Component monitoring with integrated sensors
Borris van Thiel
Peter Nyhuis
Institute of Production Systems and Logistics Leibniz University of Hannover Hannover, Germany
[email protected]
Institute of Production Systems and Logistics Leibniz University of Hanover Hannover, Germany
[email protected]
Abstract— As part of the collaborative research centre 653 (CRC 653) at the Leibniz University of Hannover, sensors and sensitive materials are being developed which can be integrated into components and register the component condition directly. The part-project “Component status driven maintenance” presented here, develops the methodical procedure for being able to use these sensors in preventive maintenance. Preventive maintenance however is not possible solely on the basis of the current status information of the components. A much more decisive factor is a database by means of which the further development of this status can be estimated. This enables the determination of possible causes of faults before they occur, and the assessment of the failure risk of components. System Engineering – Optimization, Reliability, Fault Tolerant, Maintenance
I.
MONITORING OF COMPONENT LOADS
The increasing competitive and cost pressure on companies also increases the requirements on maintenance. On the one hand components and systems should demonstrate high availability, but at the same time they must also be operated cost-effectively. Condition-monitoring systems are therefore coming into increasing use. They register the component conditions, and use this information for preventive maintenance, in order to estimate the future failure risk of the components. In the case of conventional sensors such as vibration sensors, the measurement of the status takes place indirectly by registering ancillary effects. Vibrations are caused amongst other things by slight imbalances or play, thus registered vibrations indirectly confirm the existence of such imbalances. However this does not enable the direct registration of the component status. Reliable measurements can only be obtained after slight damage has occurred, whose effect leads to an imbalance, and therefore measurable
vibrations. On the basis of this requirement, CRC 653 is developing sensors and sensitive materials which can provide direct information on the condition of the component. II.
COMPONENTS WITH INTEGRATED SENSORS
CRC 653 consists of various part-projects which are developing a wide range of sensors and storage media designed to register and store status information which is inherent to the component. The sensors consist on the one hand of microsensors for the measurement of conditions such as temperature, expansion and eddy currents [1]. In parallel with the microsensors, sensitive materials are also being developed, whose material structure changes according to the load. One approach to this task is the use of metastable, austenitic stainless steel, in which defined areas of the structure are characterised austenitically and martensitically by means of local heat treatment with a laser beam [2]. Depending on the level of a load, Austenite is converted into Martensite. This conversion can be measured by an eddy current sensor. The change in the structure can therefore be detected immediately, thus the condition of the component can be registered directly. The advantage of such sensitive material is the direct determination of the component condition on the basis of the material structure. In this way, it can be determined immediately whether a load has led to changes in the material structure, and will have an effect on the future behaviour of the component. A further advantage is that the changes in the material structure can be checked from the very beginning over the complete life cycle of a component. Overloads for example can be detected in this way. Figure 1 shows an already realised constant velocity joint as a demonstration object.
Figure 1: Constant velocity joint of sensitive material [Source: Institute of Material Sciences – University of Hannover]
A further approach consists of the specific application of internal stresses in selected peripheral surface areas of the constant velocity joint. Here tensile or compressive stresses can be induced according to the trial arrangement. These affect the working life of the component, since the internal stresses overlay the occurring loads. The internal stresses induced also change with overloading of the component, so that component loads can be registered by regular measurement of the internal stress [3]. The exact effect of loads on the internal stress is currently the subject of further investigations, in order to make these usable for maintenance purposes. Work is also in progress on an eddy current process for the easier determination of the internal stresses. So far, the determination of internal stresses has usually been carried out with the aid of an X-ray diffractometer, which registers the tensions of the lattice structure. This is however a very complicated measurement. The sensors researched by CRC 653 are particularly suitable for the monitoring of fatigue symptoms with the main influencing factors [4]: • • •
Load, Temperature and Time.
The estimation of the fatigue symptoms is based on the dynamic fatigue trials of Wöhler [5]. In his trials, Wöhler was able to demonstrate a correlation between the stress amplitude σa under vibration stress and the number of load cycles until a fatigue fracture occurs [6]. In a Wöhler trial, test specimens are subjected to a constant vibration stress until a fatigue fracture occurs. The higher the applied stress, the sooner the fatigue fracture occurs. The progression of the correlation differs according to the material. With the sensors developed by CRC 653, it is possible to detect the level of the stresses on a component. Short-term overloads in components can also be registered. The higher the load on a component, the greater the change in the material structure or the internal stress on the surface. In addition to the constant stress of a Wöhler trial, further investigations are currently being carried out in which the test specimens are subjected to short-term overloads (Figure
2). The purpose of these trials is to determine the effects of the overload on the working life of the component. However, the effect of overloads on the working life of an individual component has little significance on its own. Generally applicable conclusions can only be derived through the continual gathering of stress information in a database.
Stress (Stress Amplitude σ)
Overload σo,max
Number of Load Cycles Figure 2: Simplified representation of an overload trial
III.
USE OF A DATABASE FOR MAINTENANCE PLANNING
The newly developed sensors enable the registration of component conditions. They are however not immediately usable to determine the failure risk of the constant velocity joint. A database must be created, by means of which the further behaviour can then be estimated. For this purpose, the sensor data must be evaluated and compared with the plan data. Plan data is necessary in order to detect a deviation of the measured sensor data from the expected values. Plan data is usually based on the empirical knowledge of personnel or the specifications of the manufacturer. Components are therefore replaced as a precautionary measure from a certain degree of wear. However, due to the increasing cost pressure on maintenance, a more accurate estimation of the maintenance requirement is needed. This requires comprehensive wear models, whose development requires large volumes of condition data. This can only be done with the aid of a database.
Figure 3: Example database for the constant velocity joint
For this reason, an example database has been created (Figure 3). The database contains component types, such as the constant velocity joint, with other relevant information such as the material, integrated sensors and their measurement capabilities. Every added component is assigned an individual serial number. For the constant velocity joint for example, this produces a database which can be used for the development of wear models. The database can be used to store the load information of numerous components. This is important, since such load information must initially be available up to the point of failure of a component, before a forecast of the failure probability can be made. Newly-designed components can therefore initially only be maintained on a conditionorientated basis, until an adequate wear model is available. The more components are replaced at the end of their life cycle and their data is entered fully into the database, the more accurate will be the definition of the failure probability of subsequent components examined. Using the example of the constant velocity joint, several such components must therefore first be installed in vehicles, and then replaced following identification of a defect. As soon as this information is available, it can be statistically evaluated and used for the next components. The use of the sensors is therefore particularly helpful for components in series
production. The construction of the wear models takes place in the database by means of the regular importing of values measured by the sensors. IV.
MONITORING OF TYPICAL LOAD PROGRESSIONS
The findings from the load trials described above serve as the basis to be able to draw conclusions on the further working life and the failure probability by means of statistical frequency distributions. The method of statistical evaluation is described below using the example of simple bending or torsion loads. The frequency distributions described are only for example purposes and are currently being verified in trials by simple bending tests. In the further course of the project, the trials are to be extended to include sample components with more complex loads, such as the constant velocity joint. For the construction of the wear model, the components must first be maintained according to their condition and on the basis of the sensor information. This allows to determine the relationship between the component load and the working life for every component. Once sufficient condition information has been gathered, this can be evaluated statistically (Figure 4).
Frequency Distribution σ
Frequency Distribution N
Stress
Stress
(Stress Amplitude σ)
(Stress Amplitude σ)
σo,max Number of Load Cycles
Number of Load Cycles
N
Figure 4: Statistical evaluation of load information
The first consideration is that of the maximum load in comparison to the working life. This is done by evaluating those components which have been subjected to a similarly high maximum load σo,max over their life cycle. Every component however is individual, and can withstand a different vibration frequency of loads. For components subjected to similar maximum loads, this can be used to produce a frequency distribution of the working life (Figure 4 left). The second consideration is that of the working life in comparison to the maximum load. Also components with a similar working life could have been exposed different maximum loads. This can be used in order to check the frequency distribution σ of the maximum load (Figure 4 right). V.
ESTIMATION OF THE WORKING LIFE FOR
(Figure 5 left). Using the example of the constant velocity joint, it can now be estimated whether the working life of the component will extend to the next inspection or whether it must be replaced. The next step is to check the frequency distribution σ of all components in the database with a working life similar to Nestimated of the inspected component (Figure 5 centre). This gives a good indication of the validity and reliability of the information: a wide frequency distribution σ indicates highly fluctuating loads with an identical working life, so that the conclusiveness of Nestimated is also low. The narrower the frequency distribution σ, the more accurate is the prediction of the working life. Together they form the working-life-corridor, which shows the determined maximum load, the estimated Working-Life and the validity of the information (Figure 5 right).
MAINTENANCE PLANNING
When sufficient data is available for the derivation of the wear models, the models can be used in future for the purposes of preventive maintenance. First step of preventive maintenance is to determine the maximum load σdetermined of the inspected component, like the constant velocity joint, and then to derive from the wear models in the database the corresponding frequency distribution N. This enables an estimation of the components working life Nestimated
σdetermined Æ Nestimated
Nestimated Æ Frequency Distribution σ
Working-Life-Corridor
Stress
Stress
Stress
(Stress Amplitude σ)
(Stress Amplitude σ)
(Stress Amplitude σ)
Working-Life-Corridor
Frequency Distribution σ
σdetermined Nestimated Time
σdetermined Nestimated Time
Figure 5: Use of the failure probability for maintenance purposes
Nestimated Time
VI.
SUMMARY AND OUTLOOK
CRC 653 is developing different sensors in order to be able to register component loads directly. These are suitable for the application in preventive maintenance. The basis of preventive maintenance is an adequate wear model of the component on the relationship between the load experienced and the expected working life. For this purpose, a database was created in the part-project “Component status driven maintenance”, which collects load information on components. A method was also developed for the evaluation of the load information recorded. The method is currently being validated by means of bending trials, and is to be extended in future to components exposed to more complex loads. The existing model of linear damage accumulation [5 – based on Palmgren and Miner] is also being extended, in order to be able to use it for a more accurate delimitation of the working life corridors.
German Research Foundation (DFG) for its financial and organisational support for this project. REFERENCES [1]
[2]
[3]
[4]
ACKNOWLEDGMENT The work presented was carried out within the scope of the collaborative research centre 653 "Gentelligent components in their life cycle". The authors thank the
[5]
[6]
T. Griesbach, L. Rissing, H.H. Gatzen, “Integration of a Temperature Sensor into a Modular, Multifunctional Micro Sensor Family”, 4th I PROMS 2008 Virtual Conference on Innovative Production Machines and Systems. G. Mroz, W. Reimche, Fr.-W. Bach, “Acquisition of Discrete Component and Loading Information in the Component's Edge Region Using Innovative Sensor Technology”, 4th I*PROMS 2008 Virtual International Conference on Innovative Production Machines and Systems. B. Denkena, I. Steinbrenner, J. Ostermann, R. Dragon, “Machining Microstructures by means of a High-Dynamic Fast-Tool-Servo” Proceedings of the 9th international conference of the european society for precision engineering and nanotechnology, 2009. B. van Thiel, P. Nyhuis, “Database for a Gentelligent Maintenance Planning”, 15th Annual International Conference on Industrial Engineering - Theory, Applications & Practice, Proceedings of the IJIE 2010, ISBN 97809652558-6-8, Mexico City, 2010 B. Bertsche, “Reliability in Automotive and Mechanical Engineering - Determination of Component and System Reliability”, ISBN 978-3540-33969-4, Springer 2008 V. Läpple, „Einführung in die Festigkeitslehre - Lehr- und Übungsbuch“, Vieweg + Teubner, Wiesbaden 2008.