Thermal Fatigue Life Prediction under Temperature Uncertainty for Shot Sleeve of Squeeze Casting Machine
Round 1
Reviewer 1 Report
In this interesting work, the authors have provided a modern approach with surrogate model in estimating the influence of temperature uncertainty on thermal fatigue life prediction of a shot sleeve in squeeze casting processes. The research is quite novel, and it is worth publishing in Metals if the following comments are addressed, point by point:
- A proper coupling and modeling of various physical phenomena in metal solidification porcesses (such as in aluminum and steel casting) is still a very open research area, and is the first step in successful modeling of short term failures or fatigue. The authors should provide a wider review of multi-physics modeling in metal solidification and cite some of those works, such as:
Heat transfer and thermo-mechanical stresses in a gravity casting die - Influence of process parameters. Journal of Materials Processing Technology, 2001, 110 (2), pp.211-217.
Thermomechanical finite-element model of shell behavior in continuous casting of steel. Metall. Mater. Trans., 2004, B 35, 1151–1172
Enhanced latent heat method to incorporate superheat effects into fixed-grid multiphysics simulations. Numer. Heat Tran., Part B: Fundamentals 57, 2010, 396–413.
Multiphysics modeling of continuous casting of stainless steel, Journal of Materials Processing Technology, (2020), 278
- In equation 7, plastic strains are mentioned. However, it is not clear which materially non-linear model was used in FEA analysis (for training data generation). Was it a simple Von-Mises plasticity? Visco-plastic models (that take into consideration time dependent effects) usually provide a better approximation for metal solidification at elevated temperatures.
- Also, are the thermal strains taken into consideration? Perhaps they are in the numerical analysis, i.e. that the total strain increment is the sum of elastic, thermal and plastic strain increments? Please explain.
- Please define the “dangerous” position of the shot sleeve, is it dangerous with respect to the short term failures such as cracks and/or longer term fatigue?
- It is not clear how the surrogate model is validated in Figure 7? Was the generated data split into training and testing data, and the testing data was set aside while the surrogate model was trained with the training data? And then the testing data was tested and shown on the figure 7?
- Deep learning surrogate model are becoming increasingly popular an alternative to the Kriging model are alredau used such as in:
Deep Learning Sequence Methods in Multiphysics Modeling of Steel Solidification. Metals 2021, 11, 494
which may provide a better approximation to numerical results then the Kriging model particularly since more uncertain factors are planned to be used. Please comment on this as well (as a potential future work)
Author Response
Thank you very much for your comments. After careful consideration, our responses are as follows:
Point 1: A proper coupling and modeling of various physical phenomena in metal solidification porcesses (such as in aluminum and steel casting) is still a very open research area, and is the first step in successful modeling of short term failures or fatigue. The authors should provide a wider review of multi-physics modeling in metal solidification and cite some of those works, such as:
Heat transfer and thermo-mechanical stresses in a gravity casting die - Influence of process parameters. Journal of Materials Processing Technology, 2001, 110 (2), pp.211-217.
Thermomechanical finite-element model of shell behavior in continuous casting of steel. Metall. Mater. Trans., 2004, B 35, 1151–1172
Enhanced latent heat method to incorporate superheat effects into fixed-grid multiphysics simulations. Numer. Heat Tran., Part B: Fundamentals 57, 2010, 396–413.
Multiphysics modeling of continuous casting of stainless steel, Journal of Materials Processing Technology, (2020), 278
Response 1: We accept the suggestion and briefly describe the importance of proper coupling and modelling of various physical phenomena on Pg.2 in the revised manuscript.
Point 2: In equation 7, plastic strains are mentioned. However, it is not clear which materially non-linear model was used in FEA analysis (for training data generation). Was it a simple Von-Mises plasticity? Visco-plastic models (that take into consideration time dependent effects) usually provide a better approximation for metal solidification at elevated temperatures.
Response 2: Considering that our main research object is the injection mechanism, we simplified the molten metal as a solid heat source and the elastic-plastic isotropic material model was used. The coupling model considering the flow and solidification process will be further studied in the next work.
Point 3: Also, are the thermal strains taken into consideration? Perhaps they are in the numerical analysis, i.e. that the total strain increment is the sum of elastic, thermal and plastic strain increments? Please explain.
Response 3: We mainly considered the thermal stress and thermal strain caused by the temperature change of the shot sleeve because of the heavy heat transfer between the shot sleeve and molten metal. Since the temperature of the shot sleeve was relatively low, the thermal stress was below the yield strength of the material and just caused an elastic deformation. The mechanical stress was small in this study, which was negligible.
Point 4: Please define the “dangerous” position of the shot sleeve, is it dangerous with respect to the short term failures such as cracks and/or longer term fatigue?
Response 4: The dangerous position mentioned here refers to the position where thermal fatigue was most likely to occur on the shot sleeve, corresponding to the point where the stress and strain were maximum.
We have added the relevant description at lines 239-240 on page 8 in the revised manuscript.
Point 5: It is not clear how the surrogate model is validated in Figure 7? Was the generated data split into training and testing data, and the testing data was set aside while the surrogate model was trained with the training data? And then the testing data was tested and shown on the figure 7?
Response 5: Yes. Thirty sampling points obtained through the FEA were divided into two sets, ten of which were used as a training set and 20 as a test set. Figure 8 shows the results for the test set.
We have added the relevant description at lines 264-266 on page 9 in the revised manuscript.
Point 6: Deep learning surrogate model are becoming increasingly popular an alternative to the Kriging model are already used such as in:
Deep Learning Sequence Methods in Multiphysics Modeling of Steel Solidification. Metals 2021, 11, 494.
which may provide a better approximation to numerical results then the Kriging model particularly since more uncertain factors are planned to be used. Please comment on this as well (as a potential future work)
Response 6: We accept the suggestion and added the relevant description in Section 5 of the revised manuscript.
Reviewer 2 Report
- It would be good to give the dimensions of the sample for testing.
- Fig. 4 – illegible description, enlarge.
- It would be good to give the material properties of the materials tested.
- The topic of the work mentions fatigue life, but I do not see the description of the stand for this research. Load range, frequency ?
- Such numerical simulations without experimental confirmation are of little importance. Each result can be generated numerically.
- The first conclusion does nothing to work.
Author Response
Thank you very much for your comments. After careful consideration, our responses are as follows:
Point 1: It would be good to give the dimensions of the sample for testing.
Response 1: The punch is 63 mm in diameter and 110 mm high. The outer diameter, inner diameter, and height of the shot sleeve are 95, 63, and 360 mm, respectively.
We accept the suggestion and have added the relevant description at lines 189-191 on page 6 in the revised manuscript.
Point 2: Fig. 4 – illegible description, enlarge.
Response 2: We accept the suggestion and have enlarged Figure 4 in the revised manuscript.
Point 3: It would be good to give the material properties of the materials tested.
Response 3: We accept the suggestion and have given the material properties of the materials tested in Table 1 on page 7 in the revised manuscript.
Point 4: The topic of the work mentions fatigue life, but I do not see the description of the stand for this research. Load range, frequency ?
Response 4: The fatigue life of the shot sleeve studied in this paper was mainly caused by thermal stress. In this process, the external mechanical load had little effect, so only the thermal fatigue caused by thermal stress was considered. Load range was shown in Figure 6(b), and frequency was one cycle time 145s.
Point 5: Such numerical simulations without experimental confirmation are of little importance. Each result can be generated numerically.
Response 5: Figure 5 shows the comparison between the experiment and simulation results of temperature and deformation. The experimental results reflect the reliability of the simulation.
Point 6: The first conclusion does nothing to work.
Response 6: We have modified the first conclusion in the revised manuscript.
Reviewer 3 Report
General comments
- The modified universal slope equation is based mostly on materials mechanically tested at constant (room) temperature. Authors use it to predict fatigue in conditions of thermo/mechanical loading (mechanical loading is induced by cycling temperature). There is probably no macroscopic plastic strain, and the temperature range is relatively low (150-270•C according to Figure 6), so the model seems to be acceptable, authors should briefly discuss this aspect of using universal slope equation.
- As the temperature of pouring was 720•C and the peak temperature of wall 270•C (Figure 6) there is a question if mesh density on inner surface of shot sleeve (Figure 4) is sufficient. Did authors try perform any sensitivity study of mesh density?
Suggestions
- Qualification or selection instead of Quantification of uncertain parameters should be used on pg. 3, row 99 and in Figure 1.
- The estimated values of the points to be measured can be calculated since or if the values of weight coefficients w w 1 ,w 2 , ,w n were obtained should be used instead at pg. 3, rows 112 -114.
- Figure 3: Meaning of Tn and Dn should be described in the caption (for instance, thermocouples T1 - T5, strain gauges, micrometer rangefinders).
- The heading 3.2 should be rather FE Simulation of Injection mechanism
- Pg.7, row 198-200: The choice of heat transfer coefficients based on resouces [30,31] is OK, I miss any mention about their temperature dependence inducing additional uncertainities. Authors should comment briefly if the uncertainities are included / are not included it into the analysis.
- Pg. 7, row 222-224: As the mechanical properties of H13 depend strictly on heat treatment, authors should mention heat treatment of shot sleeve and die and provide reader with approximate yield strength at operational temperature of sleeve inner surface, to compare it with the value determined from FE analysis.
- Pg, 8: Authors should explain how strain range in table 8 had been calculated. Is it based on a component of strain tensor or on an invariant of it?
- Figure. 5: Please explain what gauges are represented by the curves or show the positions in Figure 4.
- Figure. 6: Please show the positionn of evaluated point in Figure 3 or in Figure 4.
Author Response
Thank you very much for your comments. After careful consideration, our responses are as follows:
Point 1: The modified universal slope equation is based mostly on materials mechanically tested at constant (room) temperature. Authors use it to predict fatigue in conditions of thermo/mechanical loading (mechanical loading is induced by cycling temperature). There is probably no macroscopic plastic strain, and the temperature range is relatively low (150-270•C according to Figure 6), so the model seems to be acceptable, authors should briefly discuss this aspect of using universal slope equation.
Response 1: We accept the suggestion and have added the 3rd paragraph on page 5 in the revised manuscript for description.
Point 2: As the temperature of pouring was 720•C and the peak temperature of wall 270•C (Figure 6) there is a question if mesh density on inner surface of shot sleeve (Figure 4) is sufficient. Did authors try perform any sensitivity study of mesh density?
Response 2: We have done some tests and the mesh density is reasonable. From Figure 5, the experimental and simulation results are in good agreement, which proves the accuracy and reliability of the finite element model.
Point 3: Qualification or selection instead of Quantification of uncertain parameters should be used on pg. 3, row 99 and in Figure 1.
Response 3: We accept the suggestion and have modified the item in Figure 1 on Pg.3 in the revised manuscript.
Point 4: The estimated values of the points to be measured can be calculated since or if the values of weight coefficients w w 1 ,w 2 , ,w n were obtained should be used instead at pg. 3, rows 112 -114.
Response 4: We accept the suggestion and have modified the sentence on Pg.3 row 115 in the revised manuscript.
Point 5: Figure 3: Meaning of Tn and Dn should be described in the caption (for instance, thermocouples T1 - T5, strain gauges, micrometer rangefinders).
Response 5: Measurement points were selected on the shot sleeve (three temperature points T1, T2, and T3, and three deformation points D1, D2, and D3), and the punch (two temperature points T4 and T5, and one deformation point D4).
We have added the relevant description on page 6, rows 197-200 in the revised manuscript.
Point 6: The heading 3.2 should be rather FE Simulation of Injection mechanism
Response 6: We have corrected this mistake in the revised manuscript.
Point 7: Pg.7, row 198-200: The choice of heat transfer coefficients based on resources [30,31] is OK, I miss any mention about their temperature dependence inducing additional uncertainties. Authors should comment briefly if the uncertainties are included / are not included it into the analysis.
Response 7: Here we take the constant values of heat transfer coefficients without considering their temperature dependence inducing additional uncertainties.
We have added the description on page 7, rows 216-218 in the revised manuscript.
Point 8: Pg. 7, row 222-224: As the mechanical properties of H13 depend strictly on heat treatment, authors should mention heat treatment of shot sleeve and die and provide reader with approximate yield strength at operational temperature of sleeve inner surface, to compare it with the value determined from FE analysis.
Response 8: After quenching and tempering, the yield strength of H13 should not be less than 800MPa considering the temperature of the shot sleeve inner surface [8].
We have added the description on page 8, rows 246-248 in the revised manuscript.
Point 9: Pg, 8: Authors should explain how strain range in table 8 had been calculated. Is it based on a component of strain tensor or on an invariant of it?
Response 9: Maximum principal value of total strain was adopted for strain, and the strain range was obtained by calculating the difference between the maximum value and the minimum value.
We have added the description on page 9, rows 262-264 in the revised manuscript.
Point 10: Figure. 5: Please explain what gauges are represented by the curves or show the positions in Figure 4.
Response 10: Experimental temperature points Exp1- Exp3 in Figure 5(a) respectively correspond to T1-T3 in Figure 3(a), and experimental deformation points Exp1- Exp3 in Figure 5(b) respectively correspond to D1-D3 in Figure 3(a). The positions of the corresponding simulation points and experimental points are the same.
We have added the description on page 7, rows 222-225 in the revised manuscript.
Point 11: Figure. 6: Please show the position of evaluated point in Figure 3 or in Figure 4.
Response 11: We have added Figure 7 to show the position of evaluated point on page 9 in the revised manuscript.
Round 2
Reviewer 2 Report
The authors took into account all comments of the Reviewer.
