Deepak Chowdhary 
MPM Infosoft Pvt. Ltd./India

Introduction
The green sand-casting process is one of the most economical and widely used manufacturing processes for making various metal components. The cast components manufactured by this method have widespread applications in various industries, including automobiles, agricultural farm equipment, construction, etc. Despite its extensive use across several sectors, the process itself has various inherent complexities that result in uncertainty in the outcome. The associated uncertainties arise from variations in the process parameters, return sand quality, inconsistencies in the raw material quality, weather conditions, and design complexities. These factors can result in various defects in the cast samples, ultimately leading to their rejection, which can result in financial losses for the foundry [1]. Typically, in a running foundry, the rejection varies from 6% to 8%, where the contribution of sand-related defects can be as high as 50%.
The correct identification and addressing of the cause of the generation of these defects requires a structured approach. In this regard, the Fishbone diagram, developed by Ishikawa, can be a useful tool. A thorough analysis can be obtained by classifying the possible causes into branches and cause-and-effect correlations.
In this work, a data-driven root cause analysis methodology has been developed that particularly focuses on various defects. A case of a sand inclusion defect is analysed in this study.
Methodology
This paper employs a unique data-driven method to identify the root cause analysis for the sand inclusion defect observed for a particular component. Both quantitative and qualitative data were used to develop this fishbone diagram. Apart from analyzing the quantitative data, i.e., change in prepared sand properties, other qualitative data is also presented in the fishbone diagram. The change in qualitative data is presented in the nodes: Process, Method, Machine, Material, and Others.
The step-by-step methodology for root cause analysis using a data-driven fishbone diagram is mentioned below:
1. Collect historical data on rejections/defects.
2. Identify the period of rejections when the rejection was low.
3. The parameters corresponding to a low rejection/defect period are considered reference parameters and are compared against the period when rejection is high or the period of interest.
4. The fishbone charts for the selected defect will appear with all bones and highlighted active bones where data is available.
5. The statistical analysis of parameters guided by the fishbone diagram is presented as a box plot, density plot, and correlation plot (between defects and the sand parameters).
6. Finally, consolidated results are presented to the user to guide them to focus on the parameters of interest to control the rejection.

Data Information
The study has been carried out for a component being manufactured in one of the leading foundries situated in central India. The sand-to-metal ratio for the component is 6.4. The total production of the component is around 3700 tonnes per year. A total of 12 months of data have been taken for analysis (April 2023 to March 2024). Two cases were selected for analysis, i.e., August 2023 has the highest occurrence of sand inclusion defect (3.37%) and March 2024, has the lowest occurrence of sand inclusion defect (0.45%). In this work, August 2023 is termed as the comparison period, and March 2024 is termed as the reference period.

RESULT AND DISCUSSION
The absolute change (%) of the prepared sand properties between the reference period and the comparison period is displayed. The absolute change is calculated using the mean value of the data of the sand properties between the comparison period and the reference period. It can be observed that the considerable change in the Green Compression strength (GCS) (5.65%), Temperature of sand after mix (3.76%), and Inert fines (%) (2.76%) might have resulted in the rise in the sand inclusion defect. It should be noted that the change in Moisture (%), Active clay (%), and Compactability (%) has a relatively low absolute change % that suggests its relatively little influence on the generation of defects.

By examining the individual change in the respective properties, a clearer understanding can be obtained.It can be concluded that the GCS was low during the comparison period (high defect period), whereas the temperature of sand after mix and inert fines was high.

By controlling these properties using sand additive consumption during the high rejection period, the foundry can reduce the occurrence of sand inclusion defects.

Conclusion
This study suggested that data-driven fishbone analytics can act as an effective tool to provide quick actionable strategies to reduce the sand-related defects that will help reduce the rejection of the cast sample and increase the profit margin for the foundry.