Our team was tasked with developing a new high-performance engine cylinder block for heavy-duty applications. The casting weighs 224 kg, features complex geometry with many bosses and ribs, and has a minimum wall thickness of only 6 mm. Production is fully automated on a KW static pressure molding line, using cold‑box cores and robotic handling. Initial internal scrap rates reached 15%, with rework rates exceeding 50%. Through systematic analysis and process improvements over 18 months, we reduced the scrap rate to approximately 3% and dramatically improved cleaning efficiency. This paper details the sand casting defects encountered, the root causes, and the effective countermeasures implemented.
Overview of Sand Casting Defects in the Initial Production Phase
The major sand casting defects observed are summarized in the table below. The defect distribution was dominated by mechanical damage, loose sand, sticker sand on cores, and erosion (sand wash).
| Defect Category | Description | Initial Frequency (%) | Final Frequency (%) |
|---|---|---|---|
| Rough cleaning damage | Broken bosses and ribs during robotic handling | 3.2 | 0.5 |
| Core print damage (metal tearing) | Metal pulled away when removing vent pins | 2.8 | 0.3 |
| Sand wash (erosion) | Scouring of mold surface near gates | 2.5 | 0.4 |
| Loose sand / inclusions | Sand falling into cavity from misaligned vent holes | 2.0 | 0.2 |
| Sticker sand on tappet cores | Burned-on sand in 20 mm diameter holes | 1.8 | 0.1 |
| Internal cavity sand inclusions | Sand from core flash not cleaned | 1.5 | 0.3 |
| Other (blowholes, shrinkage) | Minor defects | 1.2 | 0.5 |
Each of these sand casting defects required a tailored solution, as described in the following sections.
Damage Defects: Rough Cleaning and Core Print Tearing
Two distinct damage mechanisms were identified. The first was mechanical damage during robotic rough cleaning. The cylinder block has many small bosses (16 on the top face alone) that are easily fractured by the robot gripper. The second was metal tearing when workers broke off vent pins cast into the bosses, leaving “craters” that required welding.
Rough Cleaning Damage
We observed that bosses without support ribs were particularly vulnerable. The robot gripper applied a clamping force of approximately 600 N, and the stress concentration at the boss base exceeded the local yield strength of the cast iron (∼200 MPa). To solve this, we added small reinforcing ribs to each critical boss, increasing the cross‑sectional area by 40%. The stress was reduced below 90 MPa, as estimated by:
$$ \sigma = \frac{F}{A} \approx \frac{600\ \text{N}}{1.2 \times 10^{-3}\ \text{m}^{2}} = 0.5\ \text{MPa} $$
Furthermore, we standardized the robot gripping position and created a visual work instruction. This reduced the variability in operator technique. After these changes, the damage rate dropped from 3.2% to 0.5%.
Core Print (Vent Pin) Tearing
Each boss originally carried a cylindrical vent pin with a simple radius at the base. When the pin was knocked off, the fracture often propagated into the boss, removing a chunk of metal. We redesigned the pin to include a necked‑down “step” at the base, and reduced the pin diameter so that the pin cross‑section occupied only 80% of the boss area. The step acts as a stress‑raising notch, ensuring fracture occurs cleanly at the step, leaving the boss intact. The improvement is quantified in the table below.
| Parameter | Original Design | Redesigned Pin |
|---|---|---|
| Pin base geometry | Radius only | Step + 45° chamfer |
| Pin diameter / boss diameter | 90% | 80% |
| Metal‑tearing frequency | 2.8% | 0.3% |
Sand Wash (Erosion) Defects
Erosion occurred predominantly on the lower surface of the casting, at the intersection of two strengthening ribs. The mold surface at these acute corners had low compaction, and the molten iron velocity was high due to an unbalanced gating system. After pouring, the eroded sand was carried into the water jacket, creating leakers.
Four corrective actions were taken simultaneously:
- Molding parameters: We increased the squeeze pressure from 8.5 bar to 9.5 bar and added 5% more sand per cycle. Mold hardness was measured using a B‑scale tester; values rose from 85 to 92.
- Sand properties: The moisture content was lowered from 3.4% to 2.9%, and the green compression strength was raised to 180 kPa by increasing bentonite addition. The relationship between strength and moisture was modeled as:
$$ \sigma_{gc} = 250 – 80 \cdot (M – 2.8) \quad (\text{for } M \in [2.5, 3.2]) $$
where $\sigma_{gc}$ is green compression strength in kPa and $M$ is moisture percentage. This linear approximation helped us maintain consistent properties.
- Geometry change: The rib‑to‑body fillet radius was increased from 2 mm to 5 mm, eliminating sharp corners that are difficult to compact.
- Gating system revision: The original system had six ingates (two each on bearing caps 1,2,4,6,7) but none on caps 3 and 5. The iron flow was concentrated at cap 4, causing intense scouring. We added two additional ingates on caps 3 and 5, redistributing the flow. The new system achieved a 30% reduction in flow velocity at critical areas, as calculated from:
$$ v = \frac{Q}{A_{total}} $$
where $Q$ is the pour rate (constant at 40 kg/s) and $A_{total}$ is the total ingate area. After modification, $A_{total}$ increased from 18 cm² to 24 cm², reducing $v$ from 2.2 m/s to 1.7 m/s. The erosion defects decreased from 2.5% to 0.4%.
Loose Sand and Surface Roughness (Orange Peel)
Two related sand casting defects stemmed from the vent‑drilling operation. After molding, vent holes were drilled through the sand to reach the pre‑formed vent pins on the pattern. Misalignment between the drill coordinates and the actual pin location created a “step” where loose sand accumulated. Despite blowing the vents with compressed air, many particles remained and fell into the cavity during mold closing, causing sand inclusions.
We implemented a daily coordinate verification procedure for the drilling robot. After any power failure, a master gauge was run to check alignment. The acceptance criteria were set as: misalignment ≤ 0.5 mm. When deviations exceeded this, the drill offsets were recalibrated.
The second issue was “orange peel” roughness on the vertical faces of bosses. Investigation showed that the air blow nozzle was too long and too close to the sand surface. When operators tilted the nozzle, the high‑pressure air (6 bar) scoured the sand, leaving a rough surface. We solved this by limiting the drill stroke so that the overlap distance $A$ (distance between drill tip and vent pin top) was less than 30 mm, and by shortening the blow tube so that its insertion depth $B$ (distance from tube tip to pattern) was greater than 40 mm. The geometry is summarized below.
| Parameter | Original (mm) | Improved (mm) |
|---|---|---|
| Drill‑to‑pin overlap $A$ | 50 | 25 |
| Nozzle insertion depth $B$ | 20 | 45 |
| Frequency of orange‑peel defects | 2.0% | 0.2% |
Sticker Sand on Tappet Cores
One of the most stubborn sand casting defects was sticker sand inside the tappet bores – holes only 20 mm in diameter, completely surrounded by molten iron. With conventional silica sand, the high thermal load caused sintering and chemical bonding. The cleaning operation required manual grinding with small rotary files, which took 8 minutes per block and drastically reduced throughput.
We adopted a dual‑material core strategy. The tappet core was made from a blend of chromite sand and synthetic mullite (calcined flint clay). Chromite has a thermal conductivity about three times that of silica, and a higher melting point (2030°C vs. 1720°C). The sintering temperature threshold was raised by:
$$ \Delta T_{sinter} \approx 220^\circ\text{C} $$
However, the core cost increased by 40%. To further protect the surface, we implemented a two‑stage coating process. First, the tappet core was dipped in a high‑refractoriness zircon‑based slurry (Baumé 45) and dried at 150°C for 30 minutes. Then it was assembled into the main core package and the entire assembly was coated with a standard graphite wash (Baumé 30). The dual coating gave a total coating thickness of 0.4 mm, compared to 0.15 mm for a single coat. The results are shown below.
| Coating Method | Coating Thickness (mm) | Sticker Sand % | Reject Rate due to Sticker |
|---|---|---|---|
| Single wash (graphite) | 0.15 | 5.8 | 1.8% |
| Double coat (zircon + graphite) | 0.40 | 0.2 | 0.1% |
Internal Sand Inclusion from Core Flash
In our cylinder block, the core package consists of six cores assembled by a robot. Each core has parting lines; the resulting flash, if not removed, breaks off during metal filling and becomes sand inclusions. A typical location was the cylinder bore core, which was made in a four‑part corebox. The flash at the core print could reach 0.5 mm. Manual cleaning was inconsistent.
We designed a unique deburring ring mounted on the robot gripper. When the robot picks the core, the ring slides over the core print and cuts the flash cleanly. The ring material is polyurethane with an embedded steel blade; the cutting force is about 50 N. This automatic operation reduced the inclusion rate from 1.5% to 0.3% and eliminated manual rework. The design principle follows the simple shear stress equation:
$$ \tau = \frac{F}{A_{flash}} $$
where $F$ is the cutting force and $A_{flash}$ is the cross‑sectional area of the flash. By maintaining a sharp blade and a consistent pick‑up path, the flash is sheared at a controlled location.
Summary of Overall Process Improvements
The comprehensive attack on sand casting defects yielded a dramatic reduction in internal scrap, from 15% to below 3%. The majority of defects were eliminated through targeted modifications that addressed root causes rather than symptoms. The table below gives a final comparison of defect rates before and after the improvement campaign.
| Time Period | Internal Scrap (%) | Rework/Repair (%) | Productivity (cores/shift) |
|---|---|---|---|
| Start of production | 15.0 | 50 | 100 |
| After 6 months | 7.0 | 25 | 180 |
| After 12 months | 4.5 | 10 | 220 |
| Current (stable) | 3.0 | 2 | 280 |
Conclusion
By systematically addressing the root causes of each sand casting defect – damage, sand wash, loose sand, sticker sand, and inclusions – we were able to transform the production of a complex cylinder block casting. Key lessons include:
- Robust robotic handling requires reinforcement of fragile features and clear work standards.
- Vent pin geometry can be optimized to avoid tearing during cleaning.
- Gating system balance is critical for controlling erosion; adding ingates at under‑fed locations reduces flow velocity significantly.
- Mold hardness and sand properties must be tightly controlled; target values can be derived from simple linear models.
- Multi‑layer coating and high‑refractoriness sands are effective for small, thermally isolated cores.
- Automatic deburring integrated into the robot pick‑up sequence eliminates flash‑related sand inclusions.
These solutions have been validated over two years of high‑volume production (1000 blocks per day) and are now applied to new cylinder block developments. The systematic approach to sand casting defects described here provides a practical roadmap for similar automated foundries.