You do not have permission to edit this page, for the following reason:

You are not allowed to execute the action you have requested.


You can view and copy the source of this page.

x
 
1
<!-- metadata commented in wiki content
2
3
4
<big>'''A new calculation method of bearing reliability of tyre unloader based on heterogeneous dimensional interference model'''</big>
5
6
==Jingxiu Ling<sup>1,2, 3</sup>, Rongchang Zhang<sup>1,2</sup>, Jiacheng Shao<sup>1,2</sup>and Hao Zhang<sup>4</sup>==
7
8
<sup>1</sup>School of Mechanical and Automotive Engineering, Fujian University of Technology, Fuzhou 350118, China, <sup>2</sup>Fujian Key Laboratory of Intelligent Machining Technology and Equipment (Fujian University of Technology),<sup> 3</sup>School of Materials Science and Engineering, Fujian Institute of Technology, Fuzhou 350118, China,<sup>4</sup>CSCEC Strait Construction Development Co., Ltd., Fuzhou 350000, China.
9
10
Correspondence to: Jingxiu Ling  [mailto:ljxyxj@163.com ljxyxj@163.com]
11
12
<span id='OLE_LINK11'></span><span id='OLE_LINK12'></span>
13
-->
14
==Abstract==
15
16
Bearing is an important rotating support part of tyre unloader, and its fatigue reliability is an important part of the whole system reliability. Because of the huge alternating stress, the support bearing is required to have high fatigue life and reliability. In this paper, combined with stress-strength interference model and statistical theory, the life distribution of bearing steel material is predicted by using group test data; Based on the multi-rigid body dynamics and finite element numerical simulation platform, the reliability of the bearing of tyre unloader under different operating years was predicted by using the different dimensional interference model. The results show that the maximum resultant force of the bearing at the bottom rocker arm of the tyre unloader can reach 150kN, and the maximum transverse and longitudinal forces can reach 108kN and 78kN. When bearing the weight of the whole tyre and turning, the inertia force is the largest, the maximum stress value is 1316.2MPa, which occurs in the bearing inner ring and ball contact part. After the statistics, the stress amplitude distribution of the bearing conforms to Weibull distribution, and the life of the bearing follows lognormal distribution. After 10<sup>5</sup> tyre unloading, the fatigue reliability of the bearing is lower than 0.82, which is consistent with the actual working condition. Therefore, this model can be used to calculate the fatigue reliability of bearings conveniently and quickly, and provide certain theoretical support for the safety and fatigue reliability prediction of bearings.
17
18
'''Keywords''': Numerical simulation, fatigue life, reliability, different dimensional interference model
19
<!--==Article highlights==
20
21
<span id='OLE_LINK1'></span>
22
23
● Reliability prediction model for tyre unloader bearings based on heterogeneous interference theory.
24
25
● A set of fatigue reliability calculation method for the bearing of a tyre unloader is put forward.
26
27
● Probabilistic model for calculating bearing life using mathematical statistical methods.
28
29
● Fatigue life data of bearings obtained by using group method.
30
31
<span id='OLE_LINK2'></span>●Prediction of fatigue reliability of bearings based on the equivalent force method.
32
33
-->
34
35
==1. Introduction==
36
37
With the continuous improvement of China's machinery production process, the life and reliability of bearings has been greatly improved, but the bearings running on some large machinery, because of the huge alternating stress and complex working environment, making the life and reliability of bearings rapidly reduced. At the same time, with the rapid development of heavy machinery in China, the tonnage of tyres used is also rising. The object of this paper is a new type of giant tyre unloading machine, which unloads tyres weighing up to 6t. In the production process, the tyres need to be unloaded from a fixed position and flipped by the unloading machine after the completion of the previous process. During the unloading and flipping process, due to the huge weight of the tyre and the inertia force generated during the flipping, the tyre will collide and rub against the clamping mechanism of the tyre unloader, causing the system to vibrate and at the same time causing the bearings in key parts to be subjected to complex and variable random loads. Because of the randomness of the external load, the bearing material itself performance, size and other variability, the life distribution of bearings belongs to a probability distribution and with the growth of the use of years, bearing failure rate is on the rise. The reliability of the bearing life under different years needs to be studied to prevent bearing failure and subsequent safety accidents. The fatigue reliability of this new type of tyre unloader bearing is not systematically studied. In this paper, the fatigue reliability calculation method of the new tyre unloader bearing is proposed based on the theory of dynamic finite element and different dimensional interference model.
38
39
Current research on tyre unloaders is limited to the control methods and modes of operation. For example, Alessio et al. [1] introduced a robot that can assist tyre operators in the workshop to change tyres, which can be controlled in a variety of ways, such as automatic recognition of the user's gesture commands and remote operation through a control interface. In addition, Ján et al. [2] carried out a 3D modelling and dynamics analysis of a robot that unloads tyres, thus obtaining important mechanical parameters of the robot.
40
41
Existing bearing reliability predictions are different from the subject of this paper, and most of the analysis methods use isotropic interference models, which is the traditional stress-strength interference model. However, the strength distribution of general materials is difficult to predict and difficult to obtain strength distribution data through specific tests, and it needs to be combined with the material life distribution to obtain the strength distribution of materials through complex mathematical calculations, which is not conducive to the use of engineers. Domestic scholars Xie  and Wang  [3] proposed a heterogeneous interference model, which is no longer limited to the traditional reliability model in which the two variables must be of the same magnitude, but can be used to calculate the reliability through the stress distribution and the life distribution of the material. At the same time, the model no longer relies on tedious mathematical calculations but obtains the fatigue life distribution of the material through tests, which makes the fatigue reliability analysis much less difficult and simplifies the analysis steps. Liu et al. [4] provided a new analysis method for fatigue reliability of pipe structures by combining fatigue reliability of pipes under load with structural dynamic analysis. Zhou et al. [5] provided a theoretical basis for the life assessment and reliability analysis of rolling bearing systems through the fatigue reliability calculation formula of bearing systems under continuous load spectrum. Jin et al. [6] proposed an artificial intelligence method to analyze the fatigue reliability of aviation bearings. Qi and Liao [7] designed a system to evaluate the reliability of rolling bearings of traction motors based on MATLAB App Designer, which provided a new reference for the development and improvement of reliability evaluation systems.
42
43
Reuben et al. [8] improved this reliability assessment based on the Weibull diagram equation by estimating the variation in minimum bearing life and establishing confidence intervals using Monte Carlo simulations. Pape et al. [9] improved the calculation of bearing fatigue life by introducing residual stresses to the sub-surface region of the bearing. Wang [10] calculated the reliability life of ship unloader bearings by converting the tensile force of the ship unloader lifting mechanism into an average equivalent dynamic load on the bearings and then using the Miner criterion. Xia Xintao and Ye Liang et al. [11] proposed a new concept of rolling bearing performance retention reliability. Zhang et al. [12] established an analytical model for the contact fatigue reliability of main bearings based on the complex loading conditions of TBM main bearings and combined with fatigue cumulative damage and residual strength theory, based on a dynamic degradation model under complex conditions. Cheng et al. [13] developed a modified five-degree-of-freedom quasi-dynamic model considering multi-body interactions and used the modified fatigue life model proposed by Jones to assess the effect of angular displacement on bearing reliability. Herp et al. [14] proposed a bearing condition prediction method using temperature residuals and Bayesian probability statistics for wind turbine bearings, and this method can predict possible bearing failures within an average of 33 days. Tong et al. [15] proposed an improved model for calculating the operating torque of angular contact ball bearings (ACBB) by comparing the effect of different loading conditions on the fatigue reliability of angular misalignment on tapered roller bearings. König et al. [16] developed a model of bearing life which can determine the merits of hybrid bearings and allows for an analytical assessment of bearing life. Guillermo et al. [17] further extended a previously developed model for calculating bearing life based on high cycle fatigue by incorporating a new surface damage integral based on a creep mechanism into the model, providing a new approach to fatigue life prediction for bearings. Zhang et al. [18] used the maximum likelihood estimation method to calculate the actual value of the bearing, and used the SPSS curve and cumulative grey prediction model methods to train on some of the real data as a way of predicting the life of the bearing, and found that all three methods have some practical value in engineering, but the cumulative grey prediction was more effective. Lorenz et al. [19] developed a continuous damage mechanics (CDM) finite element (FE) model in order to investigate the effect of surface roughness on the rolling contact fatigue life of poor quality contact bodies and demonstrated the feasibility of the model. Pandey et al. [20] proposed a framework based on continuum damage mechanics and the finite element method to simulate the low circumferential fatigue crack expansion process, and developed a strain-based damage model to consider the effect of different strain ratios on fatigue damage. Cano et al. [21] developed a new intrinsic model for long-term prediction of creep deformation, damage and rupture by combining the Wilshire equation with continuum damage mechanics (CDM).
44
45
In summary, in terms of bearing reliability analysis and calculation, numerical analysis, theoretical calculation and experimental evaluation are mainly used at home and abroad to study the performance and life of various types of bearings, but there are fewer studies combining giant tyre unloaders and using heterogeneous interference models to analyse their bearing reliability. Therefore, this paper combines the existing information at home and abroad, mainly to fill the gap in the field of research on the fatigue reliability of bearings of giant tyre unloaders; Prediction of the life distribution of bearing steel materials using grouping method test data combined with statistical theory; Combining numerical simulation platforms such as dynamics and finite elements, the heterogeneous interference model is used to predict the reliability of bearings at different service lives. It also differs from traditional stress-strength interference models in that it directly uses the material life distribution for reliability calculations and is suitable for generalised applications in engineering.
46
47
==2. Reliability analysis process ==
48
49
<span id='_Ref96587959'></span>The fatigue reliability of the bearings of the giant tyre unloading machine is analysed and calculated by the numerical analysis platforms such as dynamics and finite element, using the theory related to fatigue reliability. The specific process is shown in [[#img-1|Figure 1]].
50
51
<div id='img-1'></div>
52
{| class="wikitable" style="margin: 1em auto 0.1em auto;border-collapse: collapse;width:auto;" 
53
|-style="background:white;"
54
|style="text-align: center;padding:10px;"| [[File:Ling_et_al_2023a_3475_600px-Draft_Ling_717181314-image1-c.png|600px]]
55
|-
56
| style="background:#efefef;text-align:left;padding:10px;font-size: 85%;"| '''Figure 1'''. Flow chart for calculating the fatigue reliability of a tyre unloader bearing
57
|}
58
59
==3. Maximum stress history of bearing==
60
61
===3.1 Structure and modelling of the tyre unloader===
62
63
The physical drawing of the tyre unloading machine is shown in [[#img-2|Figure 2]]. Because of the complex structure of the tyre unloader proper, the model (with tyre 2) is shown in [[#img-3|Figure 3]] by simplifying the 3D modeling of the threads and other parts of the tyre unloader blank that will not have an impact on the final result.
64
65
<div id='img-2'></div>
66
{| class="wikitable" style="margin: 1em auto 0.1em auto;border-collapse: collapse;width:auto;" 
67
|-style="background:white;"
68
|style="text-align: center;padding:10px;"| [[File:Draft_Ling_717181314-image2.png|centre]]
69
|-
70
| style="background:#efefef;text-align:left;padding:10px;font-size: 85%;"| '''Figure 2'''. Physical view of the tyre unloader
71
|}
72
73
74
<div id='img-3'></div>
75
{| class="wikitable" style="margin: 1em auto 0.1em auto;border-collapse: collapse;width:auto;" 
76
|-style="background:white;"
77
|style="text-align: center;padding:10px;font-size: 75%;"| [[Image:Draft_Ling_717181314-image3.png]]
78
79
1. Clamping plate; 2. Tyres; 3. Guide wheels; 4. Fixed ring; 5. Dynamic ring; 6. Fixed ring rotary; 7. Rocker arm
80
|-
81
| style="background:#efefef;text-align:left;padding:10px;font-size: 85%;"| '''Figure 3'''. 3D model of the tyre unloader
82
|}
83
84
85
The structure of the tyre unloader mainly consists of the Plywood 1, the Guide wheel 3, the Fixed ring 4, the Dynamic ring 5, the Rocker arm 7, etc. 16 pairs of rocker arms are mounted symmetrically on each side of the fixed ring, with the upper end connected to the moving ring by a linkage and the lower end connected to the cleat. The principle of operation is that the dynamic ring rotates at a certain angle with respect to the fixed ring, thus controlling the clamping plate on the rocker arm to clamp or unclamp the tyre. The fixed ring is equipped with 8 guide wheels, which support the fixed ring and guide it through a certain angle of rotation, while the rotation of the fixed ring and the turning of the machine are controlled by an electric motor.
86
87
===3.2 Numerical simulation of the operating conditions of the tyre unloader===
88
89
The above 3D model of the tyre unloader (including tyre) is imported into the dynamics analysis software ADAMS, and the material parameters of the model are set first. The material parameters of the whole machine and tyres are shown in [[#tab-1|Tables 1]] and [[#tab-2|2]].
90
91
<div class="center" style="font-size: 75%;">'''Table 1'''. Material parameters for tyre unloaders</div>
92
93
<div id='tab-1'></div>
94
{| class="wikitable" style="margin: 1em auto 0.1em auto;border-collapse: collapse;font-size:85%;width:auto;" 
95
|-style="text-align:center"
96
!  Materials !! Elastic modulus (MPa) !! Poisson ratio !! Density (kg/m<sup>3</sup>)
97
|-style="text-align:center"
98
|  Rubber
99
|  7.8
100
|  0.29
101
|  1200
102
|}
103
104
105
<div class="center" style="font-size: 75%;">'''Table 2'''.  Tyre material parameters></div>
106
107
{| class="wikitable" style="margin: 1em auto 0.1em auto;border-collapse: collapse;font-size:85%;width:auto;" 
108
|-style="text-align:center"
109
!  Materials !! Elastic modulus (MPa) !! Poisson ratio !! Density (kg/m<sup>3</sup>)
110
|-style="text-align:center"
111
|  Carbon structural steel
112
|  207
113
|  0.29
114
|  7801
115
|}
116
117
118
Constraint and actuation of each mechanism according to the actual motion of the tyre unloading machine. The whole mechanism of tyre unloading machine is mainly rotating sub and the connecting rod articulation, plywood and rocker arm and other rotating connections are rotating sub.The guidewheel shaft is fixed to the fixed ring using a fixed pair, and a rotating connection pair is applied between the guidewheel and the guidewheel shaft.The fixed ring and the moving ring are provided with a planar pair to prevent relative slippage caused by the rotation of the moving ring. To prevent over-constraints caused by the use of too many rotating subsets, resulting in instability of the solved system,the conflict constraints are replaced by primary subsets such as point overlap subsets or cylindrical amplitude, and coplanar subsets replace some of these planar subsets to ensure the stability and accuracy of the solved system.The virtual prototype of the tyre unloader is shown in [[#img-4|Figure 4]].
119
120
<div id='img-4'></div>
121
{| class="wikitable" style="margin: 1em auto 0.1em auto;border-collapse: collapse;width:auto;" 
122
|-style="background:white;"
123
|style="text-align: center;padding:10px;"| [[Image:Draft_Ling_717181314-image4.png|441x441px]]
124
|-
125
| style="background:#efefef;text-align:left;padding:10px;font-size: 85%;"| '''Figure 4'''. Virtual prototype of the tyre unloader 
126
|}
127
128
129
After building a virtual prototype of the tyre unloader system, the simulation is based on the actual working conditions of the tyre unloader. The simulation time is set as a cycle of 2.1s. After the simulation, the load spectrum of <math display="inline">X  </math>, <math display="inline"> Y </math> and <math display="inline"> Z </math> directions of the bearing at the bottom rocker arm is extracted, and the extraction position is illustrated in [[#img-5|Figure 5]]. The loads in the <math display="inline">X  </math> and <math display="inline"> Y </math> directions are transverse and longitudinal loads, and the loads in the <math display="inline"> Z </math> direction are those perpendicular to the bearing outward. The extraction results are shown in [[#img-6|Figure 6]].
130
131
<div id='img-5'></div>
132
{| class="wikitable" style="margin: 1em auto 0.1em auto;border-collapse: collapse;width:auto;" 
133
|-style="background:white;"
134
|style="text-align: center;padding:10px;"| [[Image:Draft_Ling_717181314-image5.png]]
135
|-
136
| style="background:#efefef;text-align:left;padding:10px;font-size: 85%;"| '''Figure 5'''. Load extraction at rocker arm bearing 
137
|}
138
139
140
<div id='img-6'></div>
141
{| class="wikitable" style="margin: 1em auto 0.1em auto;border-collapse: collapse;width:auto;" 
142
|-style="background:white;"
143
|style="text-align: center;padding:10px;"| [[Image:Draft_Ling_717181314-image6.png|500px]]
144
|-
145
| style="background:#efefef;text-align:left;padding:10px;font-size: 85%;"| '''Figure 6'''. Three-way force load spectrum at rocker arm bearing
146
|}
147
148
149
According to the three-way force load spectrum, the average load in <math display="inline">X  </math> direction and <math display="inline"> Y </math> direction of the bearing tends to 0 within 0.8s, and there is a certain impact load, which is caused by the vibration generated when the splint contacts the tyre. After the splint clamps the tyre, the fixing pair will temporarily fail. As the weight of the tyre itself is transferred to the bearing, the load values of the <math display="inline">X  </math> and <math display="inline"> Y </math> directions of the bearing at the bottom rocker arm will rise rapidly. After about 0.2s, the components of the <math display="inline">X  </math> and <math display="inline"> Y </math> directions increase and remain at about 108kN and 78kN, respectively.
150
151
===3.3 Validation of simulation results===
152
153
Because the simulation analysis model is simplified to some extent, the simulation results have some errors.Therefore, in order to verify the reasonableness and realism of the simulation results and to provide a reasonable dynamic load spectrum for the subsequent analysis and prediction of the bearing reliability. Through the proportional coefficient of the real-time output torque of the clamping motor on the control panel of the tyre unloader ([[#img-7|Figure 7]]), the output torque coefficient of the motor is converted into the actual moving ring thrust of the tyre unloader and compared with the simulation value of ADAMS, and the reasons for the error between the actual value and the simulation result are analyzed.
154
155
<div id='img-7'></div>
156
{| class="wikitable" style="margin: 1em auto 0.1em auto;border-collapse: collapse;width:auto;" 
157
|-style="background:white;"
158
|style="text-align: center;padding:10px;"| [[Image:Draft_Ling_717181314-image7-c.png]]
159
|-
160
| style="background:#efefef;text-align:left;padding:10px;font-size: 85%;"| '''Figure 7'''. Tyre unloader control panel
161
|}
162
163
164
The comparison data in the test is the thrust of the tyre unloading maneuver ring, which is the current torque value of the birth ring in the picture. The output proportional coefficient needs to be converted into the real thrust force on the moving ring, namely the lead screw thrust. The relation between output torque of reducer and motor parameters is shown in Eq. (1) 
165
166
{| class="formulaSCP" style="width: 100%; text-align: center;" 
167
|-
168
| 
169
{| style="text-align: center; margin:auto;" 
170
|-
171
| <math>T_a=9550\frac{P\cdot r\cdot \eta }{n}</math>
172
|}
173
| style="width: 5px;text-align: right;white-space: nowrap;" | (1)
174
|}
175
176
where <math display="inline"> P </math> represents the motor power, which is 0.75kW, <math display="inline"> N  </math> represents the motor speed, which is 1500r/min, 
177
<math display="inline">\eta</math> represents the transmission efficiency, which is 98%, and <math display="inline"> R</math> represents the speed ratio of motor and reducer, which is 20:1.
178
179
The maximum output torque <math display="inline">T_a</math> of reducer is 94N m.
180
181
The conversion formula between output torque of reducer and thrust Fa of the lead screw is shown in Eq. (2). The maximum thrust force Fa of the screw is calculated from the maximum output torque to be 56 N. The motor torque output scaling factor is converted into a value for the change in thrust of the kinematic ring, the torque factor and the actual kinematic ring thrust values are shown in [[#tab-3|Table 3]] 
182
183
{| class="formulaSCP" style="width: 100%; text-align: center;" 
184
|-
185
| 
186
{| style="text-align: center; margin:auto;" 
187
|-
188
| <math>T_a=\frac{F_a\times t}{2\times 3.14\times {\eta }_1}</math>
189
|}
190
| style="width: 5px;text-align: right;white-space: nowrap;" | (2)
191
|}
192
193
where <math>t</math> is the lead of the lead screw, which is 10mm, and <math>{\eta }_1</math> represents the conversion efficiency, which is 95%.
194
195
<div class="center" style="font-size: 75%;">'''Table 3'''.  Proportional coefficient of output torque and actual thrust value of moving ring (part)</div>
196
197
<div id='tab-1'></div>
198
{| class="wikitable" style="margin: 1em auto 0.1em auto;border-collapse: collapse;font-size:85%;width:auto;" 
199
|-style="text-align:center"
200
! Proportional coefficient of output torque (‰) !! Actual moving ring thrust (N)
201
|-style="text-align:center"
202
|  232
203
|  12.99 
204
|-style="text-align:center"
205
|  239
206
|  13.38 
207
|-style="text-align:center"
208
|  244
209
|  13.66 
210
|-style="text-align:center"
211
|  250
212
|  14.00 
213
|-style="text-align:center"
214
|  258
215
|  14.45 
216
|-style="text-align:center"
217
|  262
218
|  14.67 
219
|-style="text-align:center"
220
|  269
221
|  15.06 
222
|-style="text-align:center"
223
|  276
224
|  15.46 
225
|-style="text-align:center"
226
|  283
227
|  15.85 
228
|-style="text-align:center"
229
|  289
230
|  16.18 
231
|-style="text-align:center"
232
|  295
233
|  16.52 
234
|-style="text-align:center"
235
|  304
236
|  17.02 
237
|-style="text-align:center"
238
|  302
239
|  16.91 
240
|-style="text-align:center"
241
|  309
242
|  17.30 
243
|-style="text-align:center"
244
|  317
245
|  17.75 
246
|-style="text-align:center"
247
|  321
248
|  17.98 
249
|-style="text-align:center"
250
|  327
251
|  18.31 
252
|}
253
254
255
The drive torque was simulated by ADAMS with the drive position at the center of the tyre unloader. Based on the actual radius of the tyre unloader being approximately 1 meter, it can be assumed that the dynamic ring thrust value is numerically equal to the drive torque value. The actual calculated dynamic ring thrust variation was compared with the simulated value of the tyre unloader motor drive ring and the comparison results are shown in [[#img-8|Figure 8]].
256
257
<div id='img-8'></div>
258
{| class="wikitable" style="margin: 1em auto 0.1em auto;border-collapse: collapse;width:auto;" 
259
|-style="background:white;"
260
|style="text-align: center;padding:10px;"| [[Image:Draft_Ling_717181314-image10.png|500px]]
261
|-
262
| style="background:#efefef;text-align:left;padding:10px;font-size: 85%;"| '''Figure 8'''. Comparison of simulated and actual values
263
|}
264
265
266
Due to the difference between the simulated and actual driving methods of the dynamic loops, the simulated and actual values are subject to certain errors. In ADAMS, a rotational drive is applied directly to the kinematic ring, and the simulation results from [[#img-8|Figure 8]] show that the torque required to move the kinematic ring from rest to motion is large (0 to 0.1s), after which the drive torque required decreases (0.1 to 7.9s) due to the inertia of the kinematic ring itself. To maintain the movement, the torque will gradually increase, during which the collision between the moving ring and the guide wheel will also cause a sudden increase in torque, producing a peak (7.9 to 14s). In the actual working of the tyre unloader, the drive unit consists of three main parts: motor, reducer and screw. The specific mode of operation is by connecting the reducer to the motor in order to reduce the rotational speed and thus increase the output torque, which is converted into thrust on the screw by means of a screw to drive the moving ring for rotation. This process allows for customised settings of motor speed, output torque magnitude, etc. The torque magnitude is negligibly influenced by the interaction between the mechanisms during operation of the device. So the screw thrust tends to rise gradually and with little fluctuation. But the simulated force values are of the same order of magnitude as the actual force values, and the overall trend is gradually increasing, which proves that the simulation results have a certain degree of reasonableness.
267
268
===3.4 Bearing transient dynamics analysis===
269
270
From the results of the kinetic analysis, it can be seen that the bearing at the lowermost rocker is subjected to the greatest component of gravity of the tyre blank and is also furthest from the centre of rotation, where the centrifugal inertia force is greatest. The bearing selected here for analysis is therefore a deep groove ball bearing, type 61918. The bearing model is shown in [[#img-9|Figure 9]]. According to the actual bearing size parameters, the 3D modelling software is used to establish a 3D model of the bearing at the rocker arm. In order to simplify the calculation steps and save calculation resources, non-important parts such as the cage are omitted, the bearing dimensional parameters are shown in [[#tab-4|Table 4]].
271
272
<div id='img-9'></div>
273
{| class="wikitable" style="margin: 1em auto 0.1em auto;border-collapse: collapse;width:auto;" 
274
|-style="background:white;"
275
|style="text-align: center;padding:10px;"| [[Image:Draft_Ling_717181314-image11.png]]
276
|-
277
| style="background:#efefef;text-align:left;padding:10px;font-size: 85%;"| '''Figure 9'''. 3D Model of bearing 61918
278
|}
279
280
281
<div class="center" style="font-size: 75%;">'''Table 4'''. Bearing 61918 dimensional parameters</div>
282
283
<div id='tab-1'></div>
284
{| class="wikitable" style="margin: 1em auto 0.1em auto;border-collapse: collapse;font-size:85%;width:auto;" 
285
|-style="text-align:center"
286
! Bearing type !! Outer diameter (mm) !! Inner diameter (mm) !! Thickness (mm)
287
|-style="text-align:center"
288
|  61918
289
|  125
290
|  90
291
|  18
292
|}
293
294
295
The bearing model was imported into Ansys finite element analysis software, the inner and outer rings of the bearing were constrained, and the bearing material parameters were set, as shown in [[#tab-5|Table 5]]. The extracted <math display="inline"> X </math>, <math display="inline"> Y </math> and <math display="inline"> Z </math> forces are loaded onto the corresponding parts of the bearing for transient dynamics analysis, with the loads and constraints applied as shown in [[#img-10|Figure 10]]. The results of the transient dynamics analysis (equivalent stress clouds) are shown in [[#img-11|Figure 11]].
296
297
<div class="center" style="font-size: 75%;">'''Table 5'''. Bearing material parameters</div>
298
299
<div id='tab-1'></div>
300
{| class="wikitable" style="margin: 1em auto 0.1em auto;border-collapse: collapse;font-size:85%;width:auto;" 
301
|-style="text-align:center"
302
! Materials !! Elastic Modulus (MPa) !! Poisson ratio !! Yield Strength (MPa) !! Tensile strength (MPa)
303
|-style="text-align:center"
304
|  GCr15
305
|  210000
306
|  0.29
307
|  1458
308
|  1617
309
|}
310
311
312
<div id='img-10'></div>
313
{| class="wikitable" style="margin: 1em auto 0.1em auto;border-collapse: collapse;width:auto;" 
314
|-style="background:white;"
315
|style="text-align: center;padding:10px;"| [[Image:Draft_Ling_717181314-image12.png|709x709px]]
316
|-
317
| style="background:#efefef;text-align:left;padding:10px;font-size: 85%;"| '''Figure 10'''. Bearing load and constraint settings
318
|}
319
320
321
<div id='img-11'></div>
322
{| class="wikitable" style="margin: 1em auto 0.1em auto;border-collapse: collapse;width:auto;" 
323
|-style="background:white;"
324
|style="text-align: center;padding:10px;"| [[Image:Draft_Ling_717181314-image13.png|552x552px]]
325
|-
326
| style="background:#efefef;text-align:left;padding:10px;font-size: 85%;"| '''Figure 11'''. Cloud diagram of maximum equivalent stress of bearing
327
|}
328
329
330
The bearing of the lowermost rocker arm of the tyre unloader is subjected to the greatest inertial force when carrying the weight of the entyre tyre and turning it over. From the results of the transient dynamics analysis, it can be seen that the maximum stress value that the bearing is subjected to is 1316.2Mpa, so the part of the bearing where the stress is high and easy to produce fatigue is the inner ring and ball contact.
331
332
===3.5 Stress history in hazardous areas===
333
334
In order to obtain the distribution characteristics of the stresses in the hazardous parts of the bearings, the ANSYS APDL was used to extract the nodal stress time histories corresponding to the fatigue parts of the bearings according to the results of the transient dynamics. The curve for the change in stress history is shown in [[#img-12|Figure 12]].
335
336
<div id='img-12'></div>
337
{| class="wikitable" style="margin: 1em auto 0.1em auto;border-collapse: collapse;width:auto;" 
338
|-style="background:white;"
339
|style="text-align: center;"| [[Image:Draft_Ling_717181314-image14.png|504px]]
340
|-
341
| style="background:#efefef;text-align:left;padding:10px;font-size: 85%;"| '''Figure 12'''. Stress time history in fatigue bearing areas
342
|}
343
344
==4. Bearing reliability calculation based on heterogeneous interference model==
345
346
===4.1 Stress history rain flow count statistics===
347
348
From the basic theory of fatigue, it can be seen that the load average, amplitude and the number of cycles are the main factors that make the component produce fatigue damage, so the bearing fatigue part of the stress time course need to be cycle counting process. In this paper, the two-parameter rainfall counting method is used to count the amplitude and mean values of the loads and to obtain important relationships between the load amplitude, mean value and the corresponding frequency.
349
350
Rain flow counting of the stress history was carried out to convert the random variable amplitude stress into a series of load cycles, the amplitude and mean distribution of the load cycles were counted and the mean and amplitude were converted into a two-dimensional histogram and the statistical results are shown in [[#img-13|Figures 13]], [[#img-14|14]] and [[#img-15|15]].
351
352
<div id='img-13'></div>
353
{| class="wikitable" style="margin: 1em auto 0.1em auto;border-collapse: collapse;width:auto;" 
354
|-style="background:white;"
355
|style="text-align: center;padding:10px;"| [[Image:Draft_Ling_717181314-image15-c.png|500px]]
356
|-
357
| style="background:#efefef;text-align:left;padding:10px;font-size: 85%;"| '''Figure 13'''. Stress history rain flow counting
358
|}
359
360
361
<div id='img-14'></div>
362
{| class="wikitable" style="margin: 1em auto 0.1em auto;border-collapse: collapse;width:auto;" 
363
|-style="background:white;"
364
|style="text-align: center;padding:10px;"| [[Image:Draft_Ling_717181314-image16.png|500px]]
365
|-
366
| style="background:#efefef;text-align:left;padding:10px;font-size: 85%;"| '''Figure 14'''. Stress amplitude frequency statistics
367
|}
368
369
370
<div id='img-15'></div>
371
{| class="wikitable" style="margin: 1em auto 0.1em auto;border-collapse: collapse;width:auto;" 
372
|-style="background:white;"
373
|style="text-align: center;padding:10px;"| [[Image:Draft_Ling_717181314-image17.png|500px]]
374
|-
375
| style="background:#efefef;text-align:left;padding:10px;font-size: 85%;"| '''Figure 15'''. Stress mean frequency statistics
376
|}
377
378
379
Using Goodman's theory to apply an average stress correction to the stress amplitude, the stress state is converted to a load cycle with a stress ratio of -1, according to an equal life, the corrected load cycles were used as equivalent stresses for the next step of the fatigue reliability analysis. The Goodman formula is shown in Eq. (3) and the corrected results are shown in [[#img-16|Figure 16]]
380
381
{| class="formulaSCP" style="width: 100%; text-align: center;" 
382
|-
383
| 
384
{| style="text-align: center; margin:auto;" 
385
|-
386
| <math>\frac{{\sigma }_a}{{\sigma }_{-1}}+\frac{{\sigma }_m}{{\sigma }_b}=1</math>
387
|}
388
| style="width: 5px;text-align: right;white-space: nowrap;" | (3)
389
|}
390
391
where <math>{\sigma }_a</math> denotes the actual working stress amplitude (MPa), <math>{\sigma }_m</math> indicates the average stress in actual working conditions (Mpa), <math>{\sigma }_b</math>  indicates the tensile ultimate strength of the material (MPa), and <math>{\sigma }_{-1}</math> denotes the stress magnitude (MPa) for a stress ratio of -1.
392
393
<div id='img-16'></div>
394
{| class="wikitable" style="margin: 1em auto 0.1em auto;border-collapse: collapse;width:auto;" 
395
|-style="background:white;"
396
|style="text-align: center;padding:10px;"| [[Image:Draft_Ling_717181314-image19.png]]
397
|-
398
| style="background:#efefef;text-align:left;padding:10px;font-size: 85%;"| '''Figure 16'''. Modified equivalent load frequency statistics
399
|}
400
401
===4.2 Load amplitude probability distribution fitting and test===
402
403
Due to the influence of factors such as short sampling time and large amount of data in compiling load spectrum, the load history obtained cannot fully reflect the actual load history of the bearing in the whole life stage, therefore, in order to improve the reliability of the results, the probability distribution function of random loads should be calculated, and then the load distribution of bearings in the whole life interval should be predicted.
404
405
According to the rain-flow counting statistics of stress time history in fatigue parts of bearings, the probability statistics of stress amplitude are carried out and fitted by Weibull distribution. The fitting effect and test results are shown in [[#img-17|Figure 17]]. It can be seen that the distribution of stress amplitudes generally conforms to the Weibull distribution.
406
407
<div id='img-17'></div>
408
{| class="wikitable" style="margin: 1em auto 0.1em auto;border-collapse: collapse;width:auto;" 
409
|-style="background:white;"
410
|style="text-align: center;padding:10px;"| [[Image:Draft_Ling_717181314-image20.png|500px]]
411
|-
412
| style="background:#efefef;text-align:left;padding:10px;font-size: 85%;"| '''Figure 17'''. Fitting magnitude of the Weibull distribution
413
|}
414
415
416
In order to verify whether the stress amplitude conforms to Weibull distribution, [[#img-18|Figure 18]] is obtained by the graphical method. It can be concluded from the figure that the closer the data is to a straight line, the better the data fitting effect is. Therefore, it can be considered that the probability density function of stress amplitude conforms to Weibull distribution.
417
418
<div id='img-18'></div>
419
{| class="wikitable" style="margin: 1em auto 0.1em auto;border-collapse: collapse;width:auto;" 
420
|-style="background:white;"
421
|style="text-align: center;padding:10px;"| [[Image:Draft_Ling_717181314-image21.png|500px]]
422
|-
423
| style="background:#efefef;text-align:left;padding:10px;font-size: 85%;"| '''Figure 18'''. Weibull diagramming test
424
|}
425
426
427
Using the great likelihood method to solve for the Weibull correlation parameters, the relevant solution parameters are shown in [[#tab-6|Table 6]].
428
429
<div class="center" style="font-size: 75%;">'''Table 6'''. Equivalent load distribution parameters</div>
430
431
<div id='tab-6'></div>
432
{| class="wikitable" style="margin: 1em auto 0.1em auto;border-collapse: collapse;font-size:85%;width:auto;" 
433
|-style="text-align:center"
434
! Scale parameterse parameters !! Shape parameters !! Confidence level
435
|-style="text-align:center"
436
|  61.93
437
|  0.729
438
|  0.05
439
|}
440
441
442
The equivalent load probability density function is:
443
444
{| class="formulaSCP" style="width: 100%; text-align: center;" 
445
|-
446
| 
447
{| style="text-align: center; margin:auto;" 
448
|-
449
| <math>f\left(x\right)=\frac{\beta x^{\beta -1}}{{\theta }^{\beta }}\exp\left[- {\left(\frac{x}{\theta }\right)}^{\beta }\right]</math> 
450
|}
451
| style="width: 5px;text-align: right;white-space: nowrap;" | (4)
452
|}
453
454
where <math>\beta</math>  denotes the shape parameter, and <math>\theta</math>  denotes the scale parameter. Substituting the two into Eq. (4) gives:
455
456
{| class="formulaSCP" style="width: 100%; text-align: center;" 
457
|-
458
| 
459
{| style="text-align: center; margin:auto;" 
460
|-
461
| <math>f\left(S\right)=\frac{0.72S^{-0.28}}{19.51}\exp\left[- {\left(\frac{S}{61.93}\right)}^{0.72}\right]</math>
462
|}
463
| style="width: 5px;text-align: right;white-space: nowrap;" | (5)
464
|}
465
466
===4.3 Material life distribution of bearing steel under arbitrary stress===
467
468
A set of fatigue test data for bearing steel was obtained from the literature [22] and combined with Basquin's equation to derive a continuous probability life model for bearing steel using the great likelihood method. The test material was a high-carbon chromium bearing steel with the chemical composition shown in [[#tab-7|Table 7]], which was machined to give the dimensions in [[#img-19|Figure 19]]. Then using the grouping method, a four-linked cantilever beam type rotating bending fatigue tester was used to carry out the test, starting with a load of 1700 MPa stress. By reducing the weights so that the stresses fell at intervals of 50 to 100 MPa, nine sets of fatigue test data were eventually obtained, as shown in [[#tab-8|Table 8]].
469
470
<div class="center" style="font-size: 75%;">'''Table 7'''. Chemical composition of high-carbon chromium bearing steel</div>
471
472
<div id='tab-7'></div>
473
{| class="wikitable" style="margin: 1em auto 0.1em auto;border-collapse: collapse;font-size:85%;width:auto;" 
474
|-style="text-align:center"
475
! Element !! C !! Si !! Mn !! Cr !! Cu !! Ni !! Mo !! P !! S
476
|-style="text-align:center"
477
|  Content %
478
|  1.01
479
|  0.23
480
|  0.36
481
|  1.45
482
|  0.06
483
|  0.04
484
|  0.02
485
|  0.01
486
|  0.007
487
|}
488
489
490
<div id='img-19'></div>
491
{| class="wikitable" style="margin: 1em auto 0.1em auto;border-collapse: collapse;width:auto;" 
492
|-style="background:white;"
493
|style="text-align: center;padding:10px;"| [[Image:Draft_Ling_717181314-image24.jpeg|500px]]
494
|-
495
| style="background:#efefef;text-align:left;padding:10px;font-size: 85%;"| '''Figure 19'''. Test steel dimensions
496
|}
497
498
499
<div class="center" style="font-size: 75%;">'''Table 8'''. Fatigue test date</div>
500
501
<div id='tab-8'></div>
502
{| class="wikitable" style="margin: 1em auto 0.1em auto;border-collapse: collapse;font-size:85%;width:auto;" 
503
|-style="text-align:center"
504
!Stress (MPa) !! Life (cycle) !! Stress (MPa) !! Life (cycle)
505
|-style="text-align:center"
506
| rowspan="3" | 1700
507
|4480
508
| rowspan="2" |1300
509
|2480000
510
|-style="text-align:center"
511
|6100
512
|3870000
513
|-style="text-align:center"
514
|6840
515
| rowspan="10" |1250
516
|40560
517
|-style="text-align:center"
518
| rowspan="3" |1600
519
|5200
520
|52430
521
|-style="text-align:center"
522
|7790
523
|65990
524
|-style="text-align:center"
525
|10000
526
|128800
527
|-style="text-align:center"
528
| rowspan="5" |1500
529
|8990
530
|2486740
531
|-style="text-align:center"
532
|14400
533
|4983540
534
|-style="text-align:center"
535
|34700
536
|8950470
537
|-style="text-align:center"
538
|247000
539
|10325090
540
|-style="text-align:center"
541
|609000
542
|11730550
543
|-style="text-align:center"
544
| rowspan="7" |1400
545
|13460
546
|26058040
547
|-style="text-align:center"
548
|15900
549
| rowspan="8" |1200
550
|125280
551
|-style="text-align:center"
552
|17220
553
|269929
554
|-style="text-align:center"
555
|40800
556
|672410
557
|-style="text-align:center"
558
|839140
559
|8771540
560
|-style="text-align:center"
561
|1540000
562
|19548550
563
|-style="text-align:center"
564
|4970760
565
|41502050
566
|-style="text-align:center"
567
| rowspan="4" |1300
568
|15360
569
|43386180
570
|-style="text-align:center"
571
|22670
572
|61683430
573
|-style="text-align:center"
574
|39830
575
| rowspan="2" |1150
576
|23276820
577
|-style="text-align:center"
578
|2050180
579
|47700000
580
|}
581
582
583
Due to some differences in the actual size, shape and working conditions of the tyre unloader bearings used in this study and the shape and loading method used for the test specimens, therefore, the S-N relationship for the test material needs to be corrected to the S-N relationship for the actual zero component, and the correction formula is as follows:
584
585
{| class="formulaSCP" style="width: 100%; text-align: center;" 
586
|-
587
| 
588
{| style="text-align: center; margin:auto;" 
589
|-
590
| <math>S_a=\frac{{\sigma }_a}{K_f}\epsilon \beta C_L</math>
591
|}
592
| style="width: 5px;text-align: right;white-space: nowrap;" | (6)
593
|}
594
595
where <math display="inline">{\sigma }_{a}</math> corresponds to the stress tested in the material, <math display="inline">{S}_{a}</math> corresponds to zero component stress, 
596
<math display="inline">\epsilon</math>  is the dimensional coefficient of the material, the value of 0.856 was taken by consulting the «Mechanical Design Manual», <math display="inline">\beta</math> is the material surface quality factor, here taken as 1, and
597
<math display="inline">{C}_{L}</math> is the loading method for the workpiece and the steel tension is taken as 0.85.
598
599
The fatigue notch coefficient <math display="inline">K_f</math> is related to the stress concentration coefficient, the bearing stress concentration is generally related to the roughness, here it can be considered that the bearing is consistent with the roughness of the test piece, take the value of 1, the correction results are shown in [[#tab-9|Table 9]].
600
601
<div class="center" style="font-size: 75%;">'''Table 9'''. Fatigue test data after stress correction</div>
602
603
<div id='tab-9'></div>
604
{| class="wikitable" style="margin: 1em auto 0.1em auto;border-collapse: collapse;font-size:85%;width:auto;" 
605
|-style="text-align:center"
606
! Test stress (MPa) !! Corrected stress (MPa) 
607
|-style="text-align:center"
608
|  1700
609
|  1030
610
|-style="text-align:center"
611
|  1600
612
|  970
613
|-style="text-align:center"
614
|  1500
615
|  909
616
|-style="text-align:center"
617
|  1400
618
|  849
619
|-style="text-align:center"
620
|  1300
621
|  788
622
|-style="text-align:center"
623
|  1250
624
|  758
625
|-style="text-align:center"
626
|  1200
627
|  728
628
|-style="text-align:center"
629
|  1150
630
|  697
631
|}
632
633
===4.4 Maximum likelihood method for determining the P-S-N curve of bearing steel===
634
635
The general material fatigue life follows a log-normal distribution. Assuming that the bearing steel life follows a log-normal distribution, the Anderson-Darling hypothesis test was performed on the fatigue test data at different stresses in [[#tab-8|Table 8]]. As the A-D test can take different confidence intervals for testing, the smaller the confidence interval, the more obvious the test effect. So 95% confidence intervals were taken for hypothesis testing of each group of fatigue data, and the test results are shown in [[#img-20|Figures 20]] and  [[#img-21|21]] (partial) and [[#tab-10|Table 10]] (all).
636
637
<div id='img-20'></div>
638
{| class="wikitable" style="margin: 1em auto 0.1em auto;border-collapse: collapse;width:auto;" 
639
|-style="background:white;"
640
|style="text-align: center;padding:10px;"| [[Image:Draft_Ling_717181314-image26-c.png]]
641
|-
642
| style="background:#efefef;text-align:left;padding:10px;font-size: 85%;"| '''Figure 20'''. A-D test results for material life at a stress of 1030 MPa
643
|}
644
645
646
<div id='img-21'></div>
647
{| class="wikitable" style="margin: 1em auto 0.1em auto;border-collapse: collapse;width:auto;" 
648
|-style="background:white;"
649
|style="text-align: center;padding:10px;"| [[Image:Draft_Ling_717181314-image27-c.png]] 
650
|-
651
| style="background:#efefef;text-align:left;padding:10px;font-size: 85%;"| '''Figure 21'''. A-D test results for material life at a stress of 909 MPa
652
|}
653
654
655
<div class="center" style="font-size: 75%;">'''Table 10'''. Results of the A-D test for life under different stresses</div>
656
657
<div id='tab-10'></div>
658
{| class="wikitable" style="margin: 1em auto 0.1em auto;border-collapse: collapse;font-size:85%;width:auto;" 
659
|-style="text-align:center"
660
! Proof stress (MPa) !! AD !! P value
661
|-style="text-align:center"
662
|  1030
663
|  0.264
664
|  0.371
665
|-style="text-align:center"
666
|  970
667
|  0.210
668
|  0.544
669
|-style="text-align:center"
670
|  909
671
|  0.279
672
|  0.484
673
|-style="text-align:center"
674
|  849
675
|  0.562
676
|  0.091
677
|-style="text-align:center"
678
|  788
679
|  0.603
680
|  0.062
681
|-style="text-align:center"
682
|  758
683
|  0.735
684
|  0.037
685
|-style="text-align:center"
686
|  728
687
|  0.707
688
|  0.042
689
|-style="text-align:center"
690
|  697
691
|  0.250
692
|  0.277
693
|}
694
695
696
The test shows that the p-value is greater than the significant level of 0.05, which is in line with the hypothesis test, and therefore the bearing steel material life can be considered to be in line with the log-normal distribution.
697
698
The S-N curve of the bearing is estimated using Basquin's equation combined with the method of great likelihood, Basquin's equation is
699
700
{| class="formulaSCP" style="width: 100%; text-align: center;" 
701
|-
702
| 
703
{| style="text-align: center; margin:auto;" 
704
|-
705
| <math>S^m N=C</math>
706
|}
707
| style="width: 5px;text-align: right;white-space: nowrap;" | (7)
708
|}
709
710
where <math display="inline"> m </math> is the exponent, <math display="inline"> N </math> is the fatigue life and <math display="inline"> C </math> is a constant.
711
712
Taking the logarithm of both sides gives:
713
714
{| class="formulaSCP" style="width: 100%; text-align: center;" 
715
|-
716
| 
717
{| style="text-align: center; margin:auto;" 
718
|-
719
| <math>\lg N_p=\lg C_p-m \lg S</math>
720
|}
721
| style="width: 5px;text-align: right;white-space: nowrap;" | (8)
722
|}
723
724
where the subscript <math display="inline"> p </math> indicates the survival rate.
725
726
When the survival rate is 50%, the median log life is equal to the mean of the log life:
727
728
{| class="formulaSCP" style="width: 100%; text-align: center;" 
729
|-
730
| 
731
{| style="text-align: center; margin:auto;" 
732
|-
733
| <math>\mu (S)=\lg N_{50}=\lg C-m_{50}\lg S</math>
734
|}
735
| style="width: 5px;text-align: right;white-space: nowrap;" | (9)
736
|}
737
738
where <math display="inline">\mu (S)</math> is the mean log life of the sample at different stress.
739
740
When <math display="inline">S=788</math>MPa, the mean value of the log-life sample with stress of 788MPa is substituted as the parent mean value in Eqs. (8) and (9) to obtain
741
742
{| class="formulaSCP" style="width: 100%; text-align: center;" 
743
|-
744
| 
745
{| style="text-align: center; margin:auto;" 
746
|-
747
| <math>\lg C=2.90m_{50}+5.4061</math>
748
|}
749
| style="width: 5px;text-align: right;white-space: nowrap;" | (10)
750
|}
751
752
{| class="formulaSCP" style="width: 100%; text-align: center;" 
753
|-
754
| 
755
{| style="text-align: center; margin:auto;" 
756
|-
757
| <math>\mu (S)=5.4060\mbox{-}m_{50}\lg S\mbox{+}2.90m_{50}</math>
758
|}
759
| style="width: 5px;text-align: right;white-space: nowrap;" | (11)
760
|}
761
762
763
The log life of bearing steel follows a normal distribution and the mean log life is related to the standard deviation of the log life at different survival rates as follows:
764
765
{| class="formulaSCP" style="width: 100%; text-align: center;" 
766
|-
767
| 
768
{| style="text-align: center; margin:auto;" 
769
|-
770
| <math>\mu (S)-\lg N_p={\upsilon }_p\sigma (S)</math>
771
|}
772
| style="width: 5px;text-align: right;white-space: nowrap;" | (12)
773
|}
774
775
where <math display="inline">\sigma (S)</math> is the standard deviation of the log life of the member at different stress. <math display="inline">{\upsilon }_{P}</math> is the standard normal deviation corresponding to the probability of damage.
776
777
Substituting Eqs. (8) and (9) into Eq. (12) yields:
778
779
{| class="formulaSCP" style="width: 100%; text-align: center;" 
780
|-
781
| 
782
{| style="text-align: center; margin:auto;" 
783
|-
784
| <math>\sigma (S)={\left(\lg C-\lg C_p\right)}^{\frac{1}{{\upsilon }_p}}-\frac{m_{50}}{{\upsilon }_p}\lg S+\frac{m_p}{{\upsilon }_p}\lg S</math>
785
|}
786
| style="width: 5px;text-align: right;white-space: nowrap;" | (13)
787
|}
788
789
790
The sample log life standard deviation at a stress of 788Mpa is substituted into Eq. (13) as the parent log life standard deviation:
791
792
{| class="formulaSCP" style="width: 100%; text-align: center;" 
793
|-
794
| 
795
{| style="text-align: center; margin:auto;" 
796
|-
797
| <math>\sigma (788)=\sqrt{\frac{1}{\mbox{n-1}}\sum_{i=1}^n{\left(\lg N_i-\mu (788)\right)}^2}\mbox{=}{\left(\lg C-\lg C_p\right)}^{\frac{1}{{\upsilon }_p}}-</math><math>\frac{m_{50}}{{\upsilon }_p}\lg 788+\frac{m_p}{{\upsilon }_p}\lg 788</math> 
798
|}
799
| style="width: 5px;text-align: right;white-space: nowrap;" | (14)
800
|}
801
802
803
Also known as equation:
804
805
{| class="formulaSCP" style="width: 100%; text-align: center;" 
806
|-
807
| 
808
{| style="text-align: center; margin:auto;" 
809
|-
810
| <math>\sigma (S)=\sigma (788)\mbox{+}\frac{m_{50}}{{\upsilon }_p}\lg 788\mbox{-}\frac{m_p}{{\upsilon }_p}\lg 788-</math><math>\frac{m_{50}}{{\upsilon }_p}\lg S+\frac{m_p}{{\upsilon }_p}\lg S</math>
811
|}
812
| style="width: 5px;text-align: right;white-space: nowrap;" | (15)
813
|}
814
815
When the survival rate <math>P = 84.1%</math>, <math>u_p =1</math>, which is collated to give:
816
817
{| class="formulaSCP" style="width: 100%; text-align: center;" 
818
|-
819
| 
820
{| style="text-align: center; margin:auto;" 
821
|-
822
| <math>\sigma (S)=1.1344+m_{50}\lg S+m_{84.1}\lg S</math>
823
|}
824
| style="width: 5px;text-align: right;white-space: nowrap;" | (16)
825
|}
826
827
When <math>N</math> follows a log-normal distribution, its probability density function is
828
829
{| class="formulaSCP" style="width: 100%; text-align: center;" 
830
|-
831
| 
832
{| style="text-align: center; margin:auto;" 
833
|-
834
| <math>f(N)=\frac{1}{\sigma (S)\sqrt{2\pi }N}\exp\left\{-\frac{{\left(\ln N-\mu (S)\right)}^2}{2{\sigma }^2(S)}\right\}</math>
835
|}
836
| style="width: 5px;text-align: right;white-space: nowrap;" | (17)
837
|}
838
839
Substituting the different stresses and corresponding lifetimes into Eq. (17) and multiplying them together, the likelihood function is obtained as equation
840
841
{| class="formulaSCP" style="width: 100%; text-align: center;" 
842
|-
843
| 
844
{| style="text-align: center; margin:auto;" 
845
|-
846
| <math>L=\prod_{n=1}^n\frac{1}{\sigma (S_i)\sqrt{2\pi }}\exp \left\{- \frac{{\left(\ln N{-}_{}^i\mu (S_i)\right)}^2}{2{\sigma }^2(S_i)}\right\}</math> 
847
|}
848
| style="width: 5px;text-align: right;white-space: nowrap;" | (18)
849
|}
850
851
852
Take the logarithm on both sides of the above equation, and get it
853
854
{| class="formulaSCP" style="width: 100%; text-align: center;" 
855
|-
856
| 
857
{| style="text-align: center; margin:auto;" 
858
|-
859
| <math>\ln L=-\sum_{i=1}^n\left\{In\sqrt{2\pi }+\ln\sigma (S_i)+ \frac{{\left(\ln N_i-\mu (S_i)\right)}^2}{2{\sigma }^2(S_i)}\right\}</math> 
860
|}
861
| style="width: 5px;text-align: right;white-space: nowrap;" | (19)
862
|}
863
864
Convert Eq. (18) into Eq. (19) and find the minimum value of Eq. (19) to find the maximum value of the likelihood function equation
865
866
{| class="formulaSCP" style="width: 100%; text-align: center;" 
867
|-
868
| 
869
{| style="text-align: center; margin:auto;" 
870
|-
871
| <math>F(m_{50},m_{84.1})=\sum_{i=1}^n\left\{In\sqrt{2\pi }+ \ln\sigma (S_i)+\frac{{\left(\ln N_i-\mu (S_i)\right)}^2}{2{\sigma }^2(S_i)}\right\}</math> 
872
|}
873
| style="width: 5px;text-align: right;white-space: nowrap;" | (20)
874
|}
875
876
Find the minimum value of F to obtain the maximum likelihood estimates of the parameters <math>m50</math>, <math>m84.1</math>, <math>m50=15.04</math>,   <math>m84.1=5.32</math>.
877
878
The values of parameters <math>m50</math> and <math>m84.1</math> are substituted into Eqs. (11) and (16), respectively, to obtain the life distribution parameters under any stress, and the log life mean and standard deviation parameter equations are as follows:
879
880
{| class="formulaSCP" style="width: 100%; text-align: center;" 
881
|-
882
| 
883
{| style="text-align: center; margin:auto;" 
884
|-
885
| <math>\sigma (S)=29.29-9.72 \lg S</math>
886
|}
887
| style="width: 5px;text-align: right;white-space: nowrap;" | (21)
888
|}
889
890
{| class="formulaSCP" style="width: 100%; text-align: center;" 
891
|-
892
| 
893
{| style="text-align: center; margin:auto;" 
894
|-
895
| <math>\mu (S)=48.97-15.04\lg S</math> 
896
|}
897
| style="width: 5px;text-align: right;white-space: nowrap;" | (22)
898
|}
899
900
901
The parameter-stress equation derived by the great likelihood method was plotted against the sample log-life mean and standard deviation in the same graph for comparison. The sample log-life mean and log-life standard deviation are shown in [[#tab-11|Table 11]] and the results of the comparison are shown in [[#img-22|Figures 22]] and [[#img-23|23]]. Within a certain interval, the fit is good, which proves that the parameter equations derived using the great likelihood method are more reasonable.
902
903
<div class="center" style="font-size: 75%;">'''Table 11'''.  The mean and standard deviation of the log-life of the sample</div>
904
905
<div id='tab-1'></div>
906
{| class="wikitable" style="margin: 1em auto 0.1em auto;border-collapse: collapse;font-size:85%;width:auto;" 
907
|-style="text-align:center"
908
! Stress (MPa) !! Mean log -life !! Standard deviation of log-life
909
|-style="text-align:center"
910
|  1030
911
|  3.7572
912
|  0.0951
913
|-style="text-align:center"
914
|  970
915
|  3.8692
916
|  0.1433
917
|-style="text-align:center"
918
|  990
919
|  4.766
920
|  0.7922
921
|-style="text-align:center"
922
|  849
923
|  5.1407
924
|  1.0905
925
|-style="text-align:center"
926
|  788
927
|  5.406
928
|  1.1344
929
|-style="text-align:center"
930
| 758
931
|  6.0801
932
|  1.1266
933
|-style="text-align:center"
934
|  728
935
|  6.7981
936
|  1.0539
937
|-style="text-align:center"
938
|  697
939
|  7.5227
940
|  0.2203
941
|}
942
943
944
<div id='img-22'></div>
945
{| class="wikitable" style="margin: 1em auto 0.1em auto;border-collapse: collapse;width:auto;" 
946
|-style="background:white;"
947
|style="text-align: center;padding:10px;"| [[Image:Draft_Ling_717181314-image44.png|500px]]
948
|-
949
| style="background:#efefef;text-align:left;padding:10px;font-size: 85%;"| '''Figure 22'''. Median S-N curve versus sample log-life mean 
950
|}
951
952
953
<div id='img-23'></div>
954
{| class="wikitable" style="margin: 1em auto 0.1em auto;border-collapse: collapse;width:auto;" 
955
|-style="background:white;"
956
| style="text-align: center;padding:10px;"| [[Image:Draft_Ling_717181314-image45.png|500px]] 
957
|-
958
| style="background:#efefef;text-align:left;padding:10px;font-size: 85%;"| '''Figure 23'''.  Standard deviation curve versus sample log-life standard deviation
959
|}
960
961
==5. Bearing fatigue reliability==
962
963
According to the heterogeneous interference model, for different reliability of fatigue life is calculated, the model as shown in the equation, the above calculated stress distribution Eq. (4) and life distribution Eq. (17) substituted into the following Eq. (23), the fatigue life curve of the bearing can be obtained as shown in [[#img-24|Figure 24]].
964
965
{| class="formulaSCP" style="width: 100%; text-align: center;" 
966
|-
967
| 
968
{| style="text-align: center; margin:auto;" 
969
|-
970
| <math>R=\int_0^{+\infty }h(S)\int_N^{+\infty }f(n,S)dn\,dS</math>
971
|}
972
| style="width: 5px;text-align: right;white-space: nowrap;" | (23)
973
|}
974
975
where <math display="inline">R</math> is the fatigue reliability, <math display="inline">h(S)</math> is the probability density function of the equivalent load distribution and <math display="inline">f(n,S)</math> is the probability density function of the component life under different stresses.
976
977
<div id='img-24'></div>
978
{| class="wikitable" style="margin: 1em auto 0.1em auto;border-collapse: collapse;width:auto;" 
979
|-style="background:white;"
980
|style="text-align: center;padding:10px;"| [[Image:Draft_Ling_717181314-image47.png|500px]]
981
|-
982
| style="background:#efefef;text-align:left;padding:10px;font-size: 85%;"| '''Figure 24'''. Bearing fatigue reliability variation curve
983
|}
984
985
986
As can be seen from [[#img-24|Figure 24]], the fatigue reliability decreases continuously with increasing fatigue life, which is consistent with reality. The downward trend is faster until the number of cycles is 10<sup>5</sup>, after which it slows down. When the number of fatigue cycles reaches 10<sup>5</sup>, the fatigue reliability of the bearing drops to below 0.82.
987
988
==6. Conclusions==
989
990
In this paper, the reliability of the bearings of a giant tyre unloader is predicted using a multi-body dynamics, finite and other numerical simulation platform, combined with a heterogeneous interference model, for a huge alternating load during operation:
991
992
(1) The joint simulation between ADAMS and ANSYS shows that the combined force on the bearing at the lowermost rocker arm of the tyre unloader can reach a maximum of 150kN, with a maximum of 108kN and 78kN in the transverse and longitudinal directions. The bearing of the lowermost rocker arm of the tyre unloader is subjected to the greatest inertial force when bearing the weight of the whole tyre and flipping, and the maximum stress value is 1316.2Mpa.
993
994
(2) By probability fitting and testing, the equivalent stress distribution in the fatigue part of the bearing conforms to the Weibull distribution, with the Weibull distribution scale parameter and shape parameter being 61.93 and 0.72 respectively at a confidence interval of 95%.
995
996
(3) Under the modified S-N curve, the life of the bearing steel material under any stress conforms to a log-
997
998
normal distribution, and the Basquin equation exponents are 15.04 and 5.32 for a survival rate of 50% and 84.1% respectively, as obtained by the maximum likelihood method.
999
1000
(4) Based on the heterogeneous interference fatigue reliability model and according to the reliability calculation results, the reliability of the bearing shows a decreasing trend with the increase of the number of times it is used, which is in line with the actual situation, when the service life of the bearing reaches 10<sup>5</sup> times, the reliability drops to below 0.82.
1001
1002
The kinetic simulation, finite element analysis, fatigue reliability prediction and other technical methods used in this paper are feasible for the fatigue reliability prediction of tyre unloading machine bearings.
1003
1004
==Acknowledgements==
1005
1006
This work was supported by the Natural Science Foundation of Fujian Province (Grant No.2020J01871), and the China Postdoctoral Science Foundation (Granted No.2020M671956).
1007
1008
===Conflict of Interests===
1009
1010
Conflict of interests The authors declare that they have no known competing financial interests or personal relationships.
1011
1012
==References==
1013
<div class="auto" style="text-align: left;width: auto; margin-left: auto; margin-right: auto;font-size: 85%;">
1014
1015
[1] Levratti A., Riggio G., Fantuzzi C., De Vuono A.,  Secchi C. TIREBOT: A collaborative robot for the tyre workshop. Robotics and Computer-Integrated Manufacturing, 57:129-137, 2019. [https://doi.org/10.1016/j.rcim.2018.11.001 https://doi.org/10.1016/j.rcim.2018.11.001]
1016
1017
[2] Vavro J., Vavro J., Kováiková P., Bezdedová R. Kinematic and dynamic analysis of the manipulator for removal of rough tyres. Procedia Engineering, 136:120-124, 2016. [https://doi.org/10.1016/j.proeng.2016.01.184 https://doi.org/10.1016/j.proeng.2016.01.184]
1018
1019
[3] Xie L.Y, Wang Z. Heterogeneous interference model for fatigue reliability with random constant amplitude cyclic loading (in Chinese). Journal of Mechanical Engineering, (01):1-6, 2008.
1020
1021
[4] Liu K., Wang M., Yang G.L., Yu H. Fatigue reliability analysis of underground pipeline structures under multiple working conditions (in Chinese). Safety and Environmental Engineering, 26(04):106-110+126, 2019. [https://doi.org/10.13578/j.cnki.issn.1671-1556.2019.04.017 https://doi.org/10.13578/j.cnki.issn.1671-1556.2019.04.017]
1022
1023
[5] Li R.X., Zhou J.Y., Sun K.Z., et al. Fatigue reliability analysis of rolling bearing systems under random loading (in Chinese). Hydromechatronics Engineering, 40(01):157-160, 2012.
1024
1025
[6] Jin Y., Liu S.J., Zhang J.G. Fatigue reliability of high speed bearing based on genetic algorithm optimized artificial neural network (in Chinese). Journal of Aerospace Power, 2018(11):2748-2755, 2018. [https://doi.org/10.13224/j.cnki.jasp.2018.11.021 https://doi.org/10.13224/j.cnki.jasp.2018.11.021]
1026
1027
[7] Qi M.Y., Liao A.H. Design of reliability evaluation system of traction motor rolling bearing based on MATLAB app designer (in Chinese). Electronics Technology, (03):79-86, 2022. [https://doi.org/10.16180/j.cnki.issn1007-7820.2022.03.012 https://doi.org/10.16180/j.cnki.issn1007-7820.2022.03.012]
1028
1029
[8] Reuben L.C.K., Corsar M., Mba D. Bearing replacement extension without failure data. Int. J. of Risk Assessment and Management, 18(1):38-51, 2015.  [https://doi.org/10.1504/IJRAM.2015.068139 https://doi.org/10.1504/IJRAM.2015.068139]
1030
1031
[9] Pape F.,  Maiss O.,  Denkena B.,  Poll G., et al. Computational approach to improve bearings by residual stresses based on their required bearing fatigue life. International Journal of Computational Methods and Experimental Measurements, 6(4):656-666, 2018. [https://doi.org/10.2495/CMEM-V6-N4-656-666 https://doi.org/10.2495/CMEM-V6-N4-656-666]
1032
1033
[10] Wang H.B. Ship unloader rolling bearing life prediction and maintenance management platform development (in Chinese). Jilin University, 2011.
1034
1035
[11] Xia X.T., Ye L. et al. Performance continuity reliability prediction of rolling bearings (in Chinese). Bearings,(06):28-34, 2016. [https://doi.org/10.19533/j.issn1000-3762.2016.06.008 https://doi.org/10.19533/j.issn1000-3762.2016.06.008]
1036
1037
[12] Zhang X., Zhang Y.Q., Sun Y., Wena B.,  Jiang L. Reliability model of TBM main bearing based on nonlinear strength degradation theory. International Journal of Performability Engineering, 14(12):3054-3065, 2018. [https://doi.org/10.23940/ijpe.18.12.p15.30543065 https://doi.org/10.23940/ijpe.18.12.p15.30543065]
1038
1039
[13] Cheng W.W., Xiang M.H., Lyu B., Gu J., Xiao L. Influence of angular misalignment on the tribological performance of high-speed micro ball bearings considering full multibody interactions. Proceedings of the Institution of Mechanical Engineers, Part J: Journal of Engineering Tribology, 235(6):1168-1189, 2021. [https://doi.org/10.1177/1350650120948292 https://doi.org/10.1177/1350650120948292]
1040
1041
[14] Herp J., Ramezani M.H., Bach-Andersen M., Pedersen N.L., Nadimi E.S. Bayesian state prediction of wind turbine bearing failure. Renewable Energy, 116:164-172, 2018. [https://doi.org/10.1016/j.renene.2017.02.069 https://doi.org/10.1016/j.renene.2017.02.069]
1042
1043
[15] Tong V.C., Hong S.W. Improved formulation for running torque in angular contact ball bearings. International Journal of Precision Engineering and Manufacturing, 19(1):47-56, 2018. [https://doi.org/10.1007/s12541-018-0006-2 https://doi.org/10.1007/s12541-018-0006-2]
1044
1045
[16] König F., Ouald C.A., Jacobs G., Sous C. A multiscale-approach for wear prediction in journal bearing systems – from wearing-in towards steady-state wear. Wear, 426-427(Part B):1203-1211, 2019. [https://doi.org/10.1016/j.wear.2019.01.036 https://doi.org/10.1016/j.wear.2019.01.036]
1046
1047
[17] Morales-Espejel G.E., Gabelli A. A model for rolling bearing life with surface and subsurface survival: Surface thermal effects. Wear, 460–461:203446, 2020. [https://doi.org/10.1016/j.wear.2020.203446 https://doi.org/10.1016/j.wear.2020.203446]
1048
1049
[18] Zhang Y.C., Shu T., Ding R.X. Bearing life prediction based on SPSS and grey prediction model. IOP Conference Series: Earth and Environmental Science, 634(1):012051, 2021. [https://doi.org/10.1088/1755-1315/634/1/012051 https://doi.org/10.1088/1755-1315/634/1/012051]
1050
1051
[19] Lorenz S.J., Sadeghi F., Trivedi H.K., Rosado L., Kirsch M.S., Wang C. A continuum damage mechanics finite element model for investigating effects of surface roughness on rolling contact fatigue. International Journal of Fatigue, 143(3):105986, 2020. [https://doi.org/10.1016/j.ijfatigue.2020.105986 https://doi.org/ https://doi.org/10.1016/j.ijfatigue.2020.105986]
1052
1053
[20] Pandey V.B., Singh I.V., Mishra B.K. A strain-based continuum damage model for low cycle fatigue under different strain ratios. Engineering Fracture Mechanics, 242(4):107479, 2020. [https://doi.org/10.1016/j.engfracmech.2020.107479 https://doi.org/10.1016/j.engfracmech.2020.107479]
1054
1055
[21] Cano J.A.,  Stewart C.M. A continuum damage mechanics (CDM) based Wilshire model for creep deformation, damage, and rupture prediction. Materials Science and Engineering A, 799(4):140231, 2021. [https://doi.org/10.1016/j.msea.2020.140231 https://doi.org/10.1016/j.msea.2020.140231]
1056
1057
[22] Li W., Lu L.T. Probabilistic properties of the ultra-long life S-N relationship for high-carbon chromium bearing steels (in Chinese). Journal of Traffic and Transportation Engineering, (02):17-21, 2006.
1058

Return to Ling et al 2023a.

Back to Top

Document information

Published on 18/04/23
Accepted on 10/04/23
Submitted on 24/02/23

Volume 39, Issue 2, 2023
DOI: 10.23967/j.rimni.2023.04.001
Licence: CC BY-NC-SA license

Document Score

0

Views 33
Recommendations 0

Share this document

claim authorship

Are you one of the authors of this document?