In this article, we explore the diverse range of data essential for our smart model to function effectively, detailing how each type is pivotal in shaping the solutions our technology provides. From real-time operational data to historical performance metrics, we outline how comprehensive data collection forms the backbone of our cause-and-effect modelling, ultimately paving the way for groundbreaking advancements in the glass industry.
Creating a Relational Smart Model (RSM) for Analyzing Distributed Siloed Data
The initial step in using AI to analyze distributed siloed data involves creating a Relational Smart Model (RSM) that maps the interactions between various factors. An RSM requires four fundamental elements:

• Classification and Possible causes of Product Defects: One of the most critical aspects of the model is the precise definition and classification of product defects. Leantechnique not only identifies different types of defects but also maps each type to its potential causes and suggests possible remedies. This process has been executed comprehensively, providing a detailed landscape of cause-and-effect relationships along with their corresponding ons. This framework ensures that each defect is addressed effectively, enhancing overall product quality and operational efficiency.At the highest level,  the Leantechnique classifies product defects into several primary categories. This  top- level classification serves as the foundation for a more detailed breakdown, extending down to all generic types of defects. This hierarchical structure ensures a systematic approach to identifying and addressing each specific type of defect throughout the product line.
  • Optical Defects (Distortions): Optical Defects, also known as Distortions, can arise from two main sources in the glass manufacturing process: chemical inhomogeneities and physical irregularities. Chemical inhomogeneity issues often stem from poor batching quality, improper melting, or contamination. On the other hand, physical irregularities include uneven glass thickness, which may show abrupt changes, and wavy surfaces, often referred to as tin bath irregularities. These defects impact the visual quality of the glass produced.
  •   Physical Defects: Classified by defect location in the glass—top surface, body defects(inclusions), and bottom surface defects:
    • Top Surface Defects: Primarily caused by corrosion of refractory materials, inadequate sealing, and maintenance operations, as well as contamination on rollers.  They manifest as four subclasses including drips, drops, adhesions, and dirt on the surface. The depth of these defects and the size of the surrounding halo are influenced by the temperature at which the drips, drops, or adhesions occur. Defects that are closer to the glass surface and those with smaller halos typically form in relatively colder areas. Occasionally, top surface defects may also lead to or occur alongside blisters, such as damper and tweel blisters. In the RSM model, providing an accurate description of defects is crucial for effective analysis and troubleshooting. The model guides the operator through a step-by-step process to ensure defects are described as comprehensively as possible.
    • Body or Inside Defects (Inclusions): These defects are found inside the body of the glass and primarily manifest as inclusions, including gaseous inclusions (bubbles, blisters, and seeds) and solid inclusions (stones, knots, and streaks). They typically form during the melting and forming processes. However, the root causes of these defects can be traced back to factors such as raw materials, utilities, refractory materials, or deviations in the process.
    • Bottom Surface Defects: Defects on the bottom surface, such as bubbles, seeds, distortions, and specks, primarily originate in the tin bath. These defects are influenced by factors including the tin itself and the forming process, as well as the spout, lips, and conveying rolls located in the dross box and lehrs. These components can directly or indirectly come into contact with and damage the bottom surface. Additionally, improper convection currents in the melter, working end, and at the entrance of the tin bath can also contribute to these defects.
    • Shape Defects (Deformities): Deformities are defects that frequently occur during the forming process. They can be caused by a range of factors, such as the chemical composition of the batch, inhomogeneity of the melt, inadequate thermal uniformity of the melt upon reaching the forming section, or issues inherent to the forming process itself.
  • Composition Defects and Deviations: These primarily refer to variations and flaws in the glass composition from established standards.
    • Batch Behaviour: These deviations can affect the thermal behavior of a glass batch, influencing key properties such as melting temperature, refining rate, viscosity, and the homogeneity of molten glass.
    • Forming Process: Deviations in glass composition also impact the forming process. They can alter the viscosity curve within the forming temperature range, affecting crucial thermal behaviours in this range and changing the rate of viscosity development—a property referred to as “long” or “short” in the context of glass forming.
    • Glass Properties: These defects and deviations also influence various glass properties, including density and mechanical characteristics like strength, as well as the overall thermal behaviour of the glass.
• Classification and Possible causes of of Process Incidents: In the previous article, we outlined the various concurrent processes occurring within a glass furnace and discussed how their combined effects can lead to cascading and complex incidents. These incidents not only produce defects in the products and damage equipment but also obscure the root causes of the initial deviations. The main concurrent processes were categorized into three primary types,  with detailed descriptions provided in the previous article:
• Production Line Machinery Condition: The condition and proper functioning of production line machinery are vital for effective analysis. LeanTechnique rigorously monitors an extensive range of equipment, including key batch plant machinery such as weighing scales, mixers, conveyors, magnets, and metal detectors. The system also observes structural changes in critical furnace components including side and breast walls, crowns, ports, and regenerators, with a special focus on checkers. Similarly, it tracks conditions in the tin bath, noting elements like spouts, tweel or lip tile, and dross box rolls, while closely monitoring for  tin level, contamination in  tin, and possible damage to lehr rolls, fans, dampers , and cutting system devices. Additionally, the service life and status of all sensors are meticulously integrated into the model through specialized calculating indexes and input forms. This comprehensive approach guarantees thorough monitoring and maintenance planning to prevent downtime and defects.
• Managerial and Operator Skills: Managerial and operator skills are crucial factors contributing to the development of the model. In addition to the operators’ skill levels, their ongoing training, effective supervision, task delegation, and adherence to standard procedures and instructions play key roles in enhancing the model’s development, predictability, and ability to conduct cause and effect analysis. We have also established a special procedure for evaluating and integrating these data into the RSM.

For further details, please see the second part of this article where we continue our discussion on these topics.