Bernd Bertsche

Multiple failure mode and effects analysis-an approach to risk assessment of multiple failures with FMEA

By integrating the procedures of the FMEA method (failure mode and effects analysis) and the FTA method (fault tree analysis) into a combined procedure, the mFMEA (multiple failure mode and effects analysis), an inclusive reliability analysis of complex, mechatronical systems, is made possible. The advantages of both methods are brought together and integrated into a new procedure. The FMEA along with its risk analysis, risk assessment and measure controlling, paralleling systematically to the product design cycle is only applicable for single failures. The FTA expands the new procedure through a failure analysis with the help of its combination option to network failures according to Boolean logic. The new procedure allows for the consideration of multiple failures while retaining all the characteristics of FMEA. In addition, through the use of failure networks quantitative information can be derived concerning the systems availability, which delivers important information along with the results of the mFMEA.

A new approach to gathering failure behavior information about mechanical components based on expert knowledge

This paper presents a possibility with which the reliability knowledge of maintenance employees can be used in order to receive reliability data that can be used to do simulations or further calculations. Using the presented methodology opens up a big source of information. The use of employee knowledge entails the processing of imprecise data, as the knowledge is not given in the form of a failure time plot, but in the form of verbal expressions and specific information. The paper deals with the data format and precision of expert-based information. A methodology is presented which enables the transformation of certain employee (further called experts) information into reliability data. Apart from that the paper also shows the influence of imprecision of expert information and how to handle this imprecision in order to get applicable results. As a finishing subject the paper also provides a possibility with which the deviations of expert information from reality can be estimated dependent on the expert statement itself. With that approach one can tax the trustability of expert information.

Conjoint Modelling with Extended Coloured Stochastic Petri Net and Reliability Block Diagram for System Analysis

The main objective of this paper is to demonstrate the modelling power of a conjoint modelling approach, which incorporates a new developed extended coloured stochastic Petri net (ECSPN) and a reliability block diagram. The approach combines the advantages of both modelling regimes in order to keep the system model traceable. It provides a comprehensive modelling methodology to describe the logical, chronological, and dependency aspects of the system reliability, maintenance, and logistics. The ECSPN provides powerful extensions to incorporate the component age as well as queuing aspects into the system model.

Models to consider load-sharing in reliability calculation and simulation of systems consisting of mechanical components

The load-sharing case occurs in systems, if a load is shared by several components. Then the failure of a component results in a higher load share for the surviving components. This paper places emphasis on the analytic description of systems with load-sharing, the algebraic calculation and the simulative solution of their reliability. The capacity flow model, the Freund model and the state-graph method are presented as models for the analytic description. The state-graph method is the most general model: systems with individual failure behavior of the components, individual load steps as well as complex system structures can be considered. All three models are restricted to components with constant failure rates in the corresponding load level. However, in most mechanical systems the failure rates of components are not constant due to aging and wear-out. A more adequate method for the load-sharing case in mechanical systems is the application of simulation techniques. In this paper the general simulation algorithms are presented for 1-out-of-n systems consisting of components with constant failure rates and of components with time-dependent failure rates. In the case of constant failure rates, the failure times in the load levels can be sampled directly from the failure distributions due to the memory-less property of the exponential distribution. The simulation results are shown for both a 1-out-of-2 and a 1-out-of-3 system. For verification purpose the results of the simulation are compared with the analytic solution of the Freund model. In the case of time-dependent failure rates a dynamic modification of the failure distribution of the components is necessary. This is done by a time shift of the distributions and transformed random numbers. The simulation results are presented for a 1-out-of-2 and a 1-out-of-3 system. The failure behavior of the mechanical components is described by a Weibull distribution in each load level. In order to verify the simulation results the analytic bounds of the corresponding system with no increased load and maximum load are calculated.

Simulation-based risk management of product-service systems

Products are increasingly accompanied by services such as maintenance or leasing contracts. The share of services of these so-called product-service systems (PSS) is expected to gain. The provider of such PSS-related services needs a comprehensive knowledge of the product properties and service requirements - particularly the product reliability and availability as well as necessary service resources - in order to minimize its financial risk. Hence, powerful methods need to be applied to the risk management of these services. The reliability and maintenance simulation offers a great potential for the risk management of PSS-related services. The presented powerful modeling methods are able to consider various product and service parameters which are compared in an overview. The conjoint system model yields the maximum modeling power and is able to consider manifold dependencies and reciprocal effects. The Monte Carlo simulation is the established method for the analysis of the operational parameters. A wide range of applications can be covered, considering the complexity as well as the field of application. The analyzed operational parameters are an integral base for the risk management of PSS-related services as they enable to evaluate the financial risks. The application of fundamental operational parameters to the risk evaluation of several exemplary PSS-related services is summarized. It is explained how the risk management process is integrated into the different stages of the PSS life cycle. Several risk measures which are common in finance are applied to the risk evaluation. The capability of the method is demonstrated by an example: The risk management of a maintenance contract for a production line with guaranteed maximum maintenance time including a bonus malus system.

Application of ECSPN to RAMS modeling and analysis of hybrid drive systems

In our paper we present a modeling and simulation approach based on a class of Extended Colored Stochastic Petri Nets and a Monte Carlo Simulation to analyze and optimize hybrid car systems. The approach allows the modeling of various functional dependencies between the components of a hybrid car drive system and the integration of lifetime models. The Monte Carlo simulation approach allows the analysis and optimization of very detailed system characteristics. We demonstrate the abilities of our approach with a simulation study of a power split strong hybrid car drive system focusing on the occurrence of comfort and safety relevant events.

Decision support in early development phases—A case study from machine engineering

In the case study presented in this paper we consider early development phases of a mechanical product. We want to evaluate different concepts and decide which one(s) to pursue. A problem in early phases is that usually no test runs are available. In our case study, based on a standard, there are ways to compute the lifetime distributions of the components of the different concepts. Some parameters needed for these computations are not known precisely. Unfortunately, the lifetime distributions of the components are highly sensitive to these parameters. Our approach is to equip these parameters with distributions. These distributions would be called prior distributions in Bayesian terminology, but no update is possible since no test runs are available. Our approach implies that the distribution of the system lifetime for each concept is random, i.e. we get random elements in the space of lifetime distributions. Using Monte-Carlo simulations, we demonstrate several ways to compare the random lifetime distributions of the concepts. Some of these comparisons use stochastic orderings. We also introduce a new stochastic ordering which is particularly suitable for reliability purposes. Our case study, consisting of three scenarios, allows us to demonstrate some conclusions that can be reached.

Case study: influences on the availability of machine tools

For more than three years downtimes, data of a press system were observed. Influences on the availability are shown in the data analysis. Mainly in the pre-production, downtimes are caused by changing the tools and by organizational reasons like no crane for changing the tools. General downtimes of press systems are caused by tryouts of actual and new model tools or overhaul. During the production time the machine is not working due to repair, maintenance or organizational reasons. In the case study further effects on the availability are determined. One can see that downtimes of the press systems depend on the production month and on the weekday respectively. The causes of the downtimes cannot be analyzed exactly because only failure categories are given. In order to analyze the repair and maintenance behavior of the machine exact definitions of the downtimes have to be given for the manufacturer and the user of the press system. A simple method for the prediction of the availability in future for the user and manufacturer is given by the simple exponential smoothing. This method also includes seasonal variations caused by vacation employees, company holiday or maintenance of the system during vacations. The example points out that the simple exponential smoothing can be used for forecasting the availability.

SYSLEB-a tool for the calculation of the system reliability from raw failure data

In this paper, several methods are presented for the calculation of the system reliability also considering preventive maintenance actions. Additionally, the authors present the concept of importance, which allows to determine the influence of a component failure on the system failure. In a case study, the three steps for the system reliability analysis of a single-step gearing are shown. It is demonstrated how to receive the reliability function of components from failure data or from calculations respectively. Within the calculation, the B/sub 10/ lifetime is determined by means of the damage accumulation hypothesis, especially for the pinion, the gear and the ball bearings. Afterwards a database is used to determine the parameters of the three-parameter Weibull distribution. As a result, the authors received the system reliability of an analyzed gearing. The reliability tool SYSLEB was used for the evaluation of the failure data of the radial shaft seals. Additionally, the reliability functions for the components as well as for the system were determined by applying SYSLEB. It was shown that the system reliability could be increased by improving the pinion and the gear to avoid breakage. For this purpose preventive maintenance actions could also be applied.

Correlation between the failure behavior of automotive components under taxi and field operating conditions

This paper investigates the correlation of the failure behavior between components under different operating conditions, which allows the forecast of the failure behavior of the less stressed component. The authors present the reliability evaluation of a large amount of real, current and practical failure data of an automotive component under two different operating conditions-taxis and field application-and determine the correlation. One result is that the correlation factor depends on the evaluation method. They present the results for the Sudden Death and Candidate methods. In both cases, the lifetime of taxis is smaller than that of regular field application. Neglecting the knowledge of the drive-performance distribution yields identical values for the shape parameter for both operating conditions, a validation for the identical kind of failure in both cases. The correlation factor is amazingly high. However, the Candidate method yields to different shape parameters. This means the correlation cannot be described by a single factor. But, the trend of acceleration could be confirmed. One can draw the following conclusion: there is an amazingly high acceleration of the taxi failure behavior compared to the field. However, one should be careful and cautious with quantitative predicates.