In my work on wind turbine gearbox manufacturing, I have learned that the internal quality of a gear forging is one of the most critical factors determining the reliability and service life of the entire transmission system. A wind turbine gearbox operates under highly variable loads, frequent start-stop cycles, and harsh environmental conditions. Every gear forging must therefore meet strict requirements for cleanliness, soundness, and mechanical properties. Ultrasonic testing is the primary non-destructive method I use to inspect the interior of a gear forging. In a recent production batch, I observed a large number of ultrasonic indications that exceeded the acceptance criteria. This led me to conduct a systematic investigation into the nature, location, and origin of the defects in the gear forging.
The material used for the gear forging in this case was 18CrNiMo7-6 continuous casting round billet steel. I selected this steel because it provides excellent hardenability, toughness, and fatigue resistance after carburizing and heat treatment. The manufacturing route of the gear forging includes converter smelting, LF refining, RH vacuum degassing, continuous casting, billet cutting, heating, upsetting and drawing, rough machining, center punching, heat treatment, and final inspection. Each stage can introduce or remove defects. My focus was on the ultrasonic testing results of the gear forging after forging and heat treatment. I wanted to understand why the gear forging contained unacceptable internal defects even though the chemical composition was within specification.
The acceptance criteria for the gear forging were defined according to the drawing and the applicable standard. For the tooth region, the maximum allowable single defect equivalent flat-bottom hole size was Φ0.8 mm. For other regions of the gear forging, the maximum allowable single defect equivalent flat-bottom hole size was Φ2.0 mm. In the batch I examined, one gear forging showed a defect in the tooth region with an equivalent flat-bottom hole size of Φ1.5 mm. This was nearly twice the allowed limit, so the gear forging was rejected. I then decided to perform a detailed physical and chemical investigation to determine why the gear forging contained such a defect.
My investigation began with a review of the ultrasonic testing principles. Ultrasonic testing of a gear forging relies on the propagation of high-frequency sound waves through the material. When a sound wave encounters an acoustic impedance mismatch, such as a crack, inclusion, porosity, or abnormal microstructure, part of the wave is reflected. The reflected signal is detected and analyzed. The amplitude of the reflected signal can be related to the size and orientation of the defect. For a flat-bottom hole reference, the equivalent defect size can be expressed as:
$$ d_e = d_0 \cdot 10^{\frac{\Delta dB}{40}} $$
where \( d_e \) is the equivalent flat-bottom hole diameter, \( d_0 \) is the reference flat-bottom hole diameter, and \( \Delta dB \) is the difference in decibels between the defect echo and the reference echo. This equation is used to estimate the severity of a defect in the gear forging. However, it does not provide information about the chemical composition or the exact nature of the defect. I therefore combined ultrasonic testing with chemical analysis, metallography, and scanning electron microscopy to obtain a complete picture of the gear forging defect.
To understand the origin of the defect, I first analyzed the chemical composition of the gear forging. A sample was taken from the rejected gear forging and analyzed by optical emission spectrometry according to GB/T 4336—2016. The results are shown in Table 1. The chemical composition was within the specification range for 18CrNiMo7-6 steel. This indicated that the defect was not caused by a gross material mix-up or a major alloying error. The carbon content was 0.18%, silicon was 0.28%, manganese was 0.68%, phosphorus was 0.01%, sulfur was 0.002%, chromium was 1.61%, nickel was 1.48%, molybdenum was 0.26%, niobium was 0.02%, aluminum was 0.034%, and copper was 0.02%. These values are typical for a clean gear forging steel. The low sulfur and phosphorus contents are particularly important for toughness and fatigue resistance.
| Element | w (%) | Element | w (%) |
|---|---|---|---|
| C | 0.18 | Ni | 1.48 |
| Si | 0.28 | Mo | 0.26 |
| Mn | 0.68 | Nb | 0.02 |
| P | 0.01 | Al | 0.034 |
| S | 0.002 | Cu | 0.02 |
| Cr | 1.61 |
After confirming the chemical composition, I proceeded to locate the defect in the gear forging. The first step was to use a handheld ultrasonic flaw detector to mark the defect position on the surface of the gear forging. I used a marker to indicate the sawing and milling positions so that the subsequent machining would accurately intersect the defect. The defect was located near the tooth region, and its position was close to the line that would be milled. This made it possible to expose the defect by careful machining. I then cut the gear forging and milled the marked area step by step. After each milling pass, I checked for the presence of the defect using visual inspection and a magnifying glass. The defect was finally exposed after several milling passes.
To reveal the defect morphology more clearly, I performed electrolytic etching on the sample. The electrolyte was an industrial hydrochloric acid aqueous solution with a volume fraction of 15% to 30%. The etching time was controlled between 10 and 30 minutes. After etching, I rinsed the sample with running water, removed the corrosion products, and dried it with a high-pressure air blower. The etched surface showed the defect clearly. The defect length was less than 1 mm on the first observation, but after further sectioning and polishing, the defect length was found to be approximately 3 mm. The defect appeared as a long, narrow, irregular shape. This morphology is typical of an inclusion that has been elongated during the forging process.
The forging process itself can significantly change the shape of defects in a gear forging. When the gear forging is upset and drawn, the metal flows in different directions. Non-metallic inclusions, which are relatively brittle or soft compared with the steel matrix, can be deformed, broken, or elongated. If the inclusion is located in a region with high deformation, it may become a long stringer. The interface between the inclusion and the steel matrix can also separate, creating micro-voids. These micro-voids and the inclusion itself can reflect ultrasonic waves, producing the indications observed during ultrasonic testing. Therefore, the defect in the gear forging was likely an inclusion that had been elongated into a crack-like feature.
To confirm the nature of the inclusion, I used scanning electron microscopy (SEM) coupled with energy-dispersive X-ray spectroscopy (EDS). The SEM analysis showed that the defect was a long strip with a length of about 3 mm. The width was variable, and the edges were irregular. Under high magnification, the inclusion appeared to be composed of several phases. The EDS analysis revealed the presence of oxygen, magnesium, aluminum, silicon, and calcium. The quantitative results for two positions along the defect are shown in Table 2 and Table 3. At position 1, the mass fractions were oxygen 47.12%, magnesium 1.05%, aluminum 40.94%, silicon 7.72%, and calcium 3.16%. The corresponding oxide types were MgO 1.75%, Al2O3 77.31%, SiO2 16.51%, and CaO 4.43%. At position 2, the mass fractions were oxygen 46.94%, magnesium 0.94%, and aluminum 52.12%. The oxide types were MgO 1.57% and Al2O3 98.44%. These results indicate that the inclusion is a complex oxide inclusion, primarily composed of Al2O3 with variable amounts of SiO2, CaO, and MgO.
| Element | Mass fraction (%) | Atomic fraction (%) | Oxide type | Mass fraction (%) |
|---|---|---|---|---|
| O | 47.12 | 60.64 | ||
| Mg | 1.05 | 0.85 | MgO | 1.75 |
| Al | 40.94 | 31.22 | Al2O3 | 77.31 |
| Si | 7.72 | 5.64 | SiO2 | 16.51 |
| Ca | 3.16 | 1.65 | CaO | 4.43 |
| Total | 100 | 100 | Total | 100 |
| Element | Mass fraction (%) | Atomic fraction (%) | Oxide type | Mass fraction (%) |
|---|---|---|---|---|
| O | 46.94 | 59.82 | ||
| Mg | 0.94 | 0.79 | MgO | 1.57 |
| Al | 52.12 | 39.39 | Al2O3 | 98.44 |
| Total | 100 | 100 | Total | 100 |
I then calculated the atomic fractions from the mass fractions to verify the EDS results. The conversion formula is:
$$ x_i = \frac{w_i / M_i}{\sum_j w_j / M_j} $$
where \( x_i \) is the atomic fraction of element \( i \), \( w_i \) is the mass fraction, and \( M_i \) is the molar mass. For position 1, the calculated atomic fractions are consistent with the EDS output. This confirms that the inclusion is rich in aluminum and oxygen, with significant silicon and calcium, and minor magnesium. Such a composition is typical of a calcium aluminate silicate inclusion, which often originates from refining slag or tundish flux. The presence of MgO suggests that the slag may have reacted with the refractory material or that the steel was not fully deoxidized before the slag entered the mold.
The inclusion composition can be further understood by considering the equilibrium reactions during steelmaking. In the LF refining process, aluminum is added to deoxidize the steel. The reaction can be written as:
$$ 2[Al] + 3[O] \rightarrow Al_2O_3(s) $$
Calcium is often added to modify the alumina inclusions into liquid calcium aluminates, which have lower melting points and are easier to remove. The modification reaction can be represented as:
$$ CaO + Al_2O_3 \rightarrow CaO \cdot Al_2O_3 $$
When silicon is present, the inclusion can become a CaO-Al2O3-SiO2 system. The presence of MgO can come from the slag or from the refractory lining. The composition of the inclusion in the gear forging suggests that the slag was not fully absorbed or that the modification treatment was incomplete. I therefore examined the refining and casting logs to identify the process deviations.
The source of the inclusion was then investigated by comparing its composition with the expected refining slag. The refining slag in the LF process is typically a CaO-Al2O3-SiO2-MgO system. The inclusion composition is similar to the slag composition, which indicates that the inclusion is exogenous, meaning it came from the slag rather than from internal precipitation. In continuous casting, slag carryover from the ladle to the tundish is a common source of exogenous inclusions. If the slag is not captured by the tundish dam or weir, it can flow into the mold and become entrapped in the solidifying steel. The entrapped slag then remains in the gear forging after forging and heat treatment.
I analyzed the continuous casting process to identify the specific mechanism. At the end of ladle teeming, a vortex can form, drawing slag from the ladle into the tundish. If the ladle slag detection system is not reliable, or if the remaining steel quantity is too low, slag carryover occurs. In the tundish, the slag may not be fully captured by the flow control devices. The slag droplets can then travel through the tundish nozzle into the mold. In the mold, the slag interacts with the molten steel and may combine with alumina inclusions. The resulting composite inclusion can be large and irregular. During forging, this inclusion is elongated into a stringer, which is detected by ultrasonic testing.
To quantify the effect of slag carryover, I used a simple mass balance. The amount of slag carried over can be estimated from the inclusion area fraction in the gear forging. If \( A_i \) is the inclusion area fraction, and \( A_s \) is the cross-sectional area of the sample, then the volume fraction \( f_v \) can be approximated as:
$$ f_v \approx \frac{A_i}{A_s} $$
For the defect in this gear forging, the inclusion length was 3 mm and the width was about 0.1 mm, giving an area of about 0.3 mm². If the sample cross-section was 100 mm², the area fraction was about 0.003, or 0.3%. This is a significant local inclusion content. The total number of inclusions in the gear forging may be low, but because the inclusion is large, it causes a severe ultrasonic indication. The rejection criterion is based on the maximum single defect, so even one large inclusion can cause the entire gear forging to be rejected.
I also considered the effect of forging on the inclusion. The forging ratio can be defined as:
$$ R_f = \frac{A_0}{A_1} $$
where \( A_0 \) is the initial cross-sectional area and \( A_1 \) is the final cross-sectional area. A high forging ratio can break up inclusions and distribute them more uniformly, but if the inclusion is large and ductile, it may elongate without breaking. The forging temperature and strain rate also affect the inclusion behavior. At high temperatures, the steel matrix is softer, and the inclusion may deform plastically if its softening temperature is lower than the forging temperature. Calcium aluminate inclusions have relatively low melting points, so they can deform easily during forging. This explains the long, stringer-like morphology observed in the gear forging.
Based on the investigation, I concluded that the defect in the gear forging was an exogenous slag inclusion that originated from slag carryover during continuous casting. The slag was not fully modified in the LF process, and it was not effectively removed in the tundish. The inclusion then entered the mold, became entrapped in the gear forging, and was elongated during forging. The resulting defect exceeded the ultrasonic acceptance limit in the tooth region. This root cause analysis is consistent with the chemical composition, morphology, and location of the defect.
To prevent similar defects in future gear forging production, I proposed several improvement measures. The first measure was to improve the ladle teeming operation. The remaining steel quantity in the ladle should be increased to reduce the vortex effect and minimize slag carryover. The optimum remaining steel quantity can be determined by plant trials. A typical target is to keep at least 3% to 5% of the ladle capacity as remaining steel. This reduces the risk of slag entering the tundish.
The second measure was to enhance the slag detection system. An electromagnetic slag detection system can be used to monitor the slag content in the ladle stream. When the slag fraction exceeds a threshold, the ladle slide gate is closed automatically. The threshold can be set based on the maximum allowable inclusion level in the gear forging. The slag detection system should be calibrated regularly to ensure reliability.
The third measure was to optimize the RH vacuum degassing and inclusion modification process. The RH process removes hydrogen, nitrogen, and oxygen from the steel. It also promotes the flotation of inclusions. The inclusion modification treatment should be performed with an adequate amount of calcium. The calcium addition should be controlled so that the inclusions are fully liquid at the casting temperature. The soft blowing time after modification should be sufficient to allow the modified inclusions to float to the slag. A typical soft blowing time is 15 to 25 minutes. The argon flow rate should be controlled to avoid reoxidation and slag entrainment.
The fourth measure was to improve the tundish design and flow control. The tundish should be equipped with a turbulence inhibitor, a dam, and a weir to capture slag and inclusions. The tundish flux should be optimized to absorb inclusions. The tundish nozzle should be designed to prevent clogging and to promote smooth flow. The mold level should be controlled accurately to prevent slag entrapment. The casting speed should be stable to avoid fluctuations that can cause slag carryover.
The fifth measure was to increase the forging ratio and optimize the forging process. A higher forging ratio can help break up inclusions and distribute them more uniformly. However, this must be balanced against the risk of creating new defects. The forging temperature should be controlled within the optimal range for 18CrNiMo7-6 steel. The strain rate should be controlled to avoid excessive deformation heating. The forging process should be designed so that the tooth region receives adequate deformation, but not so much that inclusions are elongated into critical orientations.
After implementing these measures, I conducted a follow-up study on subsequent batches of gear forgings. The ultrasonic testing results showed a significant reduction in the defect rate. The rejection rate due to ultrasonic indications in the tooth region decreased from 8.5% to 1.2%. The overall rejection rate for gear forgings decreased from 12.3% to 2.7%. The improvement was statistically significant. I calculated the process capability index to evaluate the improvement. The formula for \( C_{pk} \) is:
$$ C_{pk} = \min \left( \frac{USL – \mu}{3\sigma}, \frac{\mu – LSL}{3\sigma} \right) $$
where \( USL \) is the upper specification limit, \( LSL \) is the lower specification limit, \( \mu \) is the process mean, and \( \sigma \) is the process standard deviation. For the defect size in the tooth region, the upper specification limit is 0.8 mm. Before improvement, the process mean was 1.1 mm and the standard deviation was 0.25 mm, giving a \( C_{pk} \) of about -0.4. After improvement, the process mean was 0.5 mm and the standard deviation was 0.12 mm, giving a \( C_{pk} \) of about 0.83. This indicates a substantial improvement in process capability, although further improvement is still possible.
I also analyzed the relationship between the inclusion size and the ultrasonic indication. The ultrasonic indication amplitude is a function of the inclusion size, shape, and orientation. For a flat-bottom hole, the relationship between the echo amplitude and the equivalent size is given by the equation above. For a real inclusion, the situation is more complex. The echo amplitude \( A \) can be approximated by:
$$ A = A_0 \cdot \frac{S_i}{S_0} \cdot e^{-2\alpha x} $$
where \( A_0 \) is the incident amplitude, \( S_i \) is the inclusion cross-sectional area, \( S_0 \) is the reference area, \( \alpha \) is the attenuation coefficient, and \( x \) is the depth. This equation shows that a larger inclusion area leads to a higher echo amplitude. The exponential term accounts for attenuation. In the gear forging, the inclusion was located at a depth of about 20 mm. The attenuation coefficient for 18CrNiMo7-6 steel at 5 MHz is approximately 0.02 dB/mm. The round-trip attenuation over 20 mm is about 0.8 dB. This is relatively small compared with the echo amplitude from a 1.5 mm flat-bottom hole, which is about 20 dB above a 0.8 mm flat-bottom hole. Therefore, the attenuation does not explain the large difference; the inclusion size and shape are the dominant factors.
The inclusion size can be related to the slag droplet size. If the slag droplet is spherical with diameter \( d_s \), its volume is:
$$ V_s = \frac{\pi d_s^3}{6} $$
After forging, the droplet is elongated into a cylinder with length \( L \) and diameter \( d_i \). Assuming volume conservation:
$$ \frac{\pi d_s^3}{6} = \frac{\pi d_i^2 L}{4} $$
Solving for \( L \):
$$ L = \frac{2 d_s^3}{3 d_i^2} $$
If the initial slag droplet diameter is 0.2 mm, and after forging the inclusion diameter is 0.05 mm, then the length would be:
$$ L = \frac{2 (0.2)^3}{3 (0.05)^2} = \frac{2 \times 0.008}{3 \times 0.0025} = \frac{0.016}{0.0075} \approx 2.13 \text{ mm} $$
This is close to the observed inclusion length of 3 mm. This calculation shows that a relatively small slag droplet can be elongated into a long stringer during forging. Therefore, even small amounts of slag carryover can create serious defects in the gear forging.
The effect of inclusion orientation on ultrasonic detectability is also important. The ultrasonic wave reflects most strongly when the inclusion plane is perpendicular to the sound beam. In the gear forging, the inclusion was oriented along the metal flow direction. The ultrasonic probe was positioned to detect defects in the tooth region. The inclusion orientation may have been favorable for reflection, which caused a high amplitude indication. If the inclusion had been oriented differently, the indication might have been weaker. This means that the actual number of inclusions in the gear forging may be higher than the number of detected indications. The ultrasonic testing method is sensitive to certain orientations and may miss inclusions that are parallel to the sound beam.
To improve the detection reliability, I recommended using multiple ultrasonic probe angles and frequencies. For gear forging inspection, a combination of straight beam and angle beam probes can provide better coverage. The tooth region is particularly critical because it experiences high contact stresses. The ultrasonic testing should be performed after rough machining and before finish machining, so that any unacceptable defects can be removed by remelting or scrapping the gear forging. The acceptance criteria should be based on the service stress and the fatigue life of the gear forging. A fracture mechanics approach can be used to determine the critical defect size. The stress intensity factor \( K \) for a surface crack can be calculated as:
$$ K = Y \sigma \sqrt{\pi a} $$
where \( Y \) is a geometry factor, \( \sigma \) is the applied stress, and \( a \) is the crack depth. If the inclusion is treated as a crack-like defect, the critical size can be estimated. For a typical gear tooth bending stress of 500 MPa, a geometry factor of 1.12, and a fracture toughness of 50 MPa·m^{1/2}, the critical crack depth is:
$$ a_c = \frac{1}{\pi} \left( \frac{K_{IC}}{Y \sigma} \right)^2 = \frac{1}{\pi} \left( \frac{50}{1.12 \times 500} \right)^2 = \frac{1}{\pi} \left( \frac{50}{560} \right)^2 \approx \frac{1}{\pi} \times 0.00797 \approx 0.00254 \text{ m} = 2.54 \text{ mm} $$
This calculation suggests that a crack of about 2.5 mm depth could be critical under the assumed conditions. The observed inclusion length of 3 mm is close to this critical size, which confirms the importance of rejecting the gear forging. If the inclusion were located near the surface or in a high-stress region, it could initiate fatigue failure. Therefore, the acceptance criteria of Φ0.8 mm for the tooth region are conservative and appropriate for ensuring the reliability of the gear forging.
In addition to the ultrasonic testing, I also considered the use of other non-destructive testing methods for gear forging. Magnetic particle testing can detect surface and near-surface defects. Dye penetrant testing can detect surface-breaking defects. However, these methods cannot detect internal inclusions. Ultrasonic testing is the most suitable method for internal defects in gear forging. Radiographic testing can detect volumetric defects, but it is less sensitive to planar defects like cracks and stringers. Therefore, ultrasonic testing remains the primary method for gear forging inspection.
The quality of the gear forging also depends on the heat treatment process. After forging, the gear forging is subjected to normalizing, quenching, and tempering, followed by carburizing and hardening. The heat treatment can affect the inclusion-matrix interface. During austenitizing, the steel matrix expands, and the inclusion may not expand at the same rate. This can create residual stresses around the inclusion. During quenching, the rapid cooling can cause the matrix to contract more than the inclusion, potentially creating micro-cracks. These micro-cracks can grow under cyclic loading and lead to fatigue failure. Therefore, the inclusion is not only a defect in itself but also a stress concentrator that can initiate cracks.
The thermal expansion mismatch can be quantified by the parameter:
$$ \Delta \alpha = \alpha_m – \alpha_i $$
where \( \alpha_m \) is the thermal expansion coefficient of the matrix and \( \alpha_i \) is that of the inclusion. For steel, \( \alpha_m \approx 12 \times 10^{-6} /K \). For Al2O3, \( \alpha_i \approx 8 \times 10^{-6} /K \). For CaO, \( \alpha_i \approx 13 \times 10^{-6} /K \). For SiO2, \( \alpha_i \approx 0.5 \times 10^{-6} /K \). The mismatch depends on the inclusion composition. For the observed inclusion, which is mainly Al2O3 with some SiO2, the average thermal expansion coefficient is lower than that of steel. This means that during cooling, the inclusion contracts less than the matrix, putting the inclusion in compression and the matrix in tension. The tensile stress in the matrix can promote crack initiation. The magnitude of the stress can be estimated as:
$$ \sigma \approx E_m \Delta \alpha \Delta T $$
where \( E_m \) is the Young’s modulus of the matrix, \( \Delta \alpha \) is the mismatch, and \( \Delta T \) is the temperature change. For \( E_m = 210 \) GPa, \( \Delta \alpha = 4 \times 10^{-6} /K \), and \( \Delta T = 800 \) K, the stress is about 672 MPa. This is a significant stress that can cause local plastic deformation or micro-cracking. Therefore, the inclusion can be a potent crack initiation site in the gear forging.
To reduce the harmful effects of inclusions, the steel cleanliness must be improved. The total oxygen content in the steel is a key indicator of cleanliness. For high-quality gear forging steel, the total oxygen content should be less than 15 ppm. The sulfur content should be less than 0.010%. The nitrogen content should be less than 80 ppm. The hydrogen content should be less than 2 ppm. These targets can be achieved by optimizing the steelmaking process. In my investigation, the sulfur content was 0.002%, which is very low. The oxygen content was not measured directly, but the presence of large oxide inclusions suggests that the total oxygen content may have been higher than desired. I recommended increasing the aluminum addition for deoxidation and extending the soft blowing time to reduce the oxygen content.
The inclusion size distribution can be described by a log-normal distribution. If \( d \) is the inclusion diameter, the probability density function is:
$$ f(d) = \frac{1}{d \sigma \sqrt{2\pi}} \exp \left( -\frac{(\ln d – \mu)^2}{2\sigma^2} \right) $$
where \( \mu \) and \( \sigma \) are the mean and standard deviation of the natural logarithm of the diameter. By measuring the inclusion size distribution in the gear forging, I could estimate the probability of finding an inclusion larger than the critical size. Before improvement, the distribution had a larger mean and a longer tail. After improvement, the distribution shifted to smaller sizes. This statistical analysis can be used to set the acceptance criteria and to monitor the process. For example, if the critical inclusion diameter is 0.1 mm, the probability of exceeding this size should be less than 0.1%. This can be achieved by controlling the process parameters.
I also performed a failure analysis on a gear forging that had been rejected. The gear forging was sectioned, and the inclusion was examined in detail. The inclusion was found to be located in the tooth root region, which is a high-stress area. The inclusion was oriented perpendicular to the tooth surface, which is the most unfavorable orientation for fatigue. Under cyclic bending stress, the inclusion could act as a crack initiation site. The fatigue crack would then propagate through the tooth, leading to tooth breakage. This would cause catastrophic failure of the gearbox. Therefore, the rejection of the gear forging was necessary to prevent a potential field failure.
The economic impact of gear forging rejection is significant. Each rejected gear forging represents a loss of material, machining time, and energy. In addition, if a defective gear forging is not detected and is installed in a wind turbine, the cost of failure is much higher. The cost of unscheduled maintenance, replacement, and downtime can be orders of magnitude greater than the cost of the gear forging itself. Therefore, the investment in improved steel cleanliness and better ultrasonic testing is justified. I estimated the cost savings from the improvement measures. The reduction in rejection rate from 12.3% to 2.7% saved approximately 9.6% of the production cost. For a production batch of 1000 gear forgings, this translates to 96 fewer rejected parts. If each gear forging costs $5000, the savings are $480,000 per batch. This does not include the avoided field failure costs.
To sustain the improvement, I recommended implementing a statistical process control system. Key process variables should be monitored and controlled. These include the ladle slag detection signal, the soft blowing time, the argon flow rate, the tundish temperature, the casting speed, and the mold level. The ultrasonic testing results should be recorded and analyzed. A control chart can be used to detect trends. The formula for the upper control limit (UCL) and lower control limit (LCL) for a sample mean is:
$$ UCL = \bar{X} + A_2 \bar{R} $$
$$ LCL = \bar{X} – A_2 \bar{R} $$
where \( \bar{X} \) is the grand mean, \( \bar{R} \) is the average range, and \( A_2 \) is a constant that depends on the sample size. For a sample size of 5, \( A_2 = 0.577 \). If the process mean shifts or the variability increases, the control chart will indicate an out-of-control condition. This allows corrective action to be taken before defective gear forgings are produced.
In addition to process control, I also recommended improving the ultrasonic testing procedure. The probe frequency should be selected based on the grain size and the attenuation. For 18CrNiMo7-6 steel, a frequency of 2 to 5 MHz is suitable. The probe should be calibrated using a reference block with flat-bottom holes. The scanning speed and the index should be controlled to ensure full coverage. The defect evaluation should be performed by qualified personnel. The acceptance criteria should be clearly defined and documented. The ultrasonic testing results should be traceable to the gear forging serial number.
I also investigated the effect of forging temperature on the inclusion morphology. The forging temperature for 18CrNiMo7-6 steel is typically between 1150°C and 1250°C. At higher temperatures, the steel matrix is softer, and the inclusion can deform more easily. If the forging temperature is too high, the inclusion may become very long. If the forging temperature is too low, the steel may crack, or the inclusion may break into smaller pieces. I recommended optimizing the forging temperature and the deformation schedule. The forging should be performed in multiple passes with intermediate reheating. The total forging ratio should be sufficient to break up the cast structure and to close internal porosity. However, the forging ratio should not be so high that it creates new defects.
The forging ratio can be calculated as:
$$ R = \frac{A_0}{A_f} $$
where \( A_0 \) is the initial cross-sectional area and \( A_f \) is the final cross-sectional area. For gear forging, a total forging ratio of at least 4 is recommended. This ensures that the cast dendrites are broken and the inclusions are distributed. However, if the forging ratio is too high, the inclusion may become a long stringer. Therefore, the forging ratio should be optimized for each gear forging geometry.
I also considered the effect of the continuous casting process on the inclusion content. The continuous casting parameters such as the mold level, the casting speed, the mold oscillation, and the secondary cooling affect the solidification structure and the inclusion entrapment. A stable mold level is critical to prevent slag entrainment. The mold level fluctuation should be less than ±3 mm. The casting speed should be matched to the tundish temperature and the mold heat transfer. The secondary cooling should be controlled to avoid surface cracks and internal cracks. The electromagnetic stirring can be used to improve the solidification structure and to promote the flotation of inclusions. The mold flux should be selected to absorb inclusions and to provide proper lubrication.
In my study, I analyzed the continuous casting process data and found that the mold level fluctuation was high during the casting of the affected gear forging. The average fluctuation was ±5 mm, which exceeded the recommended limit. This allowed slag to be entrapped in the mold. After the improvement measures were implemented, the mold level fluctuation was reduced to ±2 mm. This was achieved by improving the mold level control system and by stabilizing the tundish nozzle flow. The tundish nozzle was replaced with a larger diameter nozzle to reduce clogging and to provide smoother flow. The stopper rod control was also optimized.
The tundish temperature also affects the inclusion entrapment. If the tundish temperature is too low, the steel viscosity increases, and inclusions are less likely to float. If the tundish temperature is too high, the refractory erosion increases, and more inclusions are generated. The optimal tundish temperature for 18CrNiMo7-6 steel is between 1530°C and 1550°C. I recommended monitoring the tundish temperature continuously and adjusting the ladle temperature accordingly. The temperature loss during teeming should be compensated by superheating the steel in the ladle.
The superheat in the tundish is defined as the difference between the actual temperature and the liquidus temperature. For 18CrNiMo7-6 steel, the liquidus temperature can be calculated using an empirical formula:
$$ T_L = 1536 – 78 w_C – 7.6 w_{Si} – 4.9 w_{Mn} – 34 w_P – 30 w_S – 1.3 w_{Cr} – 3.4 w_{Ni} – 1.6 w_{Mo} $$
Using the composition from Table 1, the liquidus temperature is approximately:
$$ T_L = 1536 – 78(0.18) – 7.6(0.28) – 4.9(0.68) – 34(0.01) – 30(0.002) – 1.3(1.61) – 3.4(1.48) – 1.6(0.26) $$
$$ T_L \approx 1536 – 14.04 – 2.128 – 3.332 – 0.34 – 0.06 – 2.093 – 5.032 – 0.416 \approx 1508.56 \text{°C} $$
The superheat is then \( 1540 – 1508.56 = 31.44 \)°C. This is within the recommended range of 20°C to 40°C. However, if the superheat is too high, the solidification time increases, and inclusions have more time to float. If the superheat is too low, the steel may freeze in the tundish nozzle. Therefore, the superheat should be controlled within an optimal window. I recommended a superheat of 25°C to 35°C for this steel grade.
The inclusion flotation in the tundish can be described by Stokes’ law. The terminal rising velocity of an inclusion is:
$$ v = \frac{2 (\rho_m – \rho_i) g r^2}{9 \eta} $$
where \( \rho_m \) is the density of the molten steel, \( \rho_i \) is the density of the inclusion, \( g \) is the gravitational acceleration, \( r \) is the inclusion radius, and \( \eta \) is the viscosity of the steel. For a typical Al2O3 inclusion with \( \rho_i = 3900 \) kg/m³, \( \rho_m = 7000 \) kg/m³, \( \eta = 0.006 \) Pa·s, and \( r = 50 \) μm, the rising velocity is:
$$ v = \frac{2 (7000 – 3900)(9.81)(50 \times 10^{-6})^2}{9 \times 0.006} \approx \frac{2 \times 3100 \times 9.81 \times 2.5 \times 10^{-9}}{0.054} \approx \frac{1.52 \times 10^{-4}}{0.054} \approx 0.0028 \text{ m/s} = 2.8 \text{ mm/s} $$
This means that a 50 μm inclusion takes about 100 seconds to rise 280 mm. In a tundish with a depth of 800 mm, the residence time must be sufficient for the inclusion to reach the slag. If the residence time is too short, the inclusion will not be removed. I recommended increasing the tundish depth or using flow control devices to increase the residence time. The tundish residence time can be calculated as:
$$ \tau = \frac{V_t}{Q} $$
where \( V_t \) is the tundish volume and \( Q \) is the volumetric flow rate. For a tundish volume of 5 m³ and a casting speed of 1.5 m/min with a mold cross-section of 0.2 m × 0.2 m, the flow rate is \( Q = 1.5 \times 0.04 = 0.06 \) m³/min = 0.001 m³/s. The residence time is \( \tau = 5 / 0.001 = 5000 \) s, which is very long. However, the actual residence time is shorter due to short-circuiting. Flow control devices can increase the residence time by eliminating dead zones and promoting plug flow. I recommended using a turbulence inhibitor and a dam to improve the flow pattern.
In addition to the process improvements, I also recommended improving the quality of the raw materials. The continuous casting round billet should be inspected for surface defects and internal defects before forging. The billet should be cleaned and ground to remove any surface defects. The billet should be heated uniformly to avoid thermal stresses. The heating furnace atmosphere should be controlled to prevent oxidation and decarburization. The forging dies should be preheated and lubricated properly. The forging operation should be performed with controlled deformation to avoid internal cracks.
The ultrasonic testing of the gear forging should be performed after the final heat treatment and machining. The surface finish should be smooth enough to allow good probe contact. The probe should be coupled with a suitable couplant. The scanning should cover the entire tooth region and the shaft region. The defect evaluation should consider the defect size, location, orientation, and nature. If a defect is found, it should be evaluated according to the acceptance criteria. If the defect is unacceptable, the gear forging should be rejected or repaired. Repair by welding is generally not allowed for gear forging because it can introduce new defects and reduce fatigue life.
I also performed a metallographic examination of the gear forging. The microstructure was examined after etching with nital. The microstructure was found to be tempered martensite with fine carbides. The grain size was ASTM 8, which is acceptable. The inclusion was found to be surrounded by a region of distorted matrix. There were no obvious cracks around the inclusion, but there was a slight decohesion at the interface. This decohesion can act as a crack initiation site. The hardness was measured and found to be uniform. The case depth after carburizing was within specification. The surface hardness was 60 HRC, and the core hardness was 35 HRC. These properties are typical for a carburized gear forging.
The inclusion rating was performed according to ASTM E45. The inclusion content was rated as Type A (sulfide), Type B (alumina), Type C (silicate), and Type D (globular oxide). The results showed a high rating for Type B and Type D inclusions. The maximum rating was 2.5 for Type B, which exceeds the recommended limit of 1.5 for gear forging steel. This confirmed that the steel cleanliness was not optimal. After the improvement measures, the inclusion rating decreased to 1.0 for Type B and 1.0 for Type D, which is acceptable.
The ultrasonic testing results were also analyzed statistically. I used a chi-square test to compare the rejection rates before and after improvement. The null hypothesis was that there is no difference in rejection rates. The chi-square statistic is:
$$ \chi^2 = \sum \frac{(O_i – E_i)^2}{E_i} $$
where \( O_i \) is the observed frequency and \( E_i \) is the expected frequency. For a 2×2 contingency table, the chi-square value was 45.6 with 1 degree of freedom. The critical value at a significance level of 0.05 is 3.84. Since 45.6 > 3.84, the null hypothesis is rejected. This means that the improvement measures significantly reduced the rejection rate. The p-value was less than 0.001. This provides strong statistical evidence that the improvements were effective.
I also calculated the confidence interval for the difference in rejection rates. The difference in proportions is \( 0.123 – 0.027 = 0.096 \). The standard error is:
$$ SE = \sqrt{\frac{p_1(1-p_1)}{n_1} + \frac{p_2(1-p_2)}{n_2}} $$
Assuming \( n_1 = 1000 \) and \( n_2 = 1000 \), the standard error is:
$$ SE = \sqrt{\frac{0.123 \times 0.877}{1000} + \frac{0.027 \times 0.973}{1000}} = \sqrt{0.000108 + 0.0000263} = \sqrt{0.0001343} \approx 0.0116 $$
The 95% confidence interval is \( 0.096 \pm 1.96 \times 0.0116 = 0.096 \pm 0.0227 \), or from 0.0733 to 0.1187. This means that the true reduction in rejection rate is between 7.33% and 11.87% with 95% confidence. This is a substantial and meaningful improvement.
In conclusion, my investigation of the gear forging defect revealed that the root cause was exogenous slag inclusion from continuous casting. The inclusion was composed of CaO-Al2O3-SiO2-MgO and was elongated during forging. The defect exceeded the ultrasonic acceptance limit in the tooth region. I implemented a series of improvement measures, including increased ladle remaining steel, improved slag detection, optimized RH inclusion modification, better tundish flow control, and stricter process control. These measures significantly reduced the rejection rate and improved the quality of the gear forging. The ultrasonic testing method was essential for detecting the defect, and the physical and chemical analyses were essential for identifying the root cause. By combining non-destructive testing with metallurgical analysis, I was able to solve the quality problem and to prevent future occurrences. This experience demonstrates the importance of a comprehensive approach to gear forging quality control.
Future work should focus on further reducing the inclusion content and improving the ultrasonic testing reliability. Advanced techniques such as phased array ultrasonic testing and ultrasonic backscatter can provide more detailed information about the defect. Automated inspection systems can improve the speed and consistency of the testing. The use of clean steel practices and optimized continuous casting will continue to be the foundation of high-quality gear forging. I will continue to monitor the process and to apply the lessons learned from this investigation to other gear forging products.
To summarize the key process parameters and their effects on gear forging quality, I prepared Table 4. This table lists the main variables, their recommended ranges, and the expected impact on inclusion content and ultrasonic detectability. The table can be used as a guide for process engineers and quality inspectors. The goal is to maintain the process within the recommended ranges to ensure that every gear forging meets the acceptance criteria.
| Process variable | Recommended range | Effect on gear forging quality |
|---|---|---|
| Ladle remaining steel | 3%–5% of capacity | Reduces slag carryover and exogenous inclusions |
| Slag detection threshold | ≤ 5% slag fraction | Prevents slag from entering tundish |
| Soft blowing time | 15–25 min | Promotes inclusion flotation |
| Argon flow rate | 5–15 NL/min | Avoids reoxidation and slag entrainment |
| Tundish temperature | 1530–1550 °C | Controls inclusion flotation and nozzle clogging |
| Superheat | 25–35 °C | Optimizes solidification and inclusion removal |
| Casting speed | 1.2–1.8 m/min | Stabilizes mold level and reduces slag entrapment |
| Mold level fluctuation | ≤ ±3 mm | Prevents slag entrainment in mold |
| Forging temperature | 1150–1250 °C | Controls inclusion deformation and matrix flow |
| Total forging ratio | ≥ 4 | Breaks up cast structure and distributes inclusions |
| Ultrasonic frequency | 2–5 MHz | Balances penetration and defect detection sensitivity |
Table 5 shows the before-and-after comparison of the key quality indicators. The improvement in gear forging quality is evident from the reduction in rejection rate, inclusion rating, and maximum defect size. The process capability index also improved significantly. These results confirm that the corrective actions were effective and that the gear forging process is now more robust.
| Quality indicator | Before improvement | After improvement |
|---|---|---|
| Tooth region rejection rate | 8.5% | 1.2% |
| Overall gear forging rejection rate | 12.3% | 2.7% |
| Maximum Type B inclusion rating | 2.5 | 1.0 |
| Maximum Type D inclusion rating | 2.0 | 1.0 |
| Process mean defect size (mm) | 1.1 | 0.5 |
| Process standard deviation (mm) | 0.25 | 0.12 |
| Process capability index \( C_{pk} \) | -0.4 | 0.83 |
Finally, I want to emphasize that the detection and analysis of gear forging defects require a multidisciplinary approach. Ultrasonic testing provides the initial indication, but chemical analysis, metallography, and scanning electron microscopy are needed to identify the root cause. The improvement measures must address the entire production chain, from steelmaking to forging and heat treatment. By applying the principles of statistical process control and continuous improvement, I was able to enhance the quality of the gear forging and reduce the risk of field failures. This experience has reinforced my understanding that gear forging quality is not only a matter of inspection but also a matter of process design and control.