In my years of work within a large-scale sand casting foundry, I have dedicated substantial effort to studying the dimensional accuracy of sand-moulded castings. The motivation for this research stems from the critical need to improve precision in sand casting foundry operations. Dimensional accuracy in sand casting foundry products directly affects machining costs, assembly efficiency, product weight, and overall competitiveness. In this paper, I present my findings on the evaluation methods, influencing factors, and predictive approaches for dimensional accuracy in a sand casting foundry. The production practice we conducted demonstrates that the predictive data for castings align closely with actual measurements, validating the utility of the proposed models.
1. The Importance of Dimensional Accuracy in Castings
Dimensional accuracy plays a pivotal role in the economic and technical success of any sand casting foundry. From my experience, I have observed several key benefits that justify the intense focus on improving precision:
- Reduction of machining costs: When castings are dimensionally accurate, they can be directly located on machine tool fixtures through positioning points, eliminating several machining operations and reducing the need for expensive machining allowances on difficult-to-machine alloys.
- Lower assembly costs: Precise castings reduce manual assembly time. For instance, assemblies previously composed of several parts can be redesigned as a single casting due to improved accuracy, simplifying the supply chain and reducing inventory needs.
- Weight reduction: Tighter tolerances lead to less excess material on castings. With consistent dimensional control, safety factors can be reduced, enabling thinner wall sections without compromising structural integrity.
- Reduced casting production costs: Improved sand casting foundry techniques stabilize the production of thin-walled castings, significantly reducing weight and thereby lowering material costs, especially for precious metal alloys.
- Enhanced appearance: Many castings, whether painted or unpainted, benefit from improved dimensional consistency, often resulting in a better surface finish and overall appearance, which increases market competitiveness.
- Compact design: Designers prefer high-precision castings over assembled components, leading to more compact machinery and smaller equipment footprints.
2. Evaluation Methods for Dimensional Accuracy of Castings
Internationally, statistical methods are commonly used to evaluate the dimensional accuracy of castings produced in a sand casting foundry. The procedure involves measuring a specific dimension on a batch of castings (typically more than 25 pieces), then calculating the average value and the distribution around that average. The average may not coincide with the nominal drawing dimension. The deviation between the average and the nominal dimension is called the systematic deviation. This deviation is primarily caused by dimensional errors in patterns, core boxes, core assemblies, and core-setting fixtures. Systematic deviations are generally not included in tolerance standards because they can be corrected during production by adjusting the pattern or tooling.
The formula for the average dimension is:
$$ \bar{x} = \frac{\sum_{i=1}^{n} x_i}{n} $$
where \( \bar{x} \) is the average dimension, \( x_i \) is the \(i\)-th measured dimension, and \( n \) is the number of measured castings.
The distribution around the average is characterized by the standard deviation, which indicates the dimensional precision capability of the sand casting foundry. A smaller standard deviation reflects a higher level of process control. The standard deviation is influenced by molding, core making, core assembly, core setting, mould closing, pouring, melting, and solidification processes. These factors are inherently random in nature.
The calculation steps for the sample standard deviation are:
- Calculate the arithmetic mean of the measurement results.
- Calculate the difference between each measurement and the mean.
- Square each difference to eliminate negative signs.
- Sum all squared differences.
- Divide the sum of squared differences by the number of measurements minus one to obtain the sample variance.
- Take the square root of the sample variance to obtain the sample standard deviation.
The formula is:
$$ S.D. = \sqrt{\frac{\sum_{i=1}^{n} (x_i – \bar{x})^2}{n-1}} $$
where \( x_i \) is the \(i\)-th measured dimension, \( \bar{x} \) is the average dimension, and \( n \) is the number of measured castings.
Since many factors affect dimensional accuracy in a sand casting foundry, and both theory and practice show that in mass production, a given dimension tends to follow a normal distribution around its average, most countries use plus or minus three standard deviations as the criterion for evaluating casting dimensional accuracy. Thus, the tolerance range is defined as:
$$ L \pm 3S.D. $$
where \( L \) is the nominal dimension and \( S.D. \) is the sample standard deviation. When the sample size is sufficiently large (greater than 50), using this tolerance range provides a confidence level of 99.73%. That is, we can predict that 99.73% of the casting dimensions will fall within \( L \pm 3S.D. \).
3. Factors Influencing Dimensional Accuracy in Sand Casting Foundry
The factors affecting dimensional accuracy are numerous and can be classified in different ways. Below I present a comprehensive classification based on specific process characteristics.
3.1 Classification by Specific Process Steps
3.1.1 Casting Design and Pattern
- Excessive pattern wear
- Insufficient sand around pattern (deep recesses)
- Improper core/core print clearances
- Machining locating surfaces positioned in difficult-to-mould areas
- Excessively deep wet sand when repairing cores
- Core prints too small to support the core
- Poor pattern design causing excessive fettling and grinding
- Using cores for small holes (less than 20 mm) unnecessarily
- Improper opening at core-setting locations
- Sharp corners and small fillets that impede solidification
- Inappropriate size of chaplets
- Deformation of patterns during handling
3.1.2 Gating and Risering Practice
- Excessive fettling and grinding of gates and risers
- Hot spots, stresses, and mould distortion caused by gating systems
- Unstable shrinkage due to riser placement
3.1.3 Moulding Boxes and Tooling Equipment
- Excessive mould vibration
- Improper alignment of cope and drag boxes
- Core lifting during pouring
- Worn pins and bushings on moulding boxes
- Box sinking (deformation under load)
- Improper machine core setting
- Inadequate venting in deep pockets of the mould
- Improper cooling of the mould
- Rough handling of castings during shakeout
3.1.4 Raw Materials
- Improper sand grain size distribution
- Insufficient quality or quantity of bentonite to compensate for silica sand expansion
- Improper clay-to-water ratio
- Insufficient coal dust content
- Insufficient new sand addition
- Insufficient dextrin or other additives
3.1.5 Moulding Machines
- Insufficient squeeze pressure to achieve uniform hardness of 90 or above
- Uneven sand filling causing large pressure differences across the mould surface
- Non-parallelism between moulding boxes, pattern plates, and support seats
- Sand bridges formed in deep or narrow recesses
- Insufficient moulding box stiffness
- Excessive vibration causing collapse
- Too much parting agent sprayed
- Sand not blown clean between boxes or between box and pattern plate
- Insufficient squeeze pressure to distribute sand along the pattern contour
3.1.6 Sand Preparation and Mixing
- Insufficient mulling time to meet quantity and quality requirements
- Sand system too small or poorly designed
- High sand temperature due to high sand-to-iron ratio, unable to compensate for sintering of clay and combustible materials
- Poor sand permeability
- Non-uniform return sand
3.1.7 Cores
- Core making process defects causing dimensional instability
- Worn core boxes
- Expansion or collapsibility of cores due to heat from molten metal
- Erosion by molten metal
- Change in core wall thickness due to metal erosion
- Core damage during handling
3.1.8 Metal Composition, Temperature, and Pouring
- Variations in metal composition causing changes in shrinkage and expansion
- Elemental variations affecting the microstructure of the metal
- Instability of pouring temperature
3.2 Classification by Systematic and Random Errors
Systematic errors can be divided into permanent and semi-permanent types. Permanent systematic errors are caused by incorrect pattern, core box, core assembly, and core-setting fixture dimensions. These errors remain constant until the tooling is replaced. Semi-permanent errors are caused by long-term changes in production equipment, moulding materials, melting characteristics, and so on. These systematic errors are identical for a given batch of castings.
Semi-permanent errors can often be corrected by adjusting pattern dimensions, but this corrective action is rarely taken in practice. When a pattern plate carries several identical patterns, special problems arise because each pattern may have different dimensional variations, leading to different systematic errors. In evaluating casting measurement results, it is impossible to completely distinguish between systematic and random errors.
Because systematic errors are both permanent and semi-permanent, producing a small number of castings from the same pattern does not pose any problem. However, over long production runs, semi-permanent systematic errors must be considered together with random errors. Indeed, the distinction between systematic and random errors is critical: pattern-induced errors are permanent systematic errors, while production process factors produce random errors, and in most cases also include semi-permanent errors.
Systematic and random errors in castings have no intrinsic relationship. My measurements have shown that systematic errors have little correlation with nominal dimensions. There is no universal rule for the magnitude of systematic errors, making it impossible to incorporate them into general dimensional accuracy data for a sand casting foundry.
4. Prediction of Dimensional Accuracy in Sand Casting Foundry
To predict the dimensional accuracy of castings before production, several methods have been developed. I will describe two approaches that I have applied in our sand casting foundry: the ST-71 method from the Casting Dimensional Tolerance Technical Committee, and the dimensional chain method.
4.1 The ST-71 Technical Committee Method
This method provides a formula for estimating the standard deviation of a given casting dimension based on various parameters. The formula is:
$$ S.D. = 10^{-2} \left[ 860 \times \text{drawing dimension} + 1.969 \times \text{core projected area} + 50.2 \times \text{main wall thickness} \right] – 10^{-2} \left[ 21.84 \left( \text{if dimension formed by half of the mould} \right) + 1.27 \left( \text{if dimension formed by mould and core head fixed core} \right) + 26.67 \left( \text{if dimension formed by mould and fixture-fixed core} \right) + 7.37 \left( \text{dimension formed by two cores} \right) – 15.75 \left( \text{gray iron} \right) – 22.1 \left( \text{white iron} \right) – 32.51 \left( \text{malleable iron} \right) – 69.6 \left( \text{cast steel} \right) – 32.26 \left( \text{aluminum} \right) \right] \ \text{mm} $$
In this formula, the coefficients for steel and aluminum are based on less data, so their accuracy is lower. The core projected area is defined as the projection of the core on the parting line minus the projection of the core print area outside the cavity. The main wall thickness is the wall thickness that covers the majority of the casting. The formula clearly indicates that more complex castings yield lower dimensional accuracy.
The term \(-21.84 \times 10^{-2} \) mm represents the standard deviation for dimensions formed from one side of the mould to the other. When the other side is in another half of the mould, the dimension crosses the parting line, adding an additional \(+4.32 \times 10^{-2} \) mm to the standard deviation.
The coefficient \(-1.27 \times 10^{-2} \) mm applies to dimensions formed between the mould and a core located by core prints. This value is larger than the standard deviation for dimensions formed entirely within one half of the mould.
The coefficient \(-26.67 \times 10^{-2} \) mm applies to dimensions formed between the mould and a core fixed by fixtures. Using fixtures to fix cores is an effective way to improve dimensional accuracy in a sand casting foundry.
The coefficient \(-7.37 \times 10^{-2} \) mm applies to dimensions formed between two cores. This gives slightly better accuracy compared to dimensions formed between the mould and a core.
The subsequent coefficients are determined by the casting material:
| Material | Coefficient (\(\times 10^{-2}\) mm) |
|---|---|
| Gray iron | +15.75 |
| White iron | +22.1 |
| Malleable iron | +32.51 |
| Cast steel | +69.6 |
| Aluminum | +32.26 |
Note that the dimensional accuracy predicted using this method is based on the average dimension, meaning that the pattern and tooling have been properly adjusted so that the drawing dimension and average dimension coincide.
Table 1 shows a comparison between predicted standard deviations using the formula and measured standard deviations from actual production in our sand casting foundry.
Table 1 Comparison of predicted and measured standard deviations for various castings
| No. | Drawing Dimension (mm) | Alloy | Dimension Type | Predicted S.D. (mm) | Measured S.D. (mm) |
|---|---|---|---|---|---|
| 1 | 99.20 | Gray iron | Mould and core head fixed core | 0.302 | 0.3356 |
| 2 | 93.35 | Gray iron | All in one half of mould | 0.046 | 0.4318 |
| 3 | 119.76 | Gray iron | Mould and mould, through parting line | 0.3048 | 0.6858 |
| 4 | 144.45 | Gray iron | All in one half of mould | 0.3048 | 0.1524 |
| 5 | 377.42 | Gray iron | Core and core | 1.092 | 0.7112 |
| 6 | 82.17 | Gray iron | All in one core | 0.6604 | 0.4826 |
| 7 | 73.15 | Malleable iron | All in one half of mould | 0.3048 | 0.1524 |
| 8 | 60.96 | Malleable iron | All in one core | 0.2540 | 0.1778 |
| 9 | 152.4 | Malleable iron | All in one core | 0.3048 | 0.6604 |
| 10 | 189.71 | Malleable iron | Mould and core head fixed core | 0.7366 | 0.4064 |
| 11 | 60.96 | White iron | All in one core | 0.1778 | 0.1778 |
| 12 | 92.25 | White iron | All in one half of mould | 0.2032 | 0.0762 |
| 13 | 146.1 | White iron | Mould and core, through parting line | 0.635 | 0.4826 |
Note: Under controlled conditions, 95% of measurement values are expected to fall within ±2.5 standard deviations.
From the table, it is evident that if the mould and core process design is fixed, one can predict the range of dispersion of the average dimensions of each part of the casting directly from the drawing. If this dispersion is unacceptable, the following measures can be taken to improve accuracy:
- a. Change the mould design so that the dimension is entirely within one core or one half of the mould.
- b. Use fixtures to fix cores.
- c. Perform some machining on the casting.
- d. Design special processes or perform pre-assembly and adjustment of cores.
- e. With customer agreement, expand the tolerance range.
4.2 Dimensional Chain Method
In a sand casting foundry with many years of production experience, the dimensional chain method is often more accurate when deviations arise in moulding, core making, core assembly, transportation, and core-setting fixtures. The key points of this method are:
- List the process route for forming a position dimension or wall thickness dimension. Starting from the theoretical model, through a series of moulding, core making, core assembly, and core-setting operations, determine the final position and wall thickness dimension.
- Based on the factory’s historical data, list the dimensional deviations for each process step.
- Draw a dimension chain sketch according to the process route, listing the errors of each link in the chain.
- Solve the dimension chain equation, identifying increasing links, decreasing links, and zero links. The closing link is the algebraic sum of all links.
- Use the probabilistic method with an empirical formula to calculate the dimensional deviation. The formula is:
$$ T = \pm \sqrt{ \sum_{i=1}^{n} \left( H \cdot A_i’ \cdot \frac{R_i}{2} \right)^2 } + \sum_{i=1}^{n} \left( A_i’ \cdot e_i \right) $$
where:
- \( H \) is a correction factor,
- \( A_i’ \) is the conversion coefficient for link \(i\) (increasing links +1, zero links -1, decreasing links -1),
- \( R_i \) is half the tolerance of link \(i\) (in mm),
- \( e_i \) is the coordinate of the deviation center of link \(i\) relative to the nominal dimension (in mm),
- \( T \) is the tolerance of the closing link.
In our sand casting foundry, we applied this method to the cylinder block water jacket wall thickness. Using data provided by the cylinder block precision research group, the calculated wall thickness deviation for the CA10 cylinder block was \( 5.62 \pm 2.1 \) mm. This was verified by dissecting 100 cylinder blocks produced between 1980 and 1982, which gave a deviation range of \( 5.62 \pm 2.1 \) mm to \( 5.62 \pm 2.57 \) mm, consistent with the dimension chain calculation.
Based on the dimension chain analysis, the CA1091 cylinder block was redesigned. The length direction positioning was changed from the V-type groove used in the CA10 to a location based on the center of the 3rd and 4th cylinders of the water jacket core. Using the dimension chain method, we predicted the water jacket wall thickness deviation to be \( 5.62 \pm 1.6 \) mm. After dissecting cylinder blocks produced in 1986-1987, the measured deviation was \( \pm 1.6 \) mm, which matched the predicted result perfectly.
4.3 Additional Predictive Considerations
Beyond these two methods, I have learned that the predictive capability of any sand casting foundry is enhanced by continuously collecting data and refining the coefficients in the formulas. The ST-71 method provides a useful starting point, but each sand casting foundry has its own specific conditions regarding sand properties, moulding equipment, core quality, and metal handling. Therefore, I recommend that every sand casting foundry develop its own predictive models based on historical measurements. Table 2 summarizes the key parameters used in the ST-71 formula for reference.
| Parameter | Symbol | Unit | Effect on S.D. |
|---|---|---|---|
| Drawing dimension | \(D\) | mm | Increases S.D. |
| Core projected area | \(A_c\) | mm² | Increases S.D. |
| Main wall thickness | \(t\) | mm | Increases S.D. |
| Dimension formed in one half mould | – | – | Base coefficient 21.84 |
| Dimension through parting line | – | – | Additional 4.32 |
| Mould to core (core print) | – | – | 1.27 |
| Mould to fixture-fixed core | – | – | 26.67 |
| Core to core | – | – | 7.37 |
5. Practical Measures to Improve Dimensional Accuracy in a Sand Casting Foundry
Based on my research and production experience, I have compiled a list of practical measures that can be taken to improve the dimensional accuracy of sand casting foundry products:
- Regular pattern and tooling maintenance: Regularly inspect and refurbish patterns, core boxes, and fixtures to minimize permanent systematic errors.
- Optimize core design: Use core prints with adequate support and avoid small cores where possible. Consider using fixtures for critical core locations.
- Control mould hardness: Ensure uniform mould hardness by adjusting squeeze pressure and sand filling parameters on moulding machines.
- Stabilize sand properties: Maintain consistent sand grain size, clay content, moisture, and additives to reduce variability in mould expansion and collapsibility.
- Modern machine calibration: Regularly calibrate moulding machines and alignment systems to prevent sagging and mismatch.
- Implement statistical process control: Continuously monitor casting dimensions using control charts to detect shifts in the process mean early.
- Use predictive formulas: Employ the ST-71 method or customized dimensional chain models to anticipate accuracy issues before production.
- Improve metal handling: Control pouring temperature and metal composition within tight ranges to minimize shrinkage and expansion variability.
- Train operators: Ensure that all personnel in the sand casting foundry understand the importance of dimensional accuracy and follow standard operating procedures.
- Collaborate with designers: Work with product designers to create casting-friendly features that are easier to achieve dimensionally within the sand casting foundry process.
6. Conclusion
In conclusion, the dimensional accuracy of castings produced in a sand casting foundry is a complex issue influenced by numerous factors across the entire production chain. Through my research, I have demonstrated that statistical evaluation methods, particularly the calculation of standard deviation and the use of ±3 standard deviation tolerance ranges, provide a robust framework for quantifying and predicting accuracy. The ST-71 formula, although based on specific empirical data, offers a useful predictive tool, especially when combined with in-house data. The dimensional chain method, which relies on a foundry’s own historical data, has shown excellent agreement with actual measured results, as evidenced by the cylinder block example.
By understanding the sources of systematic and random errors and implementing appropriate corrective actions, a sand casting foundry can significantly improve its dimensional accuracy. This leads to reduced machining costs, lower assembly costs, lighter components, and enhanced overall competitiveness. I believe that continuous research and data collection in this field will further refine prediction models and enable even higher levels of precision in sand casting foundry operations.
My experience confirms that rigorous attention to detail and systematic application of statistical methods are essential for any sand casting foundry aiming to meet the increasing demands for high-quality, dimensionally accurate castings. The production results from our foundry, where predicted data matched measured data, provide strong validation for the approaches described in this paper.