Analyzing the operational processes of a large manufacturing enterprise

Objective

To develop strategies, steps, or processes aimed at enhancing the efficiency of the company and its production operations. Specifically, the goals included:

• Preventing production stoppages due to equipment failures
• Optimizing the consumption of components in the production process
• Reducing transportation logistics costs
• Minimizing energy expenses by effectively allocating production capacity

Solution 

The initial phase involved the extraction and processing of data from the enterprise's IT systems. This included unloading datasets from the implemented Manufacturing Execution System (MES) and visualizing key parameters such as dates, events, and critical values of the processes conducted. 

The second (intermediate) phase focused on processing non-automated processes, utilizing data stored solely on physical media or electronic formats that were outside the MES system's scope.

To address the task, we extracted a dataset in CSV format spanning several years. The resulting file contained over 100,000 rows of records and more than 50 columns. Further data processing was conducted using Python within Jupyter Notebook, utilizing various libraries, primarily NumPy, Pandas, and SciPy. For visualization purposes, we also employed built-in extensions from the main package, specifically Matplotlib and Seaborn. 

The selection of this technology stack was based on the following reasons: 

• Ease of creation given the multitude of interdependent parameters being analyzed.
• Convenience of visualization for subsequent analysis.
• The ability to reuse data and algorithms with minimal effort.

Challenges Encountered and Solutions Implemented
In practice, the situation proved to be more complex than initially anticipated. The team faced significant challenges while working with the data, particularly: 

• Diverse terminologies describing events that led to interruptions and stoppages in the manufacturing process, i.e., what occurred.
• Varied explanations for the causes, i.e., why the stoppages happened.

To illustrate the scope of the work required, consider that there are 100 units of identical equipment, yet the descriptions of failure causes vary over several years. Remarkably, the same individuals are responsible for servicing this equipment and recording the logs. 

Furthermore, during our research and interviews with shop floor personnel, it became evident that not all incidents, occurrences, and deviations from standardized technological processes are documented in the central MES system. Some incidents are recorded in separate logs that were established long ago in compliance with GOST standards and internal production regulations. These records are maintained using different software platforms; some are in Excel, while others are entirely on paper. 

Consequently, another challenge arose—the inability to quickly consolidate all data due to a vast number of discrepancies in terminology, equipment types and categories, shop names, event descriptions, and other classifications. 

Our colleagues had to compile an extensive amount of data into a unified and organized repository, create new categories and classifications, and address duplicates both within the overall system and across individual software platforms.

Results

We accurately determined, based on real data and measurable error margins, which equipment most frequently halts production processes across each workshop and department, taking into account specific shifts and times of day. 

Data analytics revealed the extent to which stoppages of particular equipment impact production output. The insights gained prompted an investigation into the root causes of equipment failures, including issues related to installation, production processes, and excessive load. 

The visualizations produced enabled us to assess the quality of technical maintenance by considering the timing, frequency, and specific personnel involved. This analysis highlighted deficiencies in worker qualifications, excessive workloads, and the need for additional training. 

The aggregated numerical data on equipment failures allowed us to categorize the nature of breakdowns and identify growth opportunities—specifically, the subsequent digitalization of equipment through Internet of Things (IoT) sensors, leading to the development of an automated system for planning and maintaining optimal equipment performance and service intervals.

An analysis was conducted on over 1,000 production processes involving more than 50 types of equipment, totaling over 2,000 units. 

Technologies

• Data Analytics: Python Jupyter Notebook
• Libraries: NumPy, Pandas, SciPy
• Visualization: Matplotlib, Seaborn