Client:
An industrial goods manufacturer and technology startup creating intelligent applications for Industry 4.0.
Industry:
Industry 4.0, Manufacturing
Objectives
The core challenge was to create a sophisticated algorithm that could optimize the entire production process. This required analyzing multiple variables in real-time, including temperature, humidity, and raw material type, to recommend the most efficient machine configurations.
Expectations
The project was expected to:
- Develop an algorithm to speed up machine changeovers and assist operators.
- Prove that the production process could be supported in real-time.
- Demonstrate that a cloud-based ML platform (VertexAI) was superior to on-premise infrastructure for model training and optimization.
- Deliver a solution that minimized future infrastructure maintenance costs.
Outcomes
Neontri successfully proved that VertexAI provides an agile and highly effective solution for acquiring optimal production line settings. The developed algorithm delivered a high rate of prediction accuracy and laid the groundwork for a scalable, low-maintenance production optimization system.
“Neontri delivered a quality product, meeting the client’s expectations. The team was communicative and customer-oriented, allowing them to be very understanding. Additionally, they were efficient, flexible, and proactive at problem-solving.”
Features of the ML optimization solution
The solution brings together a set of capabilities that enable real-time production optimization:
- Predictive analytics: The algorithm uses real-time data to predict optimal machine settings for operators.
- Process optimization: Directly accelerates machine changeovers, reducing downtime and increasing throughput.
- Automated MLOps: Utilized VertexAI to automate the testing, training, and optimization of machine learning models.
- Real-time data processing: Leverages BigQuery to handle large, incrementing datasets in real-time.
Cooperation
Our team worked in a highly communicative and customer-oriented manner. We were efficient, flexible, and proactive in solving the complex challenges associated with real-time industrial data analysis.
Technology
To support fast model training and real-time analytics, Neontri used the following cloud technologies:
- Machine learning platform: Google Cloud Vertex AI
- Data storage & analysis: Google Cloud Storage, Google BigQuery
Results
The project confirmed the immense value of applying machine learning to industrial challenges, delivering clear, quantifiable improvements.
- 75% rate of correct prediction of machine settings.
- Proven superiority of VertexAI for faster model training and optimization compared to on-premise infrastructure.
- Minimized costs for future infrastructure maintenance due to the cloud-native design.
- Demonstrated a powerful, agile use of cloud tools to achieve optimal production line settings faster.
