The project also looked at potential applications of AI as an improvement tool.
Eight foundries participated.
The Metal Casting Technology Station at the University of Johannesburg (UJ-MCTS) recently completed a statistical analysis of ductile iron quality and consistency in eight South African foundries. The project also looked at the potential applications of AI as an improvement tool.
“Downstream machine shops and original equipment manufacturers (OEMs) raised legitimate concerns regarding the consistency of locally produced ductile iron castings, particularly around machinability and mechanical properties. These inconsistencies have measurable business consequences, contributing to reduced productivity and, in some cases, a preference for imported castings over local alternatives, placing additional pressure on an already strained foundry sector and hastening foundry closures,” explained Daniel Sekotlong, A Technical Signatory at UJ-MCTS.
“Ductile iron castings represent approximately 12% of South Africa’s total foundry output, corresponding to an estimated 47 000 tons annually across 170 foundries. This equates to an average production of just over 270 tons per foundry per year, with an estimated annual value exceeding R14 million per facility. These castings are primarily supplied to the mining, automotive, and agricultural sectors, where reliable and consistent mechanical performance is critical.”
“Given the strategic importance of ductile iron to the South African foundry industry, this study provides the sector’s first comprehensive, multi-foundry, data-driven assessment of production consistency.”
Picture for illustration purposes courtesy Ductile Iron Organisation. Ductile iron castings represent approximately 12% of South Africa’s total foundry output, corresponding to an estimated 47 000 tons annually across 170 foundries according to the Metal Casting Technology Station at the University of Johannesburg study
“The findings reveal a dual reality: quality and cost variability driven partly by process-related gaps at foundry level, and partly by structural, industry-wide challenges tied to South Africa’s steel scrap feedstock and energy cost environment. Beyond the aggregated industry analysis presented in our findings, each participating foundry also received a tailored set of quality improvement recommendations based on its individual results, ensuring the study delivers practical value at both sector and facility level, in line with UJ-MCTS’ mission of bridging academic research and industry application.”
“The UJ-MCTS aims to contribute to the competitiveness of the South African metal casting industry through the development and transfer of specialised knowledge and technology, fostering interaction and innovation between industry and academia. This study forms part of that broader mandate, applying rigorous, data-driven research to a pressing industry challenge.”
Eight South African foundries participated in the study, during which more than 290 ductile iron castings conforming to SABS 936 Grade (SG 42) were collected and tested.
Over 1 500 tests were conducted using a range of standardised methods, including tensile testing (ASTM E8/E8M), carbon and sulphur determination by combustion (ASTM E1019), Brinell hardness testing (ASTM E10), and computer-aided microstructure analysis (ASTM A247 and ASTM E2567). Chemical composition analysis was conducted using MCTS’s recently acquired LECO GDS900 optical emission spectrometer, which enabled rapid and precise determination of elemental concentrations including manganese (Mn), copper (Cu), sulphur (S), and magnesium (Mg) across all collected samples, in line with ASTM E415. This investment in advanced analytical capability strengthened the accuracy of the chemistry-based findings underpinning the study’s conclusions.
Computer-aided microstructure analysis enabled the conversion of subjective visual observations of microstructural features into objective, quantitative measurements. A cloud-based input/output data management system, termed the Cast-Iron Information System (CIFS), was established at each foundry to facilitate the collection and correlation of operational and laboratory data. Operational data and laboratory samples were collected during the manufacturing process at the foundries, while laboratory test data were subsequently generated from the collected samples at the MCTS testing facilities. In addition to the aggregated industry findings, individual results were compiled into foundry-specific reports, forming the basis of the tailored improvement recommendations issued to each participant.
Laboratory samples were collected on a per-batch basis using a stepped specimen geometry incorporating an ASTM E8/E8M tensile test specimen. The stepped design facilitated analysis across section thicknesses ranging from 3mm to 52mm. However, only the 25mm section, corresponding to the standard SABS 936 keel block, was considered in this study.
Findings
“South African foundries face a dual competitiveness challenge: elevated and volatile energy and input material costs at industry level, combined with process-related quality gaps at individual foundry level.”
“Substantial quality variation was observed across participating foundries, with strength performance often achieved at the expense of elongation indicating opportunities for improved melt treatment and process control.”
“Ferrite fraction levels were lower than the SG 42 target range, driven by a combination of South Africa’s available steel scrap chemistry and insufficient process correction at foundry level.”
“Manganese-related variability stems from both an industry-wide feedstock constraint and inconsistent charge calculation and scrap blending practices between and within foundries.”
“Nodularity levels were below typical SG 42 expectations, likely reflecting a combination of process variables and feedstock-related trace elements that warrants further investigation.”
Challenges and observations
“Six out of eight participating foundries rely solely on chemical analysis for material verification prior to casting, while two supplement this with metallographic evaluation limited to manual nodularity counting a process that is time-consuming and susceptible to human error. Foundries conducting metallographic analysis often struggle to justify the associated costs, meaning such assessments are frequently performed only after production, typically in response to specific customer requirements.”
“Additionally, the inherent delay between production and the availability of mechanical test results constrains manufacturers’ ability to promptly identify non-conformances and implement corrective action. Together, these observations point to real, addressable process gaps in quality assurance infrastructure gaps that exist alongside, and are compounded by, the industry’s broader steel scrap and energy cost pressures.”
Individual foundry recommendations
“In addition to the aggregated, industry-level findings presented above, each of the eight participating foundries received a confidential, facility-specific report benchmarking its performance against the SABS 936 Grade SG 42 requirements and against the broader industry dataset. These reports included targeted recommendations addressing each foundry’s specific quality gaps whether related to alloy addition and charge calculation practices, melt treatment and modularisation technique, or metallurgical testing frequency and capability. This individualised feedback loop ensures that participating foundries can act on concrete, foundry-specific improvement steps, while the sector-wide findings in this article inform broader industry and policy-level engagement on shared challenges such as steel scrap supply and energy costs.”
Proposed AI solution/s
“Given the extent of the work undertaken and the volume of data generated, UJ-MCTS further assessed the potential of available AI tools for application in foundries to support the prediction of mechanical properties using microstructural and process data.”
“Preliminary findings indicate that publicly available Large Language Models (LLMs) can utilise leading production and quality variables including casting geometry, alloy chemistry, pouring temperature, and microstructural data to predict mechanical performance with respectable accuracy and repeatability. Foundries can immediately leverage these tools to support prediction of yield strength, UTS, and elongation, with accuracy improving as more operational and metallurgical data is incorporated offering a practical, low-cost route to closing some of the process gaps identified above.”
“Where data privacy and intellectual property protection are key considerations, foundries can instead deploy open-source AI models on local infrastructure, though these require model configuration, training, and validation before delivering reliable results.”
“Each participating foundry received individualised, data-driven recommendations addressing its specific quality gaps, enabling targeted process improvements alongside the sector-wide findings presented.”
A full version of the study is available from the Metal Casting Technology Station at the University of Johannesburg. For more information TEL 011 559 6952 or email mcts@uj.ac.za.
