This study generates an Excel-based parametric cradle-to-gate Life Cycle Assessment (LCA) tool for ductile iron casting to help the industry reduce its environmental impacts and help product designers make informed material selection decisions. The model is developed based on input from the literature, physics-based extrapolations, and recent material and energy data from 11 U.S. foundries representing 26% of the U.S. non-pipe ductile iron industry. In Part 1 of this study, the approach for developing the Excel-based parametric LCA model and how to use the model are presented.  In Part 2, a comparative cradle-to-grave LCA of ductile iron versus other materials for automotive applications is conducted. Case study analyses reveal that ductile iron parts with material utilizations above 50% likely have lower life cycle impacts than steel or aluminum equivalents.

The research was funded in part by a grant from AFS and the Ductile Iron Society. 

The Discovery

In all three case studies, the austempered ductile iron (ADI) casting was designed to replace products intended for other materials and the ductile iron product mass is lower than the other material products. 

The lower product mass of ADI components leads to lower use phase cumulative energy demand and greenhouse gas impacts than that of the other material products. The lower product mass and the lower impacts per kg lead to lower production impacts of ADI. However, if an ADI product is produced with a lower casting yield than 50% or higher product weight than cast aluminum, stamped steel, or forged steel product, the LCA comparison results may be reversed. Table 1 shows the cradle-to-gate and life cycle cumulative energy demand breakeven analysis results for the three case studies. If the mass of the ductile iron lower control arm in case study 2 increases from 14.1 kg to more than 18.5 kg as shown in Table 1, it may have higher manufacturing cumulative energy demand than the stamped steel equivalent. 

It should be noted that this breakeven analysis should be viewed as a reference not assertion for comparing ADI products with products made of other materials. This is because the breakeven analysis changes the product mass or casting yield assumptions, which implies changes in the underlining product and mold design of the case studies. The changes cannot be predicted without iterative simulation, experiments, and prototyping process, and will lead to high uncertainties in the cradle-to-gate or life cycle impact estimation (the case study results are most reliable with the default mass and yield assumptions). In addition, innovations in the material production process for other materials could also change the comparison results. 

The case study suggests that when making a material selection, the automotive designer should consider not only the embodied impact of materials but also the product design (e.g., mass) and manufacturing process efficiency.

Research Focus

The design freedom of 3D printing allows for new mold designs not possible with traditional approaches, such as helical sprues and spatially-varying lattice castings. However, research on the curing time of printed molds, including the aging, requires more exploration. This study describes the experiments of 3D-printed specimens in which embedded environmental sensors were fully encapsulated into sand blocks during an interruption of the binder jetting process. Subsequently, over a 28-day duration, humidity, volatile organic compound generation, temperature, and barometric pressure were captured for three environmental treatments. Mechanical testing of standard test specimens subjected to the same conditions was conducted. The sand structures held in high (uncontrolled) humidity and at reduced temperature were statistically weaker than a third treatment based on the hypothesis that high humidity and/or low temperatures impede curing. 

This project was funded through the Murchison Endowment at the University of Texas at El Paso and a grant from AFS, project number 19-20#12.

The Discovery

The project demonstrated the basic utility of monitoring environmental conditions of printed sand molds and cores to ensure that binder was sufficiently cured. This technique will be further explored to establish guidelines for a window in time for which the molds could be used for casting depending on humidity and temperature. The important conclusion points include:

• Sensors can be embedded successfully within molds during a printer interruption of sand molds to inform the extent of curing.

• For the specific BME680 sensor, the Volatile Organic Compound sensor mode would require calibration to be used in determining if sand binder had sufficiently cured to provide the mold strength required for metal pouring.

• A combined metric of temperature and humidity has been preliminarily shown to provide a higher correlation between the calculated metric and the final mold flexural strength.

• With further experiments, guidelines could be established to allow for the monitoring of temperature and humidity in a mold storage facility to calculate a minimum curing time for any specific conditions. 

• Long-duration experiments will be the focus of future work to understand how long 3D-printed sand molds remain sufficiently strong, and if these molds need to have an expiration date with the specific consideration of the environmental conditions that the molds were subjected to in the storage facility over time.

Research Focus

Machine learning (ML) is unlocking patterns and insight into data to provide financial value and knowledge for organizations. Use of ML in manufacturing environments is increasing, yet sometimes these applications fail to produce meaningful results. A critical review of how defects are classified is needed to appropriately apply machine learning in a production foundry and other manufacturing processes. Four elements associated with defect classification are proposed: Binary Acceptance Specifications, Stochastic Formation of Defects, Secondary Process Variation, and Visual Defect Inspection. These four elements create data space overlap, which influences the bias associated with training supervised machine learning algorithms. If this influence is significant enough, the predicted error of the model exceeds a critical error threshold (CET). 

There is no financial motivation to implement the machine learning model in the manufacturing environment if its error is greater than the CET. The goal is to bring awareness to these four elements, define the critical error threshold, and offer guidance and future study recommendations on data collection and machine learning that will increase the success of ML within manufacturing.
This research was originally published in the AFS International Journal of Metalcasting.

The Discovery

Supervised machine learning is a powerful tool that has been successfully applied in many industries. There have been substantial advances in image classification and natural language processing, for example, but comparing these to applications of supervised ML in manufacturing is like comparing apples and oranges. Both use ML, yet they are distinctly different and must be treated accordingly.

The challenges of supervised machine learning in manufacturing are significant due to classification issues and limitations in data collection. The four elements proposed (Binary Acceptance Specifications, Stochastic Formation of Defects, Secondary Process Variation, and Visual Defect Inspection) influence the final classification of a part. Misclassification creates data space overlap. This overlap alters the bias in the training of supervised machine learning, possibly rendering the model financially useless in a production environment. Understanding the critical error threshold provides economic guidance on when ML can be successfully applied.

Until manufacturing can establish system-wide gathering of process variables and eliminate classification issues, the success of supervised ML will be limited to highly controlled research or academic experiments. Much noise exists in the system today. This does not mean ML is to be abandoned, but instead different approaches are needed for manufacturing to see the benefit.

Beyond traditional uses of supervised ML, feature importance and unsupervised ML provide entry points for manufacturers. Feature importance is the process of utilizing machine learning to identify which variables have the most important influence on the prediction. Feature importance can provide an advantage even by identifying the few process parameters to investigate with a design of experiment (DOE). 

The potential time savings and guidance on critical input parameters feature importance can provide needs to be better understood and utilized within manufacturing. This could be a noteworthy savings in experimentation and the optimization process for diecasters. 

Unsupervised ML is learned on data inputs without any knowledge of the results. It typically focuses on clustering and anomaly detection algorithms. Using unsupervised ML for process control and anomaly detection allows for the use of machine learning in manufacturing, while creating the foundation needed for future supervised ML. This foundation is created when a company improves its classification of parts produced (reducing the bias and overlap) and optimizes the data that could improve the prediction model (reducing the inherent error). In the end, these changes will position manufacturing to benefit from accurate predictions of supervised machine learning while obtaining an improved understanding of the process.

Click here to view the column in the May 2023 Modern Casting digital edition.