I studied a complex thin-walled stainless steel bearing seat produced by the lost wax casting process. The part is used in food equipment and must satisfy strict requirements for surface quality, dimensional accuracy, coaxiality, flatness, and internal soundness. The material is 1.4308 stainless steel, and the as-cast mass is approximately 2.2 kg. The main body combines a cylindrical thin-wall tube with a rectangular thin-wall frame, reinforcing ribs, an irregular protruding plate, and several hollowed regions. The minimum wall thickness is about 1.8 mm, while the general wall thickness is in the range of 3 to 6 mm. Because of this geometry, the lost wax casting process is prone to mistun, cold shut, surface scarring, shrinkage cavity, shrinkage porosity, and deformation. My objective was to combine numerical simulation with intelligent optimization algorithms to improve the lost wax casting process, increase the process yield, and reduce the probability of casting defects.
Lost wax casting is a precision casting method that uses an expendable wax pattern, a ceramic shell, and a gravity or pressure-assisted pouring operation. In my work, lost wax casting included wax injection, pattern assembly, shell preparation, dewaxing, shell firing, alloy melting, pouring, shell removal, cut-off, and post-processing. The quality of a lost wax casting depends strongly on the gating system, pouring position, shell thickness, shell preheating, pouring temperature, and pouring speed. I therefore treated the gating system design and the process parameter optimization as two connected stages of the same lost wax casting problem.
The chemical composition of the 1.4308 stainless steel used in my lost wax casting experiments is summarized in Table 1. The alloy has good corrosion resistance, high-temperature performance, and mechanical workability. Its solidus temperature is approximately 1386 °C, and its liquidus temperature is approximately 1485 °C. These thermal properties are important because they control the filling ability, solidification sequence, and shrinkage behavior of the lost wax casting.
| Element | C | Si | Mn | P | S | Cr | Ni | Fe |
|---|---|---|---|---|---|---|---|---|
| Composition, wt.% | ≤0.07 | ≤1.50 | ≤1.50 | ≤0.040 | ≤0.030 | 18.0–20.0 | 8.0–11.0 | Balance |
I used a medium-temperature wax for the expendable patterns. The wax had good fluidity and low shrinkage, which helped maintain the dimensional precision of the lost wax casting pattern. The shell system consisted of a zircon sand face coat and mullite backup layers bonded with silica sol. Zircon sand provided high refractoriness and low thermal expansion, while mullite provided good high-temperature strength and creep resistance. The main material properties are listed in Table 2, and the shell-making parameters are listed in Table 3.
| Material | Chemical nature | Melting point, °C | Refractoriness, °C | Density, g/cm³ | Thermal expansion, 1/°C |
|---|---|---|---|---|---|
| Zircon sand | Acidic | 2750 | >1825 | 4.5–4.9 | 4.6 × 10⁻⁶ |
| Mullite sand | Neutral | 1810 | >1700 | 3.16 | 4.5 × 10⁻⁶ |
| Parameter | Face coat | Intermediate layers | Backup layers |
|---|---|---|---|
| Stucco material | 100–120 mesh zircon sand | 30–60 mesh mullite sand | 16–30 mesh mullite sand |
| Temperature | 23 ± 2 °C | 23 ± 2 °C | 23 ± 2 °C |
| Humidity | 60–80% | 40–60% | 40–60% |
| Air speed | 6–8 m/s | 6–8 m/s | 6–8 m/s |
| Drying time | 5–24 h | 8–24 h | >12 h |
For the lost wax casting experiments, I injected the wax at approximately 65 °C with a dwell time of about 35 s. After cooling, I removed flash and repaired surface defects. I assembled the patterns onto a central runner and then cleaned the pattern tree before shell making. Each shell layer involved dipping, stuccoing, and drying. After the final seal coat, I dewaxed the shell in an autoclave and dried it in air. Before pouring, I fired the shell in a roasting furnace to remove residual moisture and wax and to preheat the mold. The alloy was melted in an induction furnace, and the chemical composition was checked with a spectrometer before pouring.
Numerical simulation of lost wax casting requires coupled descriptions of fluid flow, heat transfer, solidification, and stress evolution. I treated the molten alloy as an incompressible viscous fluid and neglected turbulence in the first approximation. The continuity equation used in my simulation is
$$
\frac{\partial u}{\partial x}+\frac{\partial v}{\partial y}+\frac{\partial w}{\partial z}=0
$$
where \(u\), \(v\), and \(w\) are the velocity components in the \(x\), \(y\), and \(z\) directions. The momentum conservation equation in the \(x\) direction is
$$
\rho\left(\frac{\partial u}{\partial t}+u\frac{\partial u}{\partial x}+v\frac{\partial u}{\partial y}+w\frac{\partial u}{\partial z}\right)
=
-\frac{\partial p}{\partial x}
+\mu\left(\frac{\partial^2 u}{\partial x^2}+\frac{\partial^2 u}{\partial y^2}+\frac{\partial^2 u}{\partial z^2}\right)
+\rho g_x
$$
where \(\rho\) is density, \(p\) is pressure, \(\mu\) is dynamic viscosity, and \(g_x\) is the gravitational acceleration component. Similar equations were solved in the \(y\) and \(z\) directions. The energy equation used for the filling and solidification stages is
$$
\rho c_p\left(\frac{\partial T}{\partial t}+u\frac{\partial T}{\partial x}+v\frac{\partial T}{\partial y}+w\frac{\partial T}{\partial z}\right)
=
\lambda\left(\frac{\partial^2 T}{\partial x^2}+\frac{\partial^2 T}{\partial y^2}+\frac{\partial^2 T}{\partial z^2}\right)
+\rho L\frac{\partial f_s}{\partial t}
$$
where \(c_p\) is specific heat, \(T\) is temperature, \(\lambda\) is thermal conductivity, \(L\) is latent heat, and \(f_s\) is solid fraction. Heat transfer between the casting and the ceramic shell included conduction, convection, and radiation. Fourier’s law was used for conduction:
$$
q=-k\frac{\partial T}{\partial x}
$$
Convection between the shell and the surroundings was represented by
$$
q=\alpha(T_f-T_w)
$$
and radiation was represented by
$$
q=\varepsilon\sigma_0 T_s^4
$$
where \(\alpha\) is the convective heat-transfer coefficient, \(T_f\) is the fluid temperature, \(T_w\) is the wall temperature, \(\varepsilon\) is emissivity, \(\sigma_0\) is the Stefan-Boltzmann constant, and \(T_s\) is the absolute surface temperature.
Shrinkage cavity and shrinkage porosity are controlled by liquid and solidification contraction. I used the following simplified expression for the shrinkage volume formed in a region that cannot receive liquid feeding:
$$
V=\frac{1}{2}V_L\left[\alpha_L(t_L-t_S)+\alpha_{SL}(t_S-t_F)\right]-\alpha_{SV}V_S(t_S-t_F)
$$
where \(V_L\) is the liquid volume at the solidification temperature, \(t_L\) is the average liquid temperature, \(t_S\) is the solidification temperature, \(t_F\) is the final temperature, \(\alpha_L\) is the liquid volumetric contraction coefficient, \(\alpha_{SL}\) is the solidification contraction coefficient, and \(\alpha_{SV}\) is the solid volumetric contraction coefficient. The total shrinkage volume satisfies
$$
V_{sz}=V_{sk}+V_{ss}
$$
where \(V_{sz}\) is the total shrinkage volume, \(V_{sk}\) is the cavity volume, and \(V_{ss}\) is the porosity volume. For defect prediction, I also used the Niyama criterion:
$$
\frac{G}{\sqrt{R}}<k $$=""
where \(G\) is the local temperature gradient, \(R\) is the cooling rate, and \(K\) is a critical value. A low Niyama value indicates a high probability of shrinkage porosity in the lost wax casting. The critical solid fraction method was also used, with a critical solid fraction of approximately 0.67 for the stainless steel.
I analyzed the bearing seat structure and identified the important features for lost wax casting. The cylindrical tube must maintain coaxiality, the rectangular frame must maintain flatness, and the protruding thin plate must remain dimensionally stable. The thin ribs and hollowed regions are especially sensitive to premature solidification and mistun. Based on these requirements, I evaluated four basic pouring positions and five gating system schemes. The pouring positions included a vertical tube, a horizontal tube with the thin plate on the upper side, a horizontal tube with the thin plate on the lower side, and an inclined position. The five gating schemes are summarized in Table 4. I designed each scheme to balance filling stability, feeding efficiency, pattern-tree strength, and process yield.
| Scheme | Pouring position | Gating type | Tree type | Pattern-tree mass, kg | Process yield, % |
|---|---|---|---|---|---|
| I | Horizontal tube, thin plate lower | Combined top and side | One pattern per tree | 9.9 | 44.4 |
| II | Vertical tube, thin plate side | Step gating | One pattern per tree | 11.6 | 37.9 |
| III | Inclined tube and thin plate | Inclined pouring | One pattern per tree | 10.2 | 43.1 |
| IV | Horizontal tube, thin plate upper | Combined side gating | One pattern per tree | 10.4 | 42.3 |
| V | Vertical tube, thin plate side | Combined step gating | Four patterns per tree | 18.0 | 48.9 |
For each lost wax casting scheme, I established a three-dimensional model and imported it into a finite element casting simulation environment. I meshed the gating system with a minimum element size of 4 mm and the bearing seat with a minimum element size of 2 mm. I generated a shell with a nominal thickness of 6 mm in the initial model. The interface heat-transfer coefficient between the metal and the shell was set to 500 W/(m²·K). The initial pouring temperature was 1620 °C, the initial shell preheating temperature was 1130 °C, and the initial pouring speed was 2 kg/s. The gravity direction was aligned with the pouring cup axis. The symmetry plane was used to reduce computation time.
I simulated the filling process for all five lost wax casting schemes. In Scheme I, the metal entered the part through two side gates, then filled the lower thin plate and the tube. The filling sequence was relatively stable, but some air entrapment appeared in the bottom runner. In Scheme II, the metal was divided by two vertical runners, which reduced the impact on the shell wall, but the upper frame region filled late. In Scheme III, the inclined position produced a nonuniform liquid surface, and local splashing occurred at the beginning of filling. In Scheme IV, the metal entered from the bottom and then filled the tube; splashing was observed in the middle runner. In Scheme V, the metal first filled the rectangular frame and then progressed into the cylindrical tube through the upper thin-wall connection. The step gating arrangement allowed the liquid surface to rise in a more controlled manner. Among the five schemes, Scheme V showed the best combination of stable filling, later solidification of the runner, and higher process yield.
I then simulated the solidification process. In all schemes, the first solid fraction appeared at the thin outer walls because these regions lost heat fastest. The isolated liquid regions formed mainly at the irregular junctions between the cylindrical tube and the rectangular frame, at rib intersections, and at positions far from the gates. The predicted shrinkage cavity and porosity volumes are compared in Table 5. Scheme I produced the smallest predicted shrinkage volume, but its process yield was lower and its surface defects in trial production were not the lowest. Scheme IV produced the fewest predicted shrinkage locations, but its shrinkage volume was the largest. Scheme V produced a shrinkage volume of 0.76 cc and gave the highest process yield. When I considered both simulation and actual trial production, Scheme V was the best choice for the lost wax casting of this thin-walled bearing seat.
| Scheme | Predicted shrinkage locations | Predicted shrinkage volume, cc | Filling stability | Feeding behavior | Process yield, % |
|---|---|---|---|---|---|
| I | 14 | 0.71 | Good | Acceptable | 44.4 |
| II | 16 | 0.76 | Good | Moderate | 37.9 |
| III | 15 | 0.76 | Moderate | Moderate | 43.1 |
| IV | 12 | 0.84 | Moderate | Weak | 42.3 |
| V | 14 | 0.76 | Very good | Good | 48.9 |
I verified the simulation results with actual lost wax casting trials. The trial castings were produced with the same shell materials, melting practice, and pouring equipment. After pouring, I removed the shell, cut off the gating system, and cleaned the castings by shot blasting and acid cleaning. I inspected the external surfaces visually, examined internal defects by X-ray radiography, and sectioned selected castings by wire electrical discharge machining. The actual defects included shrinkage cavity, shrinkage porosity, and surface scarring. The number and location of internal defects generally agreed with the simulation results, which confirmed that the numerical model was useful for lost wax casting process evaluation. However, the X-ray inspection was less effective in regions where the cylindrical tube and frame overlapped, so sectioning was necessary for final verification. Scheme V showed the fewest surface defects among the four-pattern tree schemes and the best overall internal quality. I therefore selected Scheme V as the gating system for further process parameter optimization.
After fixing the gating system, I optimized the lost wax casting process parameters. I selected pouring temperature, pouring speed, shell preheating temperature, and shell thickness as the four factors. I used a four-factor, five-level orthogonal experiment to study their effects on shrinkage porosity and equivalent stress. The factor levels are listed in Table 6. The orthogonal array contained 25 simulation runs. The two quality indices were the shrinkage cavity and porosity volume and the maximum equivalent stress in the casting.
| Factor | Level 1 | Level 2 | Level 3 | Level 4 | Level 5 |
|---|---|---|---|---|---|
| Pouring temperature, °C | 1600 | 1610 | 1620 | 1630 | 1640 |
| Pouring speed, kg/s | 1.5 | 1.75 | 2.0 | 2.25 | 2.5 |
| Shell preheating temperature, °C | 1110 | 1120 | 1130 | 1140 | 1150 |
| Shell thickness, mm | 5.0 | 5.5 | 6.0 | 6.5 | 7.0 |
The orthogonal experiment results are listed in Table 7. Each row corresponds to one lost wax casting simulation. I used the shrinkage volume and equivalent stress as the responses. The results showed that the shrinkage volume ranged from 0.50 cc to 1.11 cc, while the equivalent stress ranged from 372.29 MPa to 570.80 MPa. These variations indicate that the selected process parameters have a strong effect on the lost wax casting quality.
| Run | Pouring temperature, °C | Pouring speed, kg/s | Shell preheating temperature, °C | Shell thickness, mm | Shrinkage volume, cc | Equivalent stress, MPa |
|---|---|---|---|---|---|---|
| 1 | 1600 | 1.50 | 1110 | 5.0 | 0.81 | 570.80 |
| 2 | 1600 | 1.75 | 1130 | 6.5 | 1.11 | 406.03 |
| 3 | 1600 | 2.00 | 1150 | 5.5 | 0.68 | 485.33 |
| 4 | 1600 | 2.25 | 1120 | 7.0 | 1.01 | 387.78 |
| 5 | 1600 | 2.50 | 1140 | 6.0 | 1.06 | 452.99 |
| 6 | 1610 | 1.50 | 1150 | 6.5 | 0.86 | 408.64 |
| 7 | 1610 | 1.75 | 1120 | 5.5 | 0.83 | 513.65 |
| 8 | 1610 | 2.00 | 1140 | 7.0 | 0.95 | 372.29 |
| 9 | 1610 | 2.25 | 1110 | 6.0 | 1.01 | 473.41 |
| 10 | 1610 | 2.50 | 1130 | 5.0 | 0.68 | 559.18 |
| 11 | 1620 | 1.50 | 1140 | 5.5 | 0.66 | 491.83 |
| 12 | 1620 | 1.75 | 1110 | 7.0 | 1.06 | 400.16 |
| 13 | 1620 | 2.00 | 1130 | 6.0 | 0.76 | 459.94 |
| 14 | 1620 | 2.25 | 1150 | 5.0 | 0.53 | 569.59 |
| 15 | 1620 | 2.50 | 1120 | 6.5 | 0.84 | 415.10 |
| 16 | 1630 | 1.50 | 1130 | 7.0 | 0.82 | 399.75 |
| 17 | 1630 | 1.75 | 1150 | 6.0 | 0.66 | 478.66 |
| 18 | 1630 | 2.00 | 1120 | 5.0 | 0.63 | 557.82 |
| 19 | 1630 | 2.25 | 1140 | 6.5 | 0.65 | 420.54 |
| 20 | 1630 | 2.50 | 1110 | 5.5 | 0.87 | 513.04 |
| 21 | 1640 | 1.50 | 1120 | 6.0 | 0.71 | 488.42 |
| 22 | 1640 | 1.75 | 1140 | 5.0 | 0.50 | 564.70 |
| 23 | 1640 | 2.00 | 1110 | 6.5 | 0.82 | 426.47 |
| 24 | 1640 | 2.25 | 1130 | 5.5 | 0.61 | 507.49 |
| 25 | 1640 | 2.50 | 1150 | 7.0 | 0.68 | 390.18 |
I performed a range analysis on the orthogonal experiment results. The results are shown in Table 8. For shrinkage volume, the influence order was shell thickness > pouring temperature > shell preheating temperature > pouring speed. For equivalent stress, the influence order was shell thickness > shell preheating temperature > pouring temperature > pouring speed. Shell thickness had the strongest effect on both responses. This means that the thermal resistance and heat capacity of the ceramic shell play a dominant role in the lost wax casting of this thin-walled part. A thick shell slows heat extraction and can increase the size of isolated liquid regions. A thin shell improves cooling uniformity and reduces the tendency for shrinkage defects, but it must still provide enough strength during pouring.
| Response | Index | Pouring temperature | Pouring speed | Shell preheating temperature | Shell thickness |
|---|---|---|---|---|---|
| Shrinkage volume | K1 | 4.66 | 3.85 | 4.57 | 3.14 |
| K2 | 4.33 | 4.16 | 4.01 | 3.65 | |
| K3 | 3.85 | 3.83 | 3.97 | 4.20 | |
| K4 | 3.63 | 3.81 | 3.81 | 4.27 | |
| K5 | 3.31 | 4.12 | 3.41 | 4.52 | |
| R | 1.35 | 0.35 | 1.16 | 1.38 | |
| Equivalent stress | k1 | 2302.93 | 2359.44 | 2383.87 | 2822.09 |
| k2 | 2327.18 | 2363.20 | 2362.77 | 2511.34 | |
| k3 | 2336.63 | 2301.85 | 2332.39 | 2353.42 | |
| k4 | 2369.80 | 2358.80 | 2302.35 | 2076.78 | |
| k5 | 2377.25 | 2330.49 | 2332.41 | 1950.16 | |
| R | 74.32 | 61.35 | 81.52 | 871.93 |
To build a nonlinear prediction model for lost wax casting, I used a back-propagation neural network. The four process parameters were the inputs, and the shrinkage volume and equivalent stress were the outputs. The network had four input nodes, nine hidden nodes, and two output nodes. The hidden-layer transfer function was a sigmoid function, and the output-layer transfer function was linear. The network was trained with the Levenberg-Marquardt algorithm. The maximum number of training epochs was 1000, the learning rate was 0.1, and the training target was 0.0001. The number of hidden nodes was chosen by the empirical relation
$$
M=\sqrt{m+n}+a
$$
where \(m\) is the number of input nodes, \(n\) is the number of output nodes, and \(a\) is an integer between 0 and 10. I used 20 orthogonal simulation results as training samples and 5 results as test samples.
A conventional back-propagation neural network is sensitive to initial weights and thresholds, so I optimized it with a particle swarm optimization algorithm. The particle swarm algorithm updated each particle velocity and position according to
$$
v_{ij}^{t+1}=w v_{ij}^{t}+c_1 r_1(pbest_{ij}-x_{ij}^{t})+c_2 r_2(gbest_j-x_{ij}^{t})
$$
$$
x_{ij}^{t+1}=x_{ij}^{t}+v_{ij}^{t+1}
$$
where \(w\) is the inertia weight, \(c_1\) and \(c_2\) are learning factors, \(r_1\) and \(r_2\) are random numbers between 0 and 1, \(pbest\) is the individual best position, and \(gbest\) is the global best position. I compared particle swarm optimization with a genetic algorithm for optimizing the neural network. The particle swarm method reached a stable fitness value after about 40 generations, while the genetic algorithm required about 55 generations. The particle swarm optimized network also had a lower final fitness value and better prediction accuracy. I therefore used the particle swarm optimized back-propagation network as the nonlinear mapping model between the lost wax casting parameters and the quality indices.
I then used a genetic algorithm to search for the best process parameter combination. The optimization objective was a weighted sum of the normalized shrinkage volume and normalized equivalent stress:
$$
\min f=w_1 f_1+w_2 f_2
$$
where \(f_1\) is the shrinkage response, \(f_2\) is the equivalent stress response, and \(w_1\) and \(w_2\) are weight coefficients. The parameter bounds were
$$
1600\le x_1\le 1640
$$
$$
1.5\le x_2\le 2.5
$$
$$
1110\le x_3\le 1150
$$
$$
5\le x_4\le 7
$$
where \(x_1\) is pouring temperature in °C, \(x_2\) is pouring speed in kg/s, \(x_3\) is shell preheating temperature in °C, and \(x_4\) is shell thickness in mm. The genetic algorithm used real-number coding, a population size of 50, a crossover probability of 0.8, a mutation probability of 0.2, and 100 generations. The optimization results for different weight combinations are listed in Table 9. As the weight on shrinkage volume increased, the optimized pouring temperature moved toward a lower value within the selected range, the pouring speed remained near the middle of the range, the shell preheating temperature increased, and the shell thickness decreased. Because shrinkage porosity was the main cause of rejection in the actual production of this lost wax casting, I selected the combination that reduced shrinkage volume most effectively while still improving stress.
| Weight for shrinkage | Weight for stress | Pouring temperature, °C | Pouring speed, kg/s | Shell preheating temperature, °C | Shell thickness, mm | Shrinkage volume, cc | Shrinkage improvement, % | Equivalent stress, MPa | Stress improvement, % |
|---|---|---|---|---|---|---|---|---|---|
| 0.5 | 0.5 | 1639.65 | 2.41 | 1122.00 | 6.97 | 0.68 | 10.53 | 393.11 | 14.53 |
| 0.6 | 0.4 | 1603.05 | 2.15 | 1148.89 | 5.01 | 0.54 | 28.95 | 436.62 | 5.07 |
| 0.7 | 0.3 | 1601.00 | 2.15 | 1149.96 | 5.01 | 0.49 | 35.53 | 439.84 | 4.37 |
| 0.8 | 0.2 | 1619.26 | 1.53 | 1148.70 | 5.05 | 0.42 | 44.74 | 448.86 | 2.41 |
| 0.9 | 0.1 | 1618.47 | 2.29 | 1136.91 | 5.03 | 0.58 | 23.68 | 434.85 | 5.46 |
The best compromise for my lost wax casting problem was obtained with a shrinkage weight of 0.8 and a stress weight of 0.2. The corresponding process parameters were a pouring temperature of about 1619 °C, a pouring speed of about 1.53 kg/s, a shell preheating temperature of about 1149 °C, and a shell thickness of about 5.05 mm. I validated this optimized combination by numerical simulation. The predicted shrinkage volume decreased to 0.42 cc, which was a 44.74% reduction relative to the original process. The equivalent stress decreased to 448.86 MPa, which was a 2.41% reduction. The filling process became more stable because the lower pouring speed reduced splashing and gas entrapment. The higher shell preheating temperature improved the thermal uniformity of the shell and delayed premature solidification. The thinner shell improved heat extraction and reduced the size of isolated liquid regions. These effects explain the improvement in lost wax casting quality.
I also performed actual production trials using the optimized process window. I controlled the pouring temperature at approximately 1619 °C, the shell preheating temperature at approximately 1148 °C, the shell thickness at approximately 5 mm, and the pouring time at approximately 12 s. After cooling, shell removal, cut-off, shot blasting, and acid cleaning, the bearing seats showed no visible surface shrinkage, no cold shut, and no obvious deformation. The flatness of the side bosses was acceptable, and the protruding thin plate passed the relevant checking fixture. The coaxiality and flatness requirements of the cylindrical and frame features were satisfied. X-ray inspection showed no significant internal shrinkage cavity. Because the overlap of the cylindrical tube and frame can hide defects in radiographic inspection, I sectioned a trial casting by wire electrical discharge machining. The sectioned surface showed that the shrinkage cavity and porosity that were present in the original process had essentially disappeared. These results confirmed that the combination of numerical simulation and intelligent optimization is effective for the lost wax casting of complex thin-walled stainless steel bearing seats.
The final comparison between the original process and the optimized lost wax casting process is given in Table 10. The optimized process increased the process yield and reduced the defect rate. The shrinkage volume was reduced from 0.76 cc to 0.42 cc. The equivalent stress was reduced from about 460 MPa to 448.86 MPa. Surface scarring, cold shut, and visible shrinkage defects were eliminated in the trial production. The dimensional and geometric tolerances were maintained, which is critical for the bearing seat because it must assemble with other food equipment components. The improved lost wax casting process also reduced the need for welding repair and manual finishing, which lowered the production cost and shortened the manufacturing cycle.
| Index | Original process | Optimized process | Improvement |
|---|---|---|---|
| Pouring temperature, °C | 1620 | 1619 | Nearly unchanged |
| Pouring speed, kg/s | 2.00 | 1.53 | Lower and more stable |
| Shell preheating temperature, °C | 1130 | 1149 | Higher thermal uniformity |
| Shell thickness, mm | 6.00 | 5.05 | Better heat extraction |
| Shrinkage volume, cc | 0.76 | 0.42 | 44.74% reduction |
| Equivalent stress, MPa | 460.00 | 448.86 | 2.41% reduction |
| Surface scarring | Present | Absent | Eliminated |
| Cold shut | Present | Absent | Eliminated |
| Internal shrinkage cavity | Present | Absent | Eliminated |
| Process yield, % | 48.9 | 48.9 | Maintained at higher level |
From my study, several conclusions can be drawn for the lost wax casting of complex thin-walled stainless steel parts. First, the gating system has a decisive influence on filling stability and feeding. For the bearing seat, a combined step gating system with four patterns per tree gave the best balance of process yield and defect control. Second, the pouring position must be chosen so that the thin walls fill completely before the liquid metal loses fluidity. A vertical tube position with the thin plate on the side was better than an inclined or horizontal position for this geometry. Third, the shell thickness is the most important process parameter among those studied. A shell that is too thick slows cooling and increases isolated liquid regions, while a shell that is too thin may not have sufficient strength. Fourth, the pouring speed should be high enough to avoid mistun but low enough to avoid splashing and gas entrapment. Fifth, a higher shell preheating temperature improves filling and reduces thermal gradients, but it must be balanced against grain growth and solidification time. Sixth, the back-propagation neural network optimized by particle swarm optimization can accurately map the nonlinear relationship between lost wax casting parameters and casting defects. When combined with a genetic algorithm, it provides an effective way to search for an improved process window.
For practical production, I recommend using the optimized window for the lost wax casting of this bearing seat: pouring temperature around 1619 °C, pouring speed around 1.5 kg/s, shell preheating temperature around 1148 °C, and shell thickness around 5 mm. The gating system should remain the combined step gating design with one central sprue and multiple ingates, and the pattern tree should use the four-pattern arrangement to maintain the higher process yield. The shell should be made with a zircon face coat and mullite backup layers, and the shell should be fired sufficiently before pouring. The pouring operation should be smooth and continuous, with the pouring cup kept full to provide feeding pressure. After pouring, the casting should be allowed to cool naturally, and the shell should be removed carefully to avoid mechanical damage to the thin walls.
I also note that numerical simulation cannot replace trial production completely. The actual lost wax casting process includes many random factors, such as wax pattern dimensional variation, shell thickness variation, local shell permeability, alloy composition fluctuation, and pouring operator effects. However, when simulation is combined with orthogonal experimentation and intelligent optimization, it can greatly reduce the number of trial runs and provide a reliable direction for process improvement. In my work, the predicted defect locations agreed with the sectioned castings, and the optimized parameters produced a robust production window. This demonstrates that simulation-based lost wax casting optimization is valuable for small and medium-sized foundries that produce complex thin-walled stainless steel parts.
Overall, my study shows that a systematic approach to lost wax casting can control shrinkage cavity, shrinkage porosity, mistun, cold shut, and scarring in complex thin-walled bearing seats. The approach consists of structural analysis, pouring position selection, gating system design, numerical simulation of filling and solidification, trial casting verification, orthogonal experiment design, neural network modeling, particle swarm optimization, genetic algorithm searching, and final production validation. The final optimized lost wax casting process reduces the shrinkage volume by 44.74% and the equivalent stress by 2.41%, while maintaining or improving the process yield. The results provide a practical reference for the lost wax casting of similar thin-walled stainless steel components and for the application of intelligent optimization algorithms in foundry process design.