Theory and MethodologyFailure rate distributions for flexible manufacturing systems: An empirical study
Introduction
The literature is replete with examples of how computer controlled manufacturing technologies such as flexible manufacturing systems (FMS) and computer integrated manufacturing (CIM) can be utilized to improve the strategic and competitive positions of firms. Significant improvements in inventory levels, space requirements, lead and cycle times, scrap and yield rates and other quality measures have been reported. In some cases the benefits are truly impressive and border on orders of magnitude improvements.
As more and more factories employ this new technology, the subject of maintenance management is taking on a renewed importance. The failure of a single component can not only idle a very expensive piece of equipment but, due to reduced work-in process, the failure can quickly idle an entire production system. This is compounded in a JIT environment where the flow of finished goods is disrupted, thus delaying customer shipments at the intangible but real cost of customer goodwill.
A second reason for the increased importance of maintenance management when using computer controlled manufacturing technology is the increased flexibility created by these programmable systems. Since an FMS is programmable it is expected to have a life expectancy greater than the single product or part family it was originally intended to produce. Companies are justifying these systems based on a longer expected life and the assumption they will not have future expenditures for replacement equipment (Fotsch, 1985; Meredith and Suresh, 1986). However, this is based on the assumption that the FMS's physical life will be longer than an organization's need for the system. Therefore, maintenance policies capable of keeping these systems from physically deteriorating during their extended useful life will be required. The identification and implementation of such effective policies will enable managers to avoid premature replacement costs, maintain stable production capabilities, and prevent the devaluation of the system and its component parts.
During an FMS's extended useful life it will experience a very different wear and tear history than a traditional machine tool operating during the same time period. It is estimated that an FMS will operate at 80% utilization or higher whereas a traditional machine tool would probably be utilized at only 20% (Meredith, 1988). This will result in the FMS incurring four times the wear during any given time period. It is not well known what the effect of such accelerated usage will be on the system, but it is generally agreed it will significantly increase the importance of maintenance and maintenance related activities.
Another reason for the increased importance of maintenance management for this new technology is the synergistic benefits attributed to these systems. The linking of stand-alone systems into an FMS has created qualitative benefits such as faster response to customer requests, ability to customize products, improved quality, and better production control. These benefits are synergistic and make a significant additional contribution. Consequently, the cost of an isolated failure not only includes the loss of that piece of equipment, but also includes the loss of the significant contributions of synergy. Therefore, it will be important for maintenance management to consider not only the amount of time a machine is down for maintenance but also the timing of when it is down and the resulting synergistic costs.
Finally, these technologies are less reliant on skilled craftsmen for their day to day operations. The skills required for high precision machining have been embedded in the part and operating software, thus enabling the systems to be operated by fewer personnel with less traditional machining experience. However, the elimination of the skilled personnel has also removed a valuable maintenance management resource. The highly skilled machinist not only operated the machine but also continually monitored the machine for component wear or failure. Maintenance managers now realize these operators were performing preventive and minor corrective maintenance as well as reporting potential problems before they became major failures.
The above arguments point out the need for more research aimed at understanding maintenance management for advanced manufacturing technologies such as FMS. An essential element in maintenance management research will be a knowledge of the failure and repair characteristics of FMSs. In particular, how do these machines fail, what are their failure types, what is the frequency of failure, and how can these failures and resultant repairs be characterized? This paper attempts to provide answers to these questions which should help other researchers, especially those using simulation methodology.
The purpose of this paper is twofold: (1) to help researchers and managers understand the complex reality of failures and repairs in a highly integrated, advanced technological system such as an FMS, and (2) to document the distribution types and characteristics (shape and scale parameters plus their confidence intervals) of the various modes of failures and repairs. The complexity of system failures here includes interaction between different modes of failure, misrepresentation of certain failure types (particularly human failures), and the interaction of repairs for one failure type initiating a different type of failure.
Current studies of FMSs typically have to assume most maintenance aspects about the equipment because there is so little published data available concerning failure and repair modes and characteristics. Examples are: assuming only a single failure mode, assuming a particular form of failure distribution (often the negative exponential or gamma), assuming particular values for the distribution characteristics, and assuming a particular form of repair distribution and its characteristics. However, with so many assumptions there is little chance that such studies will be representative of real situations, thereby minimizing the possibility of being useful or having an impact on practice. This is probably why, as discussed further below, there is such a gap between published maintenance theory and practice.
We believe strongly that the identification of the real complexity of system failures and the documentation of their failure and repair characteristics is an important contribution to the field and significantly advances our knowledge of this critical phenomenon. Such knowledge can help later simulation and modeling studies avoid making erroneous assumptions about how these systems operate in practice as well as the characteristics of their failure and repair distributions. As researchers, we must be diligent to not confuse “hypothetical” with “theoretical”, thereby rejecting research based on empirical, real-world data because it intuitively seems to lack a theoretical basis. Recent editorial statements in well-respected journals show that we are indeed currently making this error: “As editors, we are concerned about the number of papers submitted to the journal which are modeling papers with little empirical evidence about the nature of the real problem… In future, we shall reject, without review, all those papers which do not have supporting empirical evidence…” (Hollier et al., 1996, p. 4) and “Papers with empirical support, conducted to … motivate a mathematical model, are particularly encouraged” (Wein, 1996, p. 251).
Before proceeding, it is important to note that this paper does not seek to evaluate the efficacies of various maintenance policies although we comment on this issue in Section 8. The company whose FMS was used for our data collection was not using any preventive maintenance policy; it was simply repairing machines after they failed. Thus, the data obtained here were not biased by the company's use of a specific maintenance policy.
Section snippets
Maintenance characteristics of production systems
A highly integrated manufacturing system such as an FMS is a combination of very complex machinery integrated through an equally complex computer system. Each machine in an FMS is a combination of many sub-assemblies, where each sub-assembly is itself complex and consists of many dissimilar interdependent components. Traditional maintenance policies deal with stand-alone machinery comprised of mechanically similar components with similar failure distributions. However FMS maintenance policies
Data gathering and analysis
Our research problem consisted of gathering and analyzing data to verify and refine the earlier, limited German study for a United States FMS and to determine the failure and repair distributions for an extended group of failure types. Every failure during a five month period was identified and categorized as mechanical, electrical, electronic, hydraulic, human or software. The time between failures for each type as well as the time to repair were recorded and analyzed.
The failure and repair
Results
As stated earlier the six failure types included in the study were: mechanical, hydraulic, electronic, electrical, human, and software. The failure types represent the inherently different ways the individual machines can fail and each failure type presents its own unique challenges for maintenance personnel. For example, mechanical components often gradually wear until they fail while electronic components often show no indication of wear prior to failure. Similarly, the maintenance personnel
Failure distributions
The times between failures for each of the failure types were analyzed with respect to theoretical distributions. The chi-square goodness-of-fit test was used to determine the best theoretical distribution to represent each failure type. The general format of this analysis took the form of the following hypothesis:
Hypothesis 1. The TBF data came from a specified distribution.
This general format was applied to each of the six failure types. For example, the electrical failures resulted in the
Repair distributions
The identified probability distributions and the chi-square goodness-of-fit test results are displayed in Table 4 for the time to repair the systems. The same general tests used for the failure distributions were also applied here. The entries in the table indicate the levels of significant fit for each of the repair distributions. The shape and scale parameters and their 95% confidence intervals are also provided.
Again, the histograms (Fig. 8Fig. 9Fig. 10Fig. 11Fig. 12Fig. 13) generally
Limitations
The mean and coefficient of variation of the TTRs of an organization's machines will depend on the qualifications and availability of the organization's maintenance personnel. The repair distribution types should be consistent across organizations; however, the parameters of those distributions (mean and coefficient of variation) given here should be regarded as typical of systems with skilled maintenance personnel.
It should also be noted that human failures are the most difficult to identify.
Conclusions
This study found that the common simulation models of FMSs under failure are typically much too simplified. As shown here, actual FMSs can fail due to a number of different causes, each with their own type of distribution and parameters. Moreover, complex interactions can occur between these failure types – particularly those involving human failures – as well as between repairs of one type and failures of another.
As regards the failure types and distributions, this research confirms and
References (20)
- et al.
A survey of maintenance models for multi-unit systems
European Journal of Operational Research
(1991) Machine tool justification policies: Their effect on productivity and profitability
Journal of Manufacturing Systems
(1985)On the maintenance concept for a technical system; II. Literature review
Maintenance Management International
(1986)Availability, performance and service of FMS
The FMS Magazine
(1987)- et al.
Editorial
International Journal of Operations and Production Management
(1996) - Kennedy Jr., W.J., 1987. Issues in the maintenance of flexible manufacturing systems. Maintenance Management...
- Kvanli, A.H., Guynes, S.C., Pavur, R.J., 1989. Introduction to Business Statistics, West, New...
- Law, A.M., Kelton, W.D., 1991. Simulation Modeling and Analysis. McGraw-Hill, New...
- et al.
Availability of maintained systems: A state-of-the-art survey
AIIE Transactions
(1977) Maintenance policies for stochastically failing equipment: A survey
Management Science
(1965)
Cited by (58)
The effects of cyber threats on maintenance outsourcing and age replacement policy
2023, Computers in IndustryCitation Excerpt :The lognormal distribution and the generalized Pareto distribution displayed excellent fits to the severity of cyber loss data. Vineyard et al. (1999) investigated real-world data to determine failure and repair rate distributions occurring in the manufacturing systems industry. Their purpose was to aid researchers in conducting simulation studies of manufacturing downtime due to a variety of causes: mechanical, electrical, software, and human.
Reliability analysis of the main drive system of a CNC machine tool including early failures
2021, Reliability Engineering and System SafetyCitation Excerpt :Generally, reliability analysis plays a role in [4-6]: Determine characteristics of systems and their components such as reliability and Mean Time To Failure (MTTF); Ascertain critical reliability affecting factors that likely give rise to malfunctions of systems from either system configuration or data collection perspective; Analyse failure behaviour as a basis to clear failure propagation paths and their relations; Suggest recommendations such as corrective activities to designers, maintenance strategies to operators, and basic information to stakeholders (producers and end-users) to support their decisions. Early failure is a phenomenon that complicated systems suffer more failures at their initial installation period [7, 8]. Alongside the implementation of operational activities e.g. inspection, maintenance, and replacement of elements, early failures of systems will release, until the stable-working stage at which failure rates of systems maintain a relatively low but stable level.
Wearable shear force-sensing for augmenting manual hose connections in an automotive assembly
2021, Procedia ManufacturingMetal additive manufacturing in the commercial aviation industry: A review
2019, Journal of Manufacturing SystemsCitation Excerpt :It is well agreed that integrating assembly parts into one single component can reduce the weight of the final part and optimize the design according to its functionality. Most of the predictive models of spare parts inventory management rely on two key variables: the number of working components into the assembly and their reliability, otherwise known as Mean Time Between Failures (MTBF) [243]. Removing many sub-assembly parts could not necessarily be the optimal solution.
Evaluation of wearable visual assistance system for manual automotive assembly
2019, Procedia Manufacturing
- 1
Tel.: +1 910 678-4617; fax: +1 910 678-4151.
- 2
Tel.: +1 910 334-5666; fax: +1 910 334-5580; e-mail: [email protected].