The modern approach in Industry 4.0 centers on machines operating autonomously using real-time data collected from production processes. In concrete pavement and building blocks manufacturing, the traditional human-based approach still dominates, where operators react to production events based primarily on their senses and experience. At Quatromatic, our mission is to create data-driven factories in the building materials industry.
To develop this approach, we observed production line operators at work and documented their processes in detail. The volume of digital data collected must fully match what a human operator perceives. With this data, we can train AI models to process information just as an operator would. We have developed a step-by-step methodology to achieve this challenging goal using a comprehensive AI approach.
Measuring Finished Products: Collecting Production Process Outcome Data
We've chosen to collect data starting from the end of the process and working backward to its beginning because this approach allows each step to produce valuable output. Our first product, QuatroPBC, is a computer vision system that observes and measures all valuable characteristics of the finished product: dimensions, defects, color stability, density, etc. QuatroPBC enables real-time evaluation of production characteristics, effectively automating the visual inspection function that operators would traditionally perform while monitoring the production conveyor.
Monitoring Vibration Parameters and Machine Settings
Understanding product characteristics tells us whether everything is good or not, but this alone is insufficient to identify the causes of problems. Therefore, we need to collect more data about factors that impact the production process: press machine settings and vibration conditions. In traditional human-based factories, only press machine settings are available online. For vibration conditions, which are affected by various factors (like technical conditions, materials, etc.), operators make estimations based on experience, often just guessing. This approach is unacceptable in a data-driven factory. To progress, collecting this data is essential. At Quatromatic, we have launched a project to create a "vibration analysis sensor," which will add real vibration data and the press machine settings that cause them to the factory database.
First AI Factory Model: Press Machine Control
Using data on final product measurements, vibration conditions, and settings, we will create the first Intelligent Digital Assistant for operators (project QuatroIDA). The amount of data available at this stage makes this goal achievable. However, it's impossible to train an AI model to optimize all settings for all conditions simultaneously. Instead, we'll train the system to respond to specific industrial situations and gradually expand its capabilities.
From the operator's perspective, the process will work as follows: When the system detects a specific vibration pattern, it will generate recommendations to adjust certain settings. Over time, the number of recognized patterns will increase. Once the operator's role consists mainly of pressing the "confirm" button, we'll be ready to establish direct connections to factory settings and implement automatic adjustments.
Collecting Data from the Concrete Mixing Process
All previous steps were based on one assumption: that the mix design is good and the concrete entering the press machine is adequate for further processing. However, in a data-driven factory, we cannot rely on this assumption. We cannot depend on operators physically touching concrete to control moisture or assume aggregate granulometry remains stable. Therefore, we must develop methods to collect mix data directly without assumptions. We have initiated a project for direct moisture measurement in aggregates and granulometry calculation during each mixing cycle in the concrete mixer, as this is the only way to understand the complete set of production data (project QuatroMAC).
Second AI Factory Model: Concrete Mixing Control
At this stage, we will have compiled data about all essential characteristics: the concrete mix in the press machine, production settings, real vibration conditions, and finished product measurements. This comprehensive dataset allows us to track the entire process and make decisions based on data rather than experience. In the previous step, we trained our first model to manage the pressing process based on vibration analysis. Now, we will train an AI model to manage the concrete mixing process to prepare the optimal mix for pressing, utilizing all the collected data.
Conclusion
To transform a factory into a data-driven operation that runs autonomously, we need to collect data from three key sources:
- Measurements of finished products
- Vibration settings and actual vibration parameters
- Concrete mix recipes and real mix parameters based on aggregate measurements
Each production step requires its own AI model. For a typical concrete blocks production line, we need to train two distinct models:
- Press machine control model
- Concrete mixing control model
Quatromatic is implementing these five projects separately as valuable standalone solutions, which will ultimately combine into one comprehensive system that enables factories to operate based on data. You can begin this journey today by installing QuatroPBC at your facility. As we release additional products, your factory database will become increasingly data-rich, gradually increasing your factory's data-driven capabilities up to 100%.
