Adaslabs Roast Master : One Software for Unlock ERP And Process Excellence
All-in-one roasting software designed to streamline production, manage inventory, trace batches, and optimize every roast — whether you’re scaling up or refining your craft.
Batch Management
Roast Plan Profile
Auto Preeheating
By using a full automation system, the auto preheating feature will make it easier for operators to reach charge temperature in steady state conditions..
The consistency of the roasting results is one step ahead with this feature, because it will always get a consistent charge temperature in steady conditions (not when the RoR is rising/falling).
Temperature Calibration Feature
Bean and Environment temperature calibration function, which will improve the accuracy of temperature reading.
Background Profile
Displaying reference batch data and charts, in addition to making it easier for operators to control the process, on machines that already use automation modules, can be used to create auto events (commands such as air flow rate, burner power) that will automatically follow events in the selected background profile.
Machine Learing
In stand-alone use, modern Machine Learning technology has been embedded, which will make temperature profile predictions starting 30 seconds after charging until the end of the process. Predictions will continue to be updated according to the latest conditions.
This technology will greatly assist operators in controlling the burner and airflow early on, so that unwanted RoR spikes do not occur.
So that we can continue in developing the system, for the initial stage we will implement several modules including:
ERP for Coffee Roastery
Enterprise Resource Planning is a modern management system that integrates all aspects of business in the Roastery industry, from planning and actualizing purchases, inventory, production, quality control, and so on to sales and financial reports.
- Raw Material Receipt Module
- Inventory Modul
- The Batch Management module is integrated with machines that use the Adaslabs system, and has a Pre Roast Blend feature.
- Quality Control Module, integration of raw material quality data, actual roast profile, quality after roast (water content/weight loss, Agtron, Roast Level, Profile Note and so on).
- Finished Goods Inventory Module
- Modul Pree Roast Blend
- Traceability Module, which allows roasteries to provide detailed information (as desired) online to consumers by scanning the QR Code on the packaging.
- Sales Module
- Report Module
Physic Informed Neural Network (PINNs)
PINNs is an AI (Artificial Intelligence) system which is an integration of the Numeric Model (Physic Model) of a process with a Machine Learning model.
Physical model is a description of a process formulated in a mathematical model, which functions to predict the behavior of a process. Physical model is not yet an ideal model, because there are still limitations in adaptation to non-linear variables from outside the system that will affect the process.
Machine Learning Model is a model built based on the pattern or trend of the process that occurs. Machine Learning is able to connect more variables starting from raw materials, profiles during the process, to the final result. The weakness of Machine Learning lies in the availability and quality of reference data (Trained Data), in addition to not being able to adapt quickly if an anomaly occurs in the process
The advantages of Physics Model and Machine Learning, if integrated will produce an effective, adaptive system, and will have the ability to learn on its own to improve the process in the future. Anomalies that are not read in the Physics Model, will be covered by Machine Learning, and will become new constants/variables of the Physics model for the next process.
In addition to machine data (burner, airflow, drum speed) and profile (BT, ET) of each batch process sent in real time from the machine (IoT), Machine Learning data sources are from raw material quality data (Water Content, Type, Variety, Density, Size), Batch Planning (Roast Profile, Quality Target) and Batch Result (Actual Profile, Water Content, Agtron, Profile Note, etc.).
The data will be modeled by Machine Learning, and in the next batches will provide predictions of the final results to be achieved. Predictions will be given periodically during the roasting process, following changes in machine data and profiles, so that operators can immediately adjust the burner and airflow if the machine is still manual, or the machine will automatically make adjustments according to the results of the PINNs analysis.