Condition Monitoring of Multi-Bolted Joints with Combination of Preload Measurement and Extended Evaluation from FEA.
Bolted joints in application always involve more than one bolt even though their analytical design is performed based on idealization and the bolt encountering the maximum load. In today's engineering world multi-bolted joints must be considered as a fastening system which has influences not just from the number of bolts in the system but also the complex structure in which the bolts are present and interact with each other through transmission of forces and vibration.
Innovative products need intensified reliability against functional failure of the complete structure and also to detect component faults if there is redundancy dimensioned for the system (to find the defect by a continuous operating system) e.g. one bolt has a lower preload. Typical failure modes of bolted joints include rotational loosening (self-loosening), non-rotational loosening (seating, load induced plasticity, thermal creep), corrosive failure, fatigue and rupture. However, the basic design of bolted joints does not cover all these nonlinear behaviors in detail. An improper design of the joint can not only lead to bolt failure but also component failure based on the loading and environment. Hence it is essential to extend the design process of joints to evaluations with FE simulation and experimental testing of the complete joint.
Self-Loosening of Multi-Bolted Joints
Rotational self-loosening of bolted joint is one of the predominant failure modes, especially for bolts under vibration loading. This phenomenon is critical especially for joints where the vibrations can induce relative transverse movements in the clamped parts . A test setup developed in late 1960's by Gerhard Junker is nowadays part of a German national standard - DIN 65151 . Another standard DIN 25201-4  describes test methods based on the test setup of DIN 65151 as a design guide for railway vehicles. The above tests with an eccentric cam drive producing a relative transverse displacement between clamped parts, to induce rotational self-loosening in bolts, are good only to visualize the phenomenon but not for giving suitable design guidelines to avoid or predict it. This is due to the limited possibilities for variation of the joint design parameters. Only one bolt can be assessed at a time with different clamp length and thread engagement lengths. Recent findings have shown a significant difference in the self-loosening behavior of single bolt joints and multi bolted joints [4, 7, 8, 10]. A comprehensive study on the effect of clamped part in a multi-bolted joint on self-loosening behavior can be found in .
Condition Monitoring of Multi-Bolted Joints
Structural Health Monitoring (SHM) is a modern technique to detect faults and failure of the monitored structure. Different types of sensors are often used for monitoring e.g. strain gauges, accelerometers, etc. These sensors are often fixed at different locations in the whole structure. From the perspective of multi-bolted joints, bolts integrated with strain gauges as sensors, are particularly interesting. With this an idea of the prevailing preload and additional axial forces could be monitored. As a consequence, it is important to know how the axial additional bolt load is distributed. With this aim a unique test setup -MoHaS is developed. The abbreviation MoHaS stands for "Monitoring of Structural Health and Self-loosening". It is built to encompass different functional failures with multi-bolted joints and to study their cause and effect scenarios. On one hand structural monitoring with different load types and bolt arrays is possible and on the other hand self-loosening tests on a two bolt multi-bolted joint is possible. In this paper the focus is set to introduce the testing stand MoHaS and to show an analytical model of the main bolt arrays. For a defined experimental load case results are compared with FEA.
Another part is the bending sensitivity of bolts with integrated strain gauges.
The test configuration developed at University of Siegen - MoHaS (Figure 1) combines different objectives. It enables us to investigate the transmission behavior and the self-loosening behavior of different multi-bolted-joint arrays. For this purpose, two interfaces in the XZ plane are envisaged. The top most plate (shown in green) provides the force introduction point. This plate can be used as a multi bolted joint with two M10 bolts with different load introduction possibilities and to study the self-loosening behavior as in . The underlying plate (shown in yellow, hereafter called friction plate) is one part of the investigation in this paper and has M10 nut threads at different positions to allow studies to be performed from the top side and the bottom side. The perforations are such that the bolt positions can be varied for the study of self-loosening and condition monitoring. To minimize the influence of friction, flat bearing cages are used in the second interface. The influence of friction is minimized, to reduce the possible influences on the test configuration. The adapter plate (shown in orange) supports the flat cage and the friction plate. At the mirror imaged representation in the lower part of Figure 1 are the four red bolts which are monitored. The base plate (blue) is fixed to an angle plate of a hydraulic pulsator. The preload is measured by means of a quarter-bridge strain gauge in the shank of the bolt. Here, a strain gauge is glued into the range of homogeneous strain in the shank. It can detect the amount of preload by prior tensile calibration.
The four bolts shown in red are M10 x 50 with grade 10.9, selected in accordance with DIN EN ISO 4014 . The bolt is tightened using preloadcontrol method with a load of 29.6 kN. The coefficient of friction is assumed as 0.12 for head and thread region. A utilization factor of 90% for strength grade 8.8  is chosen. The discrepancy between strength grade 10.9 and preload for 8.8 was deliberately chosen as a safety measure owing to the diminished cross sectional area due to the integration of strain gauge in the shank.
CONDITION MONITORING WITH MOHAS
As a first step an analytical representation is made by using the method of equilibrium state (refer Fig. 2) The load F which is prefaced in the lever is divided into forces [U.sub.1] and [U.sub.2]. These forces are the loads for this bolt array and must be compensated by the bending stiffness of the bolts and friction at the interface. [F.sub.QS1/2] is the force component acting on the bolts 1 and 2 whereas [F.sub.QS3/4] is acting on the bolts 3 and 4. To account for preload variation from assembly influences, the transmitted friction forces in the interface are subtracted from the transverse force in each case. This was calculated for the bearing cage with [[mu].sub.T] = 0.01 and the existing preload. Calculation is based on the simple friction law by Coulomb. By building the ratio of the forces [F.sub.QS1/2] and [F.sub.QS3/4] and using the intercept theorem it is possible to calculate the center of rotation in every case.
Simulation was performed with Abaqus 6.11, using the complete test-stand, paying attention to the bolt array as in Fig. 2. The load is a sinusoidal force with an amplitude of [+ or -] 3 kN in Z-direction. The bolts are preloaded to 28.9 kN using a pretension section. From the simulation, no displacement of the bolt heads could be observed even though modelling is performed with contact definition without tie constraints. As a first result, preload behavior under the defined load condition is compared in Fig. 3 for simulation and experiments. It is obvious that in the test with strain gauge screws, significantly higher values for [F.sub.SA] (additional bolt force) were determined as against the simulation. The ordinate shows only the segment 25-31 kN. Since the displacement values of simulation match with those of the experiments, some questions arise as to measurement errors or unrecognized factors influencing the accuracy of the strain gauge integrated bolts. Among other things, the bending of the bolt (S-deformation), position of the strain gauge in the bore, influence of reduced cross-section of the bolt could be the contributing factors. Conversely, errors could have appeared from the numerically determined values. In preliminary studies it could be observed that quarter-bridge strain gauge screws are sensitive to bending stress and high bolt load .
The diagram Figure 4 shows the developing of the axial additional bolt load dependent on the angle of rotation of the bolt. This is interesting to know if you have bolts with transverse load and bending of the screw (not only tensile load). For this experiment, vibration test stand adapted from Junker  is used. Displacement amplitude of [+ or -] 0.5 mm was set as load for this test. The load itself is in sine form due to the rotating cam drive of vibration test stand. Speed of the testing stand was set to 50 revolutions per minute. A bolt with integrated strain gauge was used that was calibrated to a maximum of 30 kN.
In the first step the bolt is preloaded with 10 kN as measured by the strain gauge. The test is performed for measurements over a period of approximately 15 seconds. After that the machine is stopped and the preload is increased to 20 kN and then 30 kN force. The experiments were performed for 16 rotational positions of the strain gauge inside the bolt. Rotation of bolt head about the bolt axis is denoted by [phi] here. The experiments were performed at every 22.5[degrees] about bolt axis with 10, 20 and 30 kN preload respectively. To maintain the different positions, the nut component is rotated for developing the preload. From these measurement results, respectively the axial additional bolt load was recorded and averaged. Behind every data point in Figure 4 and 5 is the average value over the mentioned measurement period. The shown profiles have a sinusoidal character which repeats itself every 180[degrees] ([pi]). They have different amplitudes to the same external load. The phase shift angle a is depending on the rotation angle [phi]. Thus, a is related to the resilience of the bolt and the thread pitch.
Looking at the angle differences between the maxima of the curves and the assumption that in the experiments no slippage occurred, the difference of the angles is analogous to continue rotational angle to increase preload force. The resultant curves show a phase shift, due to the different preloads belonging to different rotation angle, the position between the load direction and strain gauge in the bolt is unequal in each measurement point. This is illustrated in Figure 5 vividly. In the background the contour of a bolt head is shown in gray. Here, the phase shift is clear by the rotation of the ribbon. The distance between the curves and the differential amplitude levels cannot infer a linear relationship, although the preload has been increased accordingly. Due to the properties of the test stand there is sliding operations below the bolt head. This is dependent on the load amplitude (constant here) and the preload level. By the preload levels of 10 kN and 20 kN this is still clearly visible to the eye. At 30 kN, it can only be detected from the measured data. This ensures that the thread-lash is exploited and the bolt gets a bending moment. Scaling refers to a maximum force of 1.6 kN and for using a screw which is described in Chapter - MoHaS test stand.
To illustrate the importance of the knowledge of preload in the fastening system, only bolt S1 was tightened with lesser preload (19,6 kN) as shown in Fig. 6. The simulation in this variant already shows a sliding of the bolt head at bolt S1 and S2 (not shown here). Such deviations can arise in application, especially in lightweight designs, leading to the risk of self-loosening which is not considered sufficiently in advance. Figure 4 shows the preload curve of all four bolts in the system. By lower preload of bolt S1, self-loosening occurs because critical displacements and rotations are achieved in conjunction with the lower preload level. After the preload of this screw has fallen below 5 kN, bolt S2 is loosening, although the loading was kept constant at [+ or -] 2.2 mm. The transverse load to be transmitted, increases due to the loss of preload in bolt S1, such that the critical self-loosening limit at bolt S2 is overcome. It is apparent that the number of load cycles is very low until complete loosening of bolt S1 (around 800 load cycles). This implies that a use of a "time resistance range with partial loosening" is not possible.
An additional aspect is the coupling of the bolt behavior within a structure by multi-bolted-joints and the sequence of self-loosening of different bolts. Hence the detachment of a multi-bolted-joint into single-bolted-joints for the purpose of design calculation is not attractive for avoiding self-loosening.
Further the measurement illustrates that the preload curve points towards a functional failure event - too low preload. By measuring the preload of bolts S1 to S4, appropriate evaluation and knowledge about the structure regarding the damage process could be predicted. Under these conditions, a preload measurement with experimental verification is also interesting for the operation in critical applications. It not only leads to a substantial increase in operational reliability but also avoids damage costs (which can grow rapidly). In the case of Fig. 4, too low preload at bolt S1 would have been obvious. If they were monitored they could have pointed to a possible failure with knowledge about the structural behavior.
EVALUATION OF SELF-LOOSENING BEHAVIOR USING FEA
With modern FE tools it is possible to simulate self-loosening behavior based on the mechanism proposed by Junker  and works of Pai and Hess . For this a complete 3 dimensional bolt with pitch of threads is essential [2, 3, 4, 5]. The detailed process of modelling to simulate the self-loosening behavior can be found in . Here the focus will be on multi-bolted joints and the influence of clamped part on the loosening behavior of the bolts. Two bolts of M10 according to DIN EN 1665 are modelled with a distance of seven times the nominal diameter to simulate self-loosening behavior. Two geometries of the clamped part are considered, one with rectangular plates (Model 1) and another with a small portion of the clamped parts cut away resulting in C like plates (Model 2). The clamp length and length of thread engagement in both models are 15mm. Material properties of Aluminum alloy - AL7075 are used for the clamped part (shown in green) and 42CrMo4 is used for the nut component (shown in yellow). For bolts, material properties of 10.9 grade is used. A preload of 25kN is generated in each of the bolt via a pretension section. Loading is displacement controlled in sinusoidal form where the top plate (green) is allowed to move freely with input load in x direction. The bottom plate (yellow) is fixed in x, y and z directions.
Influence of Clamped Part Geometry on Self-Loosening Behavior
In Figure 7 a comparison is made between model 1 and 2 with just bolt number 2 (B2) and relative transverse loading of [S.sub.Q] = 0.21mm between the clamped part and nut component, as a sine function for 10 load cycles. It can be directly observed that the self-loosening is reduced in model 2. Due to the deformation behavior of model 2, the deflection of bolt head about the bolt axis is reduced. This leads to reduced loosening gradient. The deformation behavior of model 2 also produces a torsion load on B2 which can be seen from the greater fluctuation of head loosening angle. This behavior can lead to design inputs such as tailoring the load transmission behavior in a component to reduce loosening of bolts.
In this part we extend the investigation from above to the complete joint. This leads to self-loosening behavior as shown in Figure 8. Here two bolts are considered with 25 kN preload each and the same transverse loading as above is performed. For model 1 when there are two bolts in the joint, the transmission behavior is such that some of the energy is absorbed for deformation of the cross-section between the bolts. This leads to reduced transmission of transverse slip at the end of B2. Hence B2 loosens less than B1. For model 2, the deformation behavior is such that most of the energy is absorbed as strain and direct transverse loads are not transmitted. This means that self-loosening can be influenced by clamped parts in two ways. One, with introducing a resilient member that absorbs the transverse forces as strain and another by channeling the load transmission behavior. Apart from this, comparing the results of Figure 7 and 8, addition of bolts can lead to reduction of loosening in individual bolts of the fastening system.
In this section, same models from above are loaded in another transverse direction (along z axis). This produces a rotational loading on the system as against a transverse loading along the line connecting the bolts. Figure 9 shows the resulting behavior of the two models under this modified loading condition where [S.sub.Q] = 2mm (approximately 10 times the loading in x direction). It is interesting to observe that such a loading does not produce any self-loosening in B2. This is because the loading does not produce any relative transverse displacements at B2. From this it can be concluded that based on the direction of transverse load in a mufti-bolted-joint, some bolts may or may not loosen.
The self-loosening behavior of multi-bolted-joints is coupled with the behavior of each bolt. Load cycles until complete loosening lie typically in the range from [10.sup.2] to [10.sup.3]. The structural integrity can be predicted and assured with preload measurements also with respect to self-loosening.
The deviation of [F.sub.SA] (additional bolt load) is depending on the bending sensitivity of bolts with integrated strain gauges. The bending sensitivity is depending on the preload and the orientation between the load direction and the strain gauge. The deviation in the researched case was up to 70%. Also it is possible to get rough knowledge about the rotation for the angle-controlled tightening.
From simulations, peculiar behavior was observed for self-loosening of multi-bolted joints. Apart from significant difference in a single and two bolt joint models, the influence of clamped part itself is noteworthy. Clamped parts can be designed such that they absorb or channel the transverse forces instead of direct transmission. Especially, resilient materials like Aluminum can be beneficial instead of stiff materials like Steel.
Behavior of multi-bolted joints should be assessed and optimized for applications requiring high reliability. A continuous monitoring of the preload alone can give significant information about the structure and its susceptibility to failures like self-loosening.
[1.] Junker, G., "New Criteria for Self-Loosening of Fasteners Under Vibration," SAE Technical Paper 690055, 1969, doi:10.4271/690055.
[2.] Pai, N. G., and Hess, D.P., Three dimensional finite element analysis of threaded fastener loosening due to dynamic shear load, Engineering Failure Analysis, vol. 9, pp. 383-402, 2001.
[3.] Koch, D., Beitrag zur numerischen Simulation des selbsttatigen Losdrehverhaltens von Schrauben-verbindungen. Dissertation, Shaker Verlag, Aachen, 2012.
[4.] Dinger, G., Ermittlung des selbsttatigen Losdrehens bei Mehrschraubenverbindungen. Dissertation, Shaker Verlag, Aachen, 2013.
[5.] Manoharan, S. and Friedrich, C., "Self-Loosening of Three Similar Bolted Joint Designs Using Finite Element Analysis," SAE Technical Paper 2014-28-0035, 2014, doi:10.4271/2014-28-0035.
[6.] Kyriakides, E., Polycarpou, M., Intelligent Monitoring, Conrtol, and Security of Cirtical Infrastructure Systems. Dissertation, Springer, Heidelberg, 2015.
[7.] Manoharan, S.K., Friedrich, C., Self-loosening behaviour of multi-bolted joints - requirements for engineering design in real applications, Proceedings of the 6th International Conference on Manufacturing, Machine Design and Tribology, paper no. TH-A-1-4, Okinawa, 2015.
[8.] Guggolz, D., Manoharan, S.K., Friedrich, C., Avoiding of self-loosening in components with multiple screw joints, Proceedings of IMECE 2015, paper no. 50369, ASME, Houston, 2015.
[9.] Dumpelmann, C., Guggolz, D., Friedrich, C., Montagesensitivitat und Ubertragungsverhalten von Mehrschraubenverbindungen mit VDI-Richtlinie 2230 Blatt 1 und Blatt 2 rechnerisch vorhersagen und experimentell verifizieren, VDI-Berichte 2270, VDI Tagung, Munich, 2016.
[10.] Manoharan, S.K., Friedrich, C., Design and reliability influences on self-loosening of multi-bolted joints, Proceedings of International Conference on Integrity-Reliability-Failure IRF 2016, University of Porto, paper no. 6302, Porto, 2016.
[11.] DIN 65151: Aerospace series - Dynamic testing of the locking characteristics of fasteners under transverse loading conditions (vibration test). Beuth Verlag, Berlin.
[12.] DIN EN ISO 4014 Hexagon head bolts - Product grade A and B, Beuth Verlag, Berlin.
[13.] DIN 25201-4 Design guidelines for railway vehicles and their components - Bolted joints - Part 4: Securing of bolted joints. Beuth Verlag, Berlin.
[14.] VDI 2230 Part-1, Systematic calculation of high duty bolted joints with one cylindrical bolt, Verein Deutscher Ingenieure.
[15.] DIN EN 1665, Hexagon bolts with flange, heavy series, Beuth Verlag, Berlin.
Shiva Kumar Manoharan, Christopher Duempelmann, and Christoph Friedrich
University of Siegen
Shiva Kumar Manoharan
University of Siegen
+49 271 7404399
|Printer friendly Cite/link Email Feedback|
|Title Annotation:||finite element analysis|
|Author:||Manoharan, Shiva Kumar; Duempelmann, Christopher; Friedrich, Christoph|
|Publication:||SAE International Journal of Passenger Cars - Mechanical Systems|
|Date:||Apr 1, 2017|
|Previous Article:||Development of High Fidelity Dynamic Model with Thermal Response for Single Plate Dry Clutch.|
|Next Article:||Influence of Nozzle Divergent Part Length and Throat Diameter on Vortex Control of Initially Subcooled Flashing Flow.|