This case focused on the implementation of a robust data architecture, with data engineering focused on the construction of performance indicators and OEE, ensuring data quality and efficient process orchestration.

Developed architecture was designed to generate critical KPIs, such as performance, OEE, performance, quality, stops, Pareto analysis, and the state of stations with production orders and cycle times.

To ensure scalability and reliability, Apache Airflow was adopted as an orchestrator, with pipelines built in Python and based on the medallion model (bronze, silver, gold). The focus on Data Quality ensures precise and consistent indicators that support strategic decision making and data -oriented.

The challenge of building a OEE Analytics is in highly heterogeneous data from month, ERP, sensors and applications, ensuring that these data are captured in different formats (structured, semi-structured, not structured), not structured) Validated, transformed and made available almost in real time for analysis. This requires not only a flexible intake architecture (batch and streaming), but also strict quality control, accuracy and traceability throughout the pipeline, to prevent efficiency metrics from being distorted by capture errors, delays or inconsistencies between systems.

Another major challenge is to ensure that the analyzes produced have a good value: it is not enough to generate beautiful dashboards; It is necessary to ensure that data is enriched with context (such as stopping, description of stations, operational justifications) and that insights reflect not only gross numbers, but real opportunities for improvement. This demands efficient orchestration (such as Airflow), integration between Data Lake and Data Warehouse, and a well -aligned indicators design with business questions - such as identifying bottlenecks, performance deviations and shift or machine variations.

The idea to solve this is setting up a modern layer architecture that combines intake flexibility, guaranteed data quality and intelligent consumption. This starts by connecting the sources (month, ERP, sensors, saaes) to a layer of hybrid intake (batch and streaming), ensuring that both historical and real -time data can be processed. From then on, the data would be stored in a layer architecture, with Data Lake storing gross data (bronze) and Data Warehouse storing refined data (Silver and Gold) ready for analysis.

In practice, tools such as airflow would be used to orchestrate pipelines, ensuring traceability and end monitoring. Quality rules applied in the Silver layer would guarantee consistency and accuracy, validating times, counting and benchmarks. Final consumption would be delivered via interactive dashboards, BI reports and predictive models, all fed by operational context -enriched data. This allows not only to see what happened, but understand why it happened and what to do next.

• Complete and centralized visibility
  all OEE and production data is consolidated in interactive dashboards, allowing operational and management teams to accompany availability, Performance and quality in real time.

• Predictive analysis for failure reduction
  with the graphs, it is possible to identify trends and anticipate problems before they cause stops, helping to reduce, helping to reduce stops, helping to reduce, helping to reduce downtime and operational losses.

• Continuous data-oriented improvement
  Provision of equipment, line and cause allows you to find corrective actions and validate improvement. implemented, strengthening the continuous improvement cycle.

• Agile integration with different environments (cloud/on-premises)
  

Reduction of unplanned stops happens because predictive analyzes and detail the causes of failure allow the team to act preventively, decreasing the time of stopping machine.

There is an increase in operational efficiency, as real-time indicators allow fast factory floor adjustments, ensuring that lines and equipment operate near maximum performance.

Product quality improves, as the details of quality and loss justifications allow you to identify patterns and attack the root causes, raising the end product level.

Decision making gains agility, because unified dashboards and enriched data give management a consolidated view to prioritize actions and align facts based on facts.

The model offers scalability and technological flexibility, as the cluster can be implemented in cloud or on-premises, ensuring that the solution grows along with the operation without infrastructure limitations.