SOFTWARE TOOL FOR AUTOMATED FAILURE MODES AND EFFECTS ANALYSIS (FMEA) OF HYDRAULIC SYSTEMS

Transcription

1 SOFTWARE TOOL FOR AUTOMATED FAILURE MODES AND EFFECTS ANALYSIS (FMEA) OF HYDRAULIC SYSTEMS Jacek S. Stecki*, Finn Conrad** and Beng Oh*** *Depratjnent of Mechanical Engineering, Monash University Wellington Rd., Clayton, Victoria, Australia 3168 **Department of Mechanical Engineering, Technical University of Denmark Building 358, DK Lyngby, Denmark (finn. dtu. dk) * * * Fluid Power Net Pty Ltd Bon Vue Rd., Melbourne, Victoria 3104 (bengoh@netscape. com) ABSTRACT Offshore, marine, aircraft and other complex engineering systems operate in harsh environmental and operational conditions and must meet stringent requirements of reliability, safety and maintainability. To reduce the high cost of development of new systems in these fields improved design management techniques and a vast array of computer aided techniques are applied during design and testing stages. Failure Modes and Effects Analysis (FMEA) is generally used in the above industries to reduce design risks. At present FMEA cannot be carried out automatically. Although various FMEA software tools exist, they basically provide data management for a manual analysis. It is generally recognised that failure analyses are usually carried out after the system has been designed, and often after it has been manufactured. If any fundamental design flaws or safety risks in the system are identified at this stage then modifications to the hardware are often costly. The software tool for automated FMEA is based on qualitative modelling of generalised information flow (energy, information or mass) in a system, recognising that the primary functions of mechanical, electrical, thermal and/or hydraulic systems (or their combination) as well as modem neuromechanic systems (also called mechatronic systems in the mechanical domain) are the generation, transmission and conversion of energy. The paper explains some underlying principles of software operation and provides a description of the software KEYWORDS Hydraulic, Failure analysis, FMEA, Sensor, Bond graph INTRODUCTION Offshore, marine, aircraft and other complex engineering system operate in harsh environmental and operational conditions and must meet stringent requirements of reliability, safety and maintainability. Failure Modes and Effects Analysis is generally used in the above industries to reduce design risks. It is generally recognised that failure analyses are usually carried after the system has been designed, and often after it has been manufactured. If failure analysis subsequently reveals any problems, modifications to the hardware are often costly when carried out so late in the design proccess. The software tool for automated FMEA presented in this paper is based on qualitative modelling of generalised information flow (energy, information or

2 Figure 1 Maintenance Aware Design concept mass) in a system, recognising that the primary functions of mechanical, electrical, thermal and/or hydraulic systems (or their combination) as well as modern neuromechanic systems (also called mechatronic systems in the mechanical domain) are the generation, transmission and conversion of energy. Thus, the method is independent of the technological domain of the system and mixed technology systems can be analysed. Failure analysis (FMEA and FMECA) in the context of Maintenance Aware Design (MAD) and Physical Health Monitoring (PHM), Figure 1, has additional purposes-identification of potential operational problems in the concept stages of design, selection of optimum level of diagnostic and prognostic requirements and an optimum monitoring mix, i.e. sensors. The approach to automated FMEA and software design to utilise this approach are described and description of the software is provided. FAILURE MODES AND EFFECTS ANALYSIS Failure Modes and Effects Analysis (FMEA) is a tool for the systematic analysis of failures and malfunctions which can occur in a system [1]. It is only concerned with hardware failures and malfunctions and does not include hazards caused by human error, environmental effects or any other operating and hazardous conditions outside the scope of the design limits of the components. The basic objective of FMEA is to identify all possible failure modes of the components and deduce the consequences of these faults. It is also possible to determine the criticality of each failure mode, in which case the technique becomes FMECA. Other information sought during an analysis is the nature of the failures and whether there are any redundancies, backup components or sensors which can safeguard against the failure of a component. SOFTWARE FOR AUTOMATION OF FMEA The algorithm used in the software is based on the following axioms: 1. The functional characteristics of a system are implicitly described by a mathematical model of the system 2. The energy levels in the system are determined by the values of system parameters. 3. A change in the value of a system parameter results in a change of energy level at a parameter location. 4. A local energy change propagates through the system and induces functional changes in system behaviour and represents mode of failure. 5. Changes in energy levels in various locations in the system identify the effects of a fault. 6. A change in the value of a parameter associated with a component is the result of a structural change in the component. Such change may be due to a local structural fault, control input or disturbance. 7. Each local structural fault in a system component represents mode of failure in the component. 8. Structural change in a component (local structural fault) will change the energy level at a component location in the system i.e. it represents effect of fault in a component. During analysis an energy disturbance, representing a fault, is introduced to a component and the effects observed. For example, the introduction of increased

3 pressure p down Figure 2 Fault in a hydraulic system leakage at a pump location will result in a change in the energy level at this location, Figure 2. Introduction of a disturbance at a particular location in the system causes changes in state variables of the system which are propagated through the system both in forward and backward directions from the location of the fault, Figure 3. Thus the disturbance causes changes in the energy levels at different parts of the system and thus changes in the value of power variables at these locations. It is then possible to develop a table which relates changes in power variables to specific disturbances, [2, 3]. The analysis process is performed on a modified bond graph model (for structural faults) using qualitative dynamic analysis and on a higher level models using an algorithmic procedure. The analysis is completely domain independent as it is purely based upon propagation of energy. If the system has a hierarchical structure, then the disturbances (failure mode) at the higher level are the Figure 3 Fault propagation Figure 4 Hierarchy of FMEA results of a disturbance being introduced at lower level of the hierarchy, Figure 4. At the lowest level, disturbances are caused by structural faults i.e. faults induced by changes in the geometric and/or material properties of an element and tribological pairs or changes in temperature. Before failure analysis is performed, the function of the system should be established. This will yield information about the behavioural and structural properties of the design under consideration. Software The software which has been written in JAVA, allowing it to be run under most operating systems, is currently under tests. Various components of the software are shown in Figure 5. The software is designed to facilitate stepwise or automatic determination of failure modes and effects. The process of FMEA is carried out in a number of steps. In the Modelling step the model is built using provided models of components, Figures 6 and 7. After the model is constructed the FMEA model is translated into bond graph model and the integrity of the model is automatically checked. At the same time controllability of the system can be assessed using method described in [4], Figure 8. In the Analysis step the top event (a component whose operation is under consideration) is nominated and faults are propagated through the system. The next step is generation of FMECA reports (in XML and HTML format) which are produced using default criticality assessment. The criticality maybe subsequently

4 Figure 5 Components of software re-evaluated by accessing criticality tool,[5]. This tool allow either fuzzy (linguistic) or crisp (numeric) assessment of criticality, Figure 9. After propagation is completed sensor selection procedure takes place. In the first step minimum number of sensors using complementary algorithm, [6], and their location is evaluated for each level, and a choice of sensor locations is presented, Figure 10. If designer is not happy with a sensor at a particular location he may investigate sensors on a lower level which will provide indirect indication of faults on the higher level. When the selection of the sensor is completed the observability of the system using selected sensors is checked by analysis of the bond graph model of the system. Lower Figure 7 Modelling icons level (component level) faults are modelled directly using the bond graph approach, the pre-defined faults are then associated with faults in the component. These faults are determined by propagation in the bond graph model of component. The results of this analysis is passed to procedure which generates FMECA reports. Again criticality of component faults can be assessed either using fuzzy or crisp evaluation. Faults which cannot be evaluated using energy disturbance method are represented as paired concepts (e.g. contamination-wear, vibration-noise) and evaluated using Fuzzy Cognitive Maps, Figure 11, using approach outlined in [7]. This technique when completed will be integrated with functional analysis of the system. CONCLUSIONS This paper describes engineering principles and the implementation of an automated Failure Mode and Effects Analysis. The method is based on a qualitative analysis of power flow in the system in response to disturbances and the separation of the functional and structural Figure 6 FMEA-Model of hydraulic system

5 Figure 8 Java Bond-bond graph automatically translated from FMEA model (insert shows results of simulation) Figure 9 Fuzzy RPN-fuzzy logic based evaluation of criticality (insert shows panel for entering severity, criticality and occurrence of faults) Figure 10 Fuzzy Minimise - selection of minimum number of sensors

6 Figure 11 Fuzzy Cognitive Map - modelling of cause-effect concepts modes of failures. By separating the system functional behaviour from the behaviour of its components, it is possible to develop a suite of failure models of standard components which are re-usable. Further work will be directed towards developing techniques for handling feedback systems, multiple faults and redundancies in a system. ACKNOWLEDGEMENTS Some part of the software presented in this paper were based, with kind permission, on software developed by Mr. Stephen Mohr (FCM) and Robert Bajzat (Bond graph). Fuzzy logic toolbox FuzzyJ was used with a permission of NRC (Canada). Above contributions are greatly appreciated. This continuing project is carried out in collaboration with Impact Technologies LLC (USA) and Hood Technology Corporation (USA). REFERENCES 1. BS5760: 1991,"Reliability of systems, Equipment and components", Part 5, Guide to Failure Modes, Effects and Criticality Analysis (FMEA and FMECA), Kohda, T., Inoue, K., Takata, S. and Asama, H., Disturbance analysis Using a system bond-graph, IEEE Proc. Annual. Reliability and Maintainability Symposium, 1999, pp Bull, D. R., Stecki, J. S., Edge, K. A. and Burrows, C. R., Failure modes and effects analysis of a valvecontrolled hydrostatic drive, in Fluid Power Engineering: Challenges and Solutions, Tenth Bath International Fluid Power Workshop, Research Studies Press Ltd., 1997,. pp Sueur, C. and Dauphin-Tanguy, G., Structural Controllability/Observability of Linear Systems Represented by Bond Graphs, J. of Franklin Institute, 1989, 326, 6, pp Peleaz, C. E. and Bowles, J. B., Using Fuzzy Logic for System Criticality Analysis, Proc Annual Reliability and Maintainability Symposium, 1994, pp Reddecliffe, O. A. and Stecki, J. S., Synthesis of Fluid Logic Control Systems Using Digital Computers, Proc. Conf.. Computers in Engineering, 1974, I. E. Aust., pp.59-63, Sydney, Australia. 7. Peleaz, C. E. and Bowles, J. B., Using Fuzzy Cognitive Maps as a Model for Failure modes and Effects Analysis, Information Sciences, 1996, 88, pp