and consumer abuse detection are important issues in warranties. Most warranty policies explicitly exclude failures that are caused by consumers, whether intentionally or not. However, it is not always easy to determine if a failure has been induced by a consumer. This aspect of warranty returns is often costly for manufacturers. By tracking the usage profile data during the life cycle of a product, a company can verify the field usage conditions and identify the cause of the failure for returned products. PHM is capable of detecting customer-induced failures that should be excluded from warranty coverage. In this way, prognosticsbased warranty can directly reduce warranty costs. The case study presented in this section demonstrates the potential opportunities for PHM in warranty validation. A research study by SquareTrade, a third-party warranty provider for consumer electronics and appliances, found that 31% of iPhones failed in the first 22 months (over 10,000 iPhones, including iPhone and iPhone 3G models, were examined) [29]. The authors of this study determined that only one-third of these failures were due to malfunctions from
normal use, and the other two-thirds were caused by user abuse or accidental damage. Water damage accounted for over 25% of the customer-induced damage cases. According to Apple’s quarterly reports, it sold approximately 318 million iPhones between April 2007 and December 2012. If the failure rate during this time period is consistent with that in the SquareTrade study, almost 16.5 million iPhones sold have been or will be damaged by water. When iPhones are damaged by water or liquids containing water, users often return their phones to their local Apple retailer to seek repair or replacement under warranty. Apple has stated that the personnel receiving these problematic devices are often unqualified or untrained to determine whether the devices have failed because of manufacturing defects or consumer abuse [30]. To help its personnel identify water damage, Apple has installed liquid contact indicators (LCIs) in its iPhones. Each iPhone is equipped with an LCI at the rear of the headphone jack, the iPhone 3G/3GS and iPhone 4/4S models also have an LCI at the rear of the dock-connector housing, and iPhone 5 has an LCI located on the sidewall of the dock-connector housing [31]. When water comes into contact with the edge of LCI, it changes color permanently from white to red. Apple refuses to honor warranties on products with triggered LCIs, assuming that the customer has caused the problem. This policy applies to both the one-year limited warranty and the AppleCare Protection Plan. Apple expects the LCI to give a quick indication of whether a device has suffered from water damage to prevent costly replacements due to customer abuse. However, some iPhone customers have claimed that the external LCI can be triggered by normal usage conditions, such as in high-humidity and condensation environments in locations such as Singapore and Hong Kong [32], and during activities such as exercise [33][34]. Apple is facing two class-action lawsuits in the United States for its use of external sensors to deny warranty service without opening the phone to determine the actual causes of the failure of the phone [35][36]. In South Korea, a 13-year-old girl sued Apple for refusing to service her iPhone based on a triggered LCI indicator [37]. In response to the criticism and lawsuits, Apple has made a change to its water damage policy for iPhones and similar products. The new policy allows the customer to request further internal inspection of the phone to verify if the device is eligible for warranty service [38][39]. Detecting customer-induced product failures is desirable for Apple and other companies so that they can exclude costly outof-warranty claims. However, as the LCI can only provide a binary (yes/no) indication without further information, it is possible that an LCI could give a false alarm (false positive error, where the LCI turns red without water damage). A prognostics-based warranty could alleviate this problem by providing health monitoring data to allow root cause analysis for identifying whether the failure is under warranty. For example, a humidity sensor inside a cell phone could monitor the humidity of the mother board in situ. If the cell phone is designed to easily access the health monitoring data, by using a USB cable or other device, the warrantor will understand the humidity change over time in the mother board of the cell phone, and be able to determine if the phone experienced a
long period of high humidity environment or an intense water submerge. In-situ monitoring of environmental and operational loads can also provide warning of overstress and alert the user of improper usage conditions. Through this warning, PHM can prevent abusive behavior. For instance, if a PHM system indicates that the moisture level is too high for a device, the user might wipe the condensation off of the surface of the item or shut it down to avoid potential water damage. Prognosticsbased warranty will be a credible way to detect and prevent user abuse. V.
DISCUSSION
PHM offers solutions for implementing proactive and predictive warranty service at a lower cost, achieving better warranty strategies and optimum decisions, and reducing system downtime and losses, compared to traditional warranty strategies. At the same time, we need to understand the possible challenges that prognostics-based warranties may face and try to overcome them when incorporating PHM into warranties. The cost of the PHM system itself and the confidence level of the fault detection and RUL prediction are the first challenges to the manufacturers. The manufacturers need to estimate the cost of PHM and the savings that the prognosticsbased warranty and other PHM applications, such as conditionbased maintenance, can bring. When estimating the costs and savings, the confidence level of the PHM technology should be taken into account, as PHM may not detect the failures or predict the RUL with 100% accuracy. Manufacturers need to make optimized decisions when implementing PHM-based warranties, where ROI analysis [40] can be used to verify the economic benefits. Problems may arise if PHM data are only known by the customers and not by the manufacturer or supplier. Customers may not be able to comprehend the data or the value offered by the PHM data. If the PHM system is too complicated for customers to understand (and for marketers to market), the prognostics-based warranty will not work. Some customers might also abuse the PHM system for personal gain. For example, if an mp3 player can be fixed based on early anomaly detection from PHM or replaced after the occurrence of failure, the user may take no action with the PHM alert and let it fail in order to get a new device from the manufacturer. Therefore, a prognostics-based warranty should make a clear connection between indicators and potential failures, find a method for the manufacturer to collect the health data, and require a customer to contact the warrantor with early degradation information in the policy. Data privacy and security is also an issue for prognosticsbased warranties when product usage data are collected by manufacturers from customers. The data should be stored in an encrypted form to avoid abuse from a hacker or a competitor. Warrantors may not collect the usage data to achieve their own commercial purposes without the consumers’ permission, as this would violate the law. Warrantors need to be aware of legal issues when obtaining and analyzing customers’ PHM data.
Another potential challenge is that customers may report potential failures based on PHM prediction prior to the end of the stipulated warranty period. Customers would expect the warrantor to perform necessary maintenance before the warranty period ends in order to reduce their own maintenance expenses after the warranty coverage. With the PHM program, customers have a better chance of finding an indicator to claim under-warranty maintenance. To avoid potential problems, the warranty policies should make clear the rights and obligations of both parties. Moreover, an appropriate examination and reliability evaluation at the end of the base warranty can be used to start an extended warranty, where the manufacturer can make a profit. The end-warranty evaluation and possible maintenance is also a way to maintain good customer relationships. VI.
CONCLUSIONS
While the cost of warranties is a heavy burden for many manufacturers, manufacturers are driven to offer attractive warranties to differentiate themselves in the intense global market. PHM provides opportunites for competitive warranty approaches with reduced warranty costs, by enabling a proactive and predictive warranty paradigm that improves the warranty decision-making process. The in-situ health monitoring data can be used to identify defective parts and unreliable components, which enable a partbased warranty return. Moreover, long-term part-based warranty return data can be used to evaluate and select suppliers. PHM will permit companies to offer longer warranties at lower prices, which can further enable lifetime warranties with failures under control. PHM enables customized extended warranties based on reliability evaluation and RUL prediction for individual systems. The customized extended warranty will be flexible for different products and customer needs. Warranty validation based on health monitoring data and through root-cause identification distinguishes user-induced failures from failures due to product defects. This directly reduces warranty costs by excluding outof-coverage warranty claims. Prognostics-based warranties will result in not only warranty cost reduction, but also new business models. Prognostics-based warranties can create value beyond monetary gain, in terms of customer relationship management, competitiveness, and next-generation product design. ACKNOWLEDGMENT The authors would like to thank the more than 100 companies and organizations that support research activities at the Center for Advanced Life Cycle Engineering at the University of Maryland annually, specifically the CALCE Prognostics and Health Management Group. Also, the authors would like to thank Kelly Smith and Mark Zimmerman for copyediting.
[2] [3]
[4]
[5]
[6]
[7]
[8] [9] [10]
[11]
[12] [13] [14] [15]
[16]
[17]
[18]
[19]
[20]
[21]
[22]
[23]
REFERENCES [1]
D. N. P. Murthy and I. Djamaludin, “New product warranty: a literature review,” International Journal of Production Economics, vol 73, pp. 231260, October 2002.
[24]
W. R. Blischke and D. N. P. Murthy, Product Warranty Handbook. NY: Marcel Dekker, Inc., 1996. B. H. Franklin, Product Warranties & Servicing: Responsive Business Approaches to Consumer Needs. Washington, DC: U.S. Department of Commerce, 1992. B. Choi and J. Ishii, “Consumer perception of warranty as signal of quality: an empirical study of powertrain warranties,” Working Paper, Amherst College, September 2009. T. Cain, “Hyundai usa market share growth 1999 – 2010”. January 2011. [Cited: 2013.01.27]; Available from: http://www.goodcarbadcar.net/2011/01/hyundai-usa-market-sharegrowth-1999.html. “Top 100 warranty providers of 2009,” WarrantyWeek, April 2010. [cited 2013.01.27]; Available from: http://www.warrantyweek.com/archive/ww20100401.html. E. Arnum, “Update on the warranty industry, report of Warranty Week,” The Seventh Annual Warranty Chain Management Conference, San Deigo, CA, March 15-17, March 2011. E. Arnum, Warranty Week, May 2007. W. R. Blischke and D. N. P. Murthy, Warranty Cost Analysis. NY: Marcel Dekker, Inc., 1994. D. N. P. Murthy and W. R. Blischke, “Strategic warranty management: a life-cycle approach,” IEEE Transactions on Engineering Management, vol. 47, pp. 40-54, February 2000. M. Pecht, “Establishing a relationship between warranty and reliability,” IEEE Transactions on Electronics Packaging Manufacturing, vol 29, pp. 184-190, July 2006. D. N. P. Murthy, “Product warranty and reliability,” Annals of Operations Research, vol. 143, pp. 133-146, March 2006. S. Cheng, M. Azarian, and M. Pecht, “Sensor systems for prognostics and health management,” Sensors, vol. 10, pp. 5774-5797, 2010. “Approaching zero downtime: The Center For Intelligent Maintenance Systems,” Harbor Research Pervasive Internet Report, 2003. A. Rahman and G. Chattopadhyay, “Lifetime warranty policies: complexities in modelling and potential for industrial application,” the Fifth Asian Pacific Industrial Engineering and Management Systems Conference, Gold Coast, Australia, 2004. G. Chattopadhyay and A. Rahman, “Modelling environmental and human factors in maintenance of high volume infrastructure components,” The Third Asian Pacific Conference on System Integrity and Maintenance, Hilton, Cairns, Australia, 2002. S. Krueger, R. Müller-Fiedler, S. Finkbeiner, and H.Trah, “Microsystems for automotive industry,” January 2007. [Cited: 2013.01.27]; Available from: http://www.mstonline.de/microsystems/industrial_applications_I/automo tive. B. L. Capehart And L. C. Capehart, “Facility energy efficiency and controls: automobile technology applications,” in Encyclopedia of Energy Engineering and Technology, Edited by B. L. Capehart. CRC Press. Lilburn, GA, U.S.A. Vol 2, pp. 671-678, 2007. J. Barkai, “Shadow drivers of warranty cost,” Warranty Week, June 13, 2006. [cited 2013.01.27]; Available from: http://www.warrantyweek.com/archive/ww20060613.html. D. A. Thomas, K. Ayers, and M. Pecht, “The ‘trouble not identified’ phenomenenon in automotive electronics,” Microelectronics Reliablilty, vol. 42, pp. 641-651, April-May 2002. H. Qi, S. Ganesan, and M. Pecht, “No-fault-found and intermittent failures in electronic products,” Microelectronics Reliablilty, vol. 48, pp. 663-674, 2008. M. Pecht, “A prognostics and health management roadmap for information and electronics-rich systems,” IEICE Fundamentals Review, vol. 3, pp. 25-32, 2010. “Ford's powertrain warranties,” WarrantyWeek, July 2006. [cited 2013.01.27]; Available from: http://www.warrantyweek.com/archive/ww20060725.html. J. Reed, “GM woos Europe with ‘lifetime warranty’,” August 2010. [Cited: 2013.01.27]; Available from:
[25]
[26]
[27]
[28]
[29]
[30]
[31]
[32]
http://www.ft.com/intl/cms/s/0/e4caae36-a0b7-11df-badd00144feabdc0.html#axzz1frFiykWk. “Lifetime powertrain limited warranty,” Chrysler Group LLC. [Cited: 2013.01.27]; Available from: http://www.dodge.com/crossbrand/warranty/pdf/X_lifetime_warranty.pd f. G. Chattopadhyay and A. Rahman, “Development of lifetime warranty policies and models for estimating costs,” Reliability Engineering & System Safety, vol. 93, pp. 522-529, 2006. S. Nakate, “Car life expectancy,” June 2010. [Cited: 2013.01.27]; Available from: http://www.buzzle.com/articles/car-lifeexpectancy.html. “Vehicle service contract administrators,” Warranty Week, September 2010. [cited 2013.01.27]; Available from: http://www.warrantyweek.com/archive/ww20100909.html. A. Sands and V. Tseng, “One-third of iphones fail over 2 years, mostly from accidents,” SquareTrade Research, 2009. [cited 2013.01.27]; Available from: http://www.squaretrade.com/htm/pdf/SquareTrade_iPhone_Study_0609. pdf. T. M. Johnson, R. H. M. Dinh, and T.Y. Tan, “Consumer abuse detection system and method,” Pending Patent. August 2009. [cited 2013.01.27]; Available from: http://www.faqs.org/patents/app/20090195394. “iPhone and iPod: liquid damage is not covered by warranty,” AppleSupport, October 18, 2012. [cited 2013.04.02]; Available from: http://support.apple.com/kb/ht3302. I. J. Liu, “Is Hong Kong so humid it could wreck your iphone?” South China Morning Post, August 2010.
[33] D. Martin, “Sweaty workouts killing iphones?” April 2009. [cited 2013.01.27]; Available from: http://reviews.cnet.com/8301-19512_710214153-233.html. [34] M. Klurfeld, “Apple iPhone abuse detection sensors: who is abusing whom?” 2009. [cited 2013.01.27]; Available from: http://techgeist.net/2009/09/apple-iphone-abuse-detectionsensorsabusing-2/. [35] Class action, “Complanint for equitable relief and damages,” Charlene Gallion, on behalf of herself and all other similarly situated, Plaintiff v. Apple, Inc., a California coporation, and DOES 1-100, inclusive, Defendants, Case: 101610. United States District Court, Northern District of California, Filed April 15, 2010. [36] M. Dinzeo, “iPhone warranty voided on pretext, class says,” 2010. [cited 2013.01.27]; Available from: http://www.courthousenews.com/2010/04/16/26463.htm. [37] “13-year-old south korean girl sues apple for not paying iphone repair costs,” BNO News, 2010. [cited 2013.01.27]; Available from: http://wireupdate.com/wires/11426/13-year-old-south-korean-girl-suesapple-for-not-paying-iphone-repair-costs-2/. [38] Z. Epstein, “Apple amends internal ipod water damage policy,” Boy Genius Report, November 9, 2010. [cited 2013.01.27]; Available from: http://www.bgr.com/2010/11/09/apple-amends-internal-ipod-waterdamage-policy/. [39] C. Warren, “Apple changes policy on water-damaged ipods,” January 13, 2011. [cited 2013.01.27]; Available from: http://mashable.com/2011/01/31/apple-water-damage-policy/. [40] K. Feldman, T. Jazouli, and P. Sandborn, “A methodology for determining the return on investment associated with prognostics and health management,” IEEE Transaction on Reliability, vol. 58, pp. 305316, June 2009.
