Published on

Collecting Data from Industrial Machines: Methods and Tools

Authors
  • avatar
    Name
    Martin Szerment

Introduction

Effective management of data from industrial machines is critical for predicting failures and optimizing processes. In this article, we discuss two approaches:

  1. Lack of existing machine data, where models for predictive maintenance need to be built from scratch.
  2. Utilizing existing datasets from platforms such as Statista and Kaggle, which provide real-world industrial data.

1. No Machine Data: How to Start Collecting Data?

OMNIMES by Multiprojekt

If a factory lacks a data collection system, implementing an MES (Manufacturing Execution System) such as OMNIMES by Multiprojekt is a good solution.

What is OMNIMES?

  • OMNIMES is an advanced MES system that facilitates data collection, analysis, and visualization from industrial machines.
  • It enables monitoring of production parameters, such as cycle times, number of units produced, and machine failure statuses.
  • OMNIMES data can be used to build predictive models, such as:
    • Predicting machine failures.
    • Optimizing maintenance schedules.

Key Features of OMNIMES:

  1. Real-time Data Collection: The system gathers data directly from machines equipped with appropriate PLCs.
  2. Performance Analysis: OMNIMES calculates KPIs such as Overall Equipment Effectiveness (OEE).
  3. Integration with Existing Systems: It supports integration with ERP and SCADA.

For more details, visit: OMNIMES – Multiprojekt.

2. Using Public Data Sources

When a factory does not have its own data, publicly available datasets can be a viable option. Here are two popular sources:

a) Statista

Statista is a global data platform offering insights into various industrial sectors. For predictive maintenance, you can find:

  • Reports on machine failure rates.
  • Analyses of production efficiency across industries.
  • Statistical data on machine maintenance and downtimes.

Example Use Case: Reports from Statista can be used to build basic predictive models based on industry statistics.

b) Kaggle

Kaggle provides free datasets and tools for data analysis and machine learning. You can find:

  • Data collected from real industrial equipment.
  • Datasets on vibrations, temperature, or energy consumption.
  • Ready-to-use models and scripts for predictive analysis.

Example Use Case: Kaggle offers time-series data from real machines that can be used to build failure prediction models with algorithms like LSTM, XGBoost, or TensorFlow.

Comparison of Two Approaches

CriterionSystem Implementation (OMNIMES)Public Data (Statista/Kaggle)
CostMedium (system implementation)Low (data often free)
Adaptation to FacilityIdeal, data from your machinesLimited, generalized data
Model PrecisionVery highDepends on data quality
Implementation TimeModerate (up to one month for MES setup)Short (ready-to-use data)
Historical Data AccessImmediately available after OMNIMES setupUsually available immediately

Conclusion

  1. No Data: If no data collection system exists, consider implementing a solution like OMNIMES by Multiprojekt. It enables real-time data collection and analysis, simplifying the creation of accurate predictive models.

  2. Public Data: For quick prototyping, platforms like Statista and Kaggle offer ready-to-use industrial datasets.

The optimal approach depends on your budget, available resources, and timeline. Long-term, implementing an MES like OMNIMES ensures precise data and greater benefits for your production facility.

← Back to the blog