Structured Problem Solving Techniques for Manufacturing Datasets to Enhance Yield

Abstract

Structured problem-solving techniques have emerged as a critical approach for enhancing manufacturing yield and optimizing production processes. This review explores the application of structured problem-solving methodologies, such as Six Sigma, Lean Manufacturing and the Theory of Constraints, in the context of manufacturing datasets. The importance of data collection, preparation and analysis techniques, including statistical process control, regression analysis, machine learning algorithms for anomaly detection, clustering techniques and time-series analysis, is discussed. This study also highlights the role of data visualization tools in facilitating problem-solving and decision-making processes. Case studies from diverse manufacturing sectors, including semiconductor, automotive, pharmaceutical and electronics assembly, have demonstrated the successful implementation of structured problem-solving approaches and their impact on yield enhancement. This review delineates how Industry 4.0 technologies such as real-time data analysis and predictive maintenance, enhance traditional structured problem-solving techniques by enabling more proactive and data-driven decision-making processes. Challenges and limitations, including data quality issues, resistance to change and the need for cross-functional collaboration, were addressed. The paper concludes by discussing future trends in manufacturing data analysis, such as the application of artificial intelligence and edge computing and the growing importance of sustainability in manufacturing problem solving. This review underscores the significance of structured problem-solving techniques in driving continuous improvement and maintaining competitiveness in the evolving manufacturing landscape.

Keywords: Structured problem-solving (SPS), Manufacturing yield, Six sigma, Lean manufacturing, Data analysis, Industry 4.0

1. Introduction

Yield enhancement is a critical aspect of manufacturing processes with a focus on increasing the percentage of defect-free products or components produced. As manufacturing technologies become increasingly sophisticated and competitive pressure intensifies, the importance of yield enhancement cannot be overstated. Improved yields lead to reduced waste, lower production costs and increased profitability, making them a key driver of success in modern manufacturing environments.

In recent years, data analysis has emerged as a powerful tool in modern manufacturing, playing a pivotal role in yield enhancement and problem-solving². The advent of advanced sensors, Internet of Things (IoT) devices and big data technologies has enabled manufacturers to collect and analyze vast amounts of real-time data from their production processes. This wealth of information allows for more accurate identification of root causes, predictive maintenance and data-driven decision-making, ultimately leading to more effective problem solving and yield optimization strategies³.

By leveraging structured problem-solving techniques in conjunction with advanced data analysis, manufacturers can address complex challenges, optimize their processes and achieve significant improvements in yield and overall operational performance. This approach not only enhances productivity, but also contributes to the long-term sustainability and competitiveness of manufacturing organizations in an increasingly data-driven industrial landscape.

2. Overview of the Common SPS Methodologies

A.   Six sigma (DMAIC)

Six Sigma is a data-driven methodology used to improve business processes by reducing defects and variability. It employs statistical techniques and quality management principles to identify and eliminate sources of errors or inefficiencies. The goal of Six Sigma is to achieve near-perfect quality, with only 3.4 defects per million opportunities.

This data-driven systematic approach to problem solving consists of five distinct phases⁴.

·Define: Clearly identify the problem, project goals and customer requirements.

·Measure: Collect relevant data to establish a baseline for current performance.

·Analyze: Examine the data to identify root causes of problems and inefficiencies.

·Improve: Develop, test and implement solutions to address identified issues.

·Control: Monitoring the improved process to ensure sustained success and prevent regression.

B.    Lean manufacturing

Lean Manufacturing emphasizes the elimination of eight types of waste: defects, overproduction, waiting, non-utilized talent, transportation, inventory, motion and extra processing. By streamlining processes and reducing waste organizations can improve efficiency, reduce costs and enhance customer satisfaction.

This methodology focuses on maximizing customer value while minimizing waste⁵. Key principles include:

·Identifying value from the customer's perspective.

·Mapping the value stream to eliminate non-value-adding activities.

·Creating flow by organizing work processes to reduce interruptions and delays.

·Implementing pull systems to produce only what is needed when it is needed.

·Pursuing perfection through continuous improvement.

C.   Theory of constraints (TOC)

This management philosophy, developed by Dr. Eliyahu M. Goldratt, focuses on identifying and addressing the most significant limiting factor (constraint) that hinders an organization from achieving its goals. The TOC process involves the following steps.

·Identifying the system's constraint.

·Deciding how to exploit the constraint.

·Subordinating everything else to the above decision.

·Elevating the system's constraint.

·Repeating the process if a new constraint emerges.

By concentrating on the primary bottleneck organizations can achieve significant improvements in the overall system performance. This approach recognizes that the strength of any chain is determined by its weakest link and aims to strengthen that link systematically.

These methodologies, while distinct, share the common goals of improving organizational performance, reducing inefficiencies and enhancing customer value. Many organizations integrate elements from multiple approaches to create a comprehensive improvement strategy tailored to their specific needs and challenges. By implementing these methods, businesses can identify and address issues more effectively, streamline their processes, reduce costs, improve quality and achieve better overall performance and competitiveness in their respective markets.

3. Data Collection and Preparation

Data collection and preparation are crucial steps in leveraging data analytics in manufacturing processes. Various manufacturing data sources provide valuable insights into different aspects of production. Sensors embedded in machinery and equipment collect real-time data on performance, temperature, vibration and other operational parameters. Quality control systems generate data regarding product specifications, defects and inspection results. Enterprise Resource Planning (ERP) systems capture information on inventories, supply chains and production planning. Additionally, Manufacturing Execution Systems (MES) provide data on work-in-progress, production schedules and resource allocation.

The importance of data cleaning and preprocessing cannot be overstated in manufacturing analytics. Raw data often contain errors, inconsistencies and noise, which can lead to inaccurate analyses and flawed decision making. Data cleaning involves identifying, correcting or removing inaccurate, incomplete or irrelevant data⁶. Preprocessing techniques such as normalization, scaling and feature engineering help transform raw data into a format suitable for analysis⁷. These steps ensure that the data used for analytics are reliable, consistent and meaningful, ultimately improving the accuracy and effectiveness of predictive models and decision-support systems.

Dealing with the manufacturing data presents several challenges. One significant issue is the high volume and velocity of data generated by sensors and automated systems, which require a robust infrastructure for data storage and processing. Data integration from disparate sources with varying formats and structures can be complex, necessitating careful data mapping and standardization. Ensuring data quality and consistency across different production lines or facilities is another challenge because variations in equipment, processes or data collection methods can lead to discrepancies. Additionally, manufacturing environments often face issues related to data security and privacy, particularly when dealing with proprietary or customer data. Addressing these challenges requires a comprehensive data management strategy that includes robust data governance policies, advanced data integration tools and scalable data storage and processing solutions.

3. Data Analysis Techniques

Building on traditional statistical methods, such as regression analysis, modern manufacturing increasingly relies on machine learning algorithms to uncover deeper insights and detect anomalies in production data.

Regression analysis is a powerful statistical method for examining the relationships between variables. This allows researchers to model and predict outcomes using one or more independent variables⁹. This technique is extensively used in various fields, including economics, social sciences and engineering, to understand cause-and-effect relationships and to make data-driven decisions.

Machine-learning algorithms for anomaly detection have gained significant attention in recent years. These algorithms can automatically identify unusual patterns or outliers in large datasets, rendering them invaluable for fraud detection, network security and industrial equipment monitoring¹⁰. By learning from historical data, these algorithms can adapt to evolving patterns and detect subtle anomalies that may be missed by traditional methods.

Clustering techniques are essential for identifying patterns and grouping of similar data points. It is an unsupervised learning method that partitions a dataset into subsets or clusters, where objects within the same cluster are more similar to each other than to those in other clusters¹¹. This technique is particularly useful for market segmentation, customer behavior analysis and image recognition. By organizing data into meaningful clusters, researchers can uncover hidden structures and relationships within complex datasets, leading to more targeted strategies and improved decision-making¹².

Time series analysis is a specialized technique for analyzing data points collected over time. This method is crucial for forecasting trends, understanding seasonal patterns and identifying cyclical behaviors in various domains such as finance, economics and environmental sciences. Time series analysis enables researchers to make predictions based on historical data and understand the underlying factors influencing temporal patterns¹³.

4. Visualization Tools

Various chart types serve different purposes in data visualization and problem solving. Pareto charts are particularly useful for identifying the most significant factors contributing to a problem following the 80/20 principle. These charts help prioritize issues and effectively allocate resources. Fishbone diagrams, also known as Ishikawa diagrams, are valuable for root cause analysis, allowing teams to explore and categorize the potential causes of a problem systematically. Scatter plots are effective in revealing relationships between variables, helping identify correlations and potential causal links. Other chart types, such as bar charts, line graphs and pie charts, each have strengths in presenting specific types of data and relationships.

Modern interactive visualization tools have revolutionized the field of data analysis and problem-solving. These tools allow users to dynamically explore data, zoom in to specific areas of interest and manipulate variables in real time. Interactive dashboards provide a comprehensive view of key metrics and allow for drill-down capabilities, enabling users to investigate the underlying data points. Features, such as brushing and linking across multiple visualizations, enhance the ability to discover complex relationships within datasets. Additionally, modern tools often incorporate machine learning algorithms to suggest relevant visualizations or highlight significant patterns automatically. These advancements in interactive visualization empower users to engage with data more deeply, leading to more thorough problem analyses and innovative solutions.

5. Case Studies

B. Automotive industry

An automotive parts supplier struggled with high rejection rates of precision-machined components. Applying lean manufacturing principles, they utilized value stream mapping to visualize the entire production process. This reveals bottlenecks and quality issues in specific machining operations. By implementing poka-yoke (error-proofing) devices and optimizing tool change procedures, they reduced defects by 40% and increased the overall equipment effectiveness (OEE) by 25%.

C. Pharmaceutical manufacturing
 A pharmaceutical company has reported inconsistent yields in the production of a critical active ingredient. Using the design of experiments (DOE), key process parameters such as temperature, pH and reagent concentrations were systematically varied. This structured approach allowed them to optimize the reaction conditions, resulting in a 30% increase in the yield and improved batch-to-batch consistency. The DOE methodology also provides a deeper understanding of the underlying chemical processes, facilitating future process improvements.

D.   Electronics assembly

An electronics manufacturer faces quality issues with printed circuit board assemblies. By implementing failure mode and effects analysis (FMEA), they identified the critical failure points in their soldering process. This led to targeted improvements in flux application and reflow oven profiles. Combined with statistical process control (SPC) to monitor key variables, they achieved a 50% reduction in solder defects and a corresponding increase in first-pass yield.

These case studies highlight how structured problem-solving techniques, such as Six Sigma, lean manufacturing, design of experiments and failure mode analysis, can be applied across various manufacturing sectors to achieve significant yield improvements. The common thread is a systematic approach to identifying root causes and implementing data-driven solutions.

6. Integration with Industry 4.0

Real-time data analysis is a cornerstone of Industry 4.0, which significantly impacts structured problem-solving. Advanced sensors and Internet of Things (IoT) devices continuously collect vast amounts of data from manufacturing equipment and processes. These data are instantly analyzed using sophisticated algorithms, enabling the rapid detection of anomalies or deviations from optimal performance. Consequently, problems can be identified and addressed much faster than with traditional methods, often before they escalate into more significant issues that could disrupt production.

Predictive maintenance is another key area in which structured problem-solving techniques have been revolutionized by Industry 4.0¹⁴. By leveraging machine learning and artificial intelligence, manufacturers can analyze historical and real-time data to predict when the equipment is likely to fail or require maintenance. This proactive approach allows scheduled maintenance interventions that minimize downtime and optimize resource allocation. Structured problem solving in this context shifts from reactive troubleshooting to preemptive action, reducing the frequency and severity of equipment-related issues.

The integration of smart manufacturing principles also enhances root cause analysis and continuous improvement. With access to comprehensive data and advanced analytics tools, problem-solvers can easily identify patterns, correlations and underlying factors contributing to issues. This data-driven approach enables more accurate and nuanced problem definitions, thereby leading to more effective solutions. Furthermore, the iterative nature of Industry 4.0, which allows for continuous monitoring and refinement of implemented solutions, fosters a culture of ongoing improvement and adaptation.

As structured problem-solving techniques evolve within the industry 4.0 framework, they are becoming more collaborative and interconnected. Cloud-based platforms and digital twins enable cross-functional teams to work together seamlessly, share insights and collaborate on solutions, regardless of physical location. This interconnectedness extends to the supply chain, allowing for more holistic problem-solving approaches that consider the entire value stream, rather than isolated processes.

7. Challenges and Limitations

Data quality issues significantly affect the effectiveness of problem-solving efforts. Incomplete, inaccurate or outdated data can lead to flawed analysis and misguided conclusions. Organizations may struggle with data silos where critical information is scattered across different departments or systems, making it challenging to gain a comprehensive view of the problem. Furthermore, the sheer volume of data available can be overwhelming, requiring advanced analytical skills to extract meaningful insights.

Cross-functional collaboration is essential for successful problem-solving; however, it presents its own set of challenges. Different departments may have conflicting priorities or perspectives, making it difficult to align problem definitions and potential solutions. Communication barriers, including technical jargon or differing expertise levels, can impede effective information sharing. Moreover organizational structures that promote siloed thinking may discourage the cross-pollination of ideas necessary for innovative problem-solving.

To overcome these limitations organizations must invest in training programs to develop problem-solving skills at all levels. The implementation of robust data governance practices can improve data quality and accessibility. Finally, fostering a culture of collaboration and establishing clear processes for cross-functional teamwork can break down silos and enhance problem solving capabilities. By addressing these challenges systematically organizations can create an environment conducive to effective structured problem solving, leading to better decision-making and improved outcomes.

8. Future Trends

Advanced analytics in yield optimization hold immense potential for manufacturers seeking to maximize their output while minimizing waste. By leveraging big data and sophisticated statistical models, companies can gain deeper insight into their production processes and identify bottlenecks and inefficiencies. Predictive analytics can forecast yield rates based on various factors, allowing manufacturers to adjust their processes proactively for optimal results. Moreover, prescriptive analytics can suggest specific actions to improve yield, considering multiple variables such as raw material quality, equipment performance and environmental conditions.

Thus, the growing importance of sustainability in manufacturing cannot be overstated. As environmental concerns become increasingly pressing, manufacturers are under pressure to reduce their carbon footprints and adopt more eco-friendly practices. Data analysis plays a crucial role in this shift towards sustainability, enabling companies to monitor and optimize their energy consumption, reduce waste and improve resource efficiency. Advanced analytics can help identify opportunities for implementing circular economy principles such as recycling and reusing materials within the production process. Data-driven insights can support the development of more sustainable products by analyzing lifecycle assessments and identifying areas for improvement in product design and manufacturing processes.

9. Conclusion

The integration of advanced data analytics, artificial intelligence and Industry 4.0 concepts with traditional problem-solving techniques offers even greater potential for yield optimization. Real-time data analysis, predictive maintenance and digital twins have revolutionized problem-solving strategies, enabling more proactive and efficient approaches.

Successful implementation requires overcoming challenges, such as data quality issues, resistance to change and the need for cross-functional collaboration. Organizations must invest in training and robust data management practices and foster a culture of continuous improvement.

The future of manufacturing problem-solving is likely to be shaped by emerging technologies and a growing emphasis on sustainability. Manufacturers must adapt their approaches to address yield, efficiency, environmental impact and resource conservation.

In conclusion, structured problem-solving techniques remain essential for manufacturers seeking to enhance their yield and maintain competitiveness in an increasingly complex and data-driven landscape. By embracing these methodologies and adapting them to incorporate new technologies and sustainability considerations, manufacturers can achieve long-term success.

10. References

1.  Choo S, Nag R and Xia Y. “The role of executive problem solving in knowledge accumulation and manufacturing improvements,” Journal of Operations Management, 2015;36: 63-74.
2. Khakifirooz M, Chien CF and Chen YJ. “Bayesian inference for mining semiconductor manufacturing big data for yield enhancement and smart production to empower industry 4.0,” Applied Soft Computing, 2017;68: 990-999.
3. Cho M, Song M, Lee J, Park G, Kum E and Lee B. “Discovery of Resource-Oriented Transition Systems for Yield Enhancement in Semiconductor Manufacturing,” IEEE Transactions on Semiconductor Manufacturing, 2021;34: 17-24.
4. Girmanová L, Mikloš V, Divoková A, Kliment J and Šolc M. “Application of Six Sigma Using DMAIC Methodology in the Process of Product Quality Control in Metallurgical Operation,” Acta Technologica Agriculturae, 20: 104-109.
5. Singh J, Singh G and Singh H. “Productivity improvement using lean manufacturing in manufacturing industry of Northern India,” International Journal of Productivity and Performance Management, 67: 1394-1415.
6. Xu S, et al. “Data cleaning in the process industries,” Reviews in Chemical Engineering, 2015;31.
7. Rahul K and Banyal RK. “Detection and Correction of Abnormal Data with Optimized Dirty Data: A New Data Cleaning Model,” International Journal of Information Technology and Decision Making, 2021;20: 809-841.
8. De Vries and Reneau JK. “Application of statistical process control charts to monitor changes in animal production systems1,” Journal of Animal Science, 2010;88: E11-E24.
9. Marill KA. “Advanced statistics: linear regression, part II: multiple linear regression.,” Academic Emergency Medicine, 2004;11: 94-102.
10. Chien CF and Chuang SC. “A Framework for Root Cause Detection of Sub-Batch Processing System for Semiconductor Manufacturing Big Data Analytics,” IEEE Transactions on Semiconductor Manufacturing, 2014;27: 475-488.
11. Irani J, Pise N and Phatak M. “Clustering Techniques and the Similarity Measures used in Clustering: A Survey,” International Journal of Computer Applications, 2016;134: 9-14.
12. Madhulatha TS. “Comparison between K-Means and K-Medoids Clustering Algorithms,” springer berlin heidelberg, 2011: 472-481.
13. Misra P and Siddharth S. “Machine learning and time series: Real world applications,” 2017;3176: 389-394.
14. Zhu L, Chen J and Chen CI. “Prognostics of tool failing behavior based on autoassociative Gaussian process regression for semiconductor manufacturing,” 2020: 316-321.
15. Zhang L and Zhang L. “Artificial Intelligence for Remote Sensing Data Analysis: A review of challenges and opportunities,” IEEE Geoscience and Remote Sensing Magazine, 2022;10: 270-294.
16. Babou SM, Kashihara S, Niang I, Fall D and Kadobayashi Y. “Home Edge Computing (HEC): Design of a New Edge Computing Technology for Achieving Ultra-Low Latency,” springer, 2018: 3-17.