High gas prices and CO2 emission regulations urge large industrial bakeries to reduce gas consumption. On top of that, these bakeries experience scarcity of expert bakers and factory operators which causes a risk for product quality and product consistency.
These challenges led a manufacturer of high-speed industrial bakery equipment to look for an innovative solution to support their customers, which are large industrial bakeries. The goal was to reduce gas consumption for them, without compromising product quality and consistency, and without the need for extra personnel. Our contribution was to build the technological solution.
We designed a Machine Learning (ML) solution, called ”the Sustainable Oven Service (SOS)”, that enables our customer to work with its customers to improve baking processes and oven performance.
The solution consists of two parts: a “data producer” part and a “data consumer” part.
Data producing is about gathering all necessary raw data from machines and processing that data into time sensitive series of information about performance. Data consumption is about generating insights and acting upon them.
For the data producer part, we connected several types of PLCs to the IXON industrial IoT platform. The time series data is stored in InfluxDB, hosted on Azure. For maximum automation, we supported the customer to standardize data models for several oven types. These data models contain information on oven components and measurable data such as oven temperature and humidity or actual gas usage.
After implementation of the solution, we collected benchmark data for one month to build a mathematical data model of the oven that can simulate various scenarios and calculate valuable insights, such as “how much energy is actually absorbed by the product itself and how much is lost to its surroundings?”.
Our Python built rules engine automatically generates recommendations for the baker to further improve the baking processes. For example, to change temperature setpoints, adjust vapor exhaust settings, or reduce idle time. Oven specialists can manually add expert recommendations.
To get the recommendations to the factory operators controlling the ovens (the data consumers), we’ve built an Angular web application with an intuitive user interface. The operators can monitor all aspects of the ovens and are notified of recommended changes. Each time they accept or decline a change, the software asks for a brief explanation which is logged into the database. This information can be used to continuously train the machine learning model, so it can generate even better recommendations in the future.
At first, bakeries had to get used to receiving recommendations from our AI rules engine. But in less than three months the average gas consumption (and associated CO2 emission) was 20+% lower while producing over 7% more products.
On a yearly basis, this means the bakery saves 138 tons of CO2 emissions on just one oven, which equals the average annual CO2 emission of 17 households combined.
On top of that, the feedback of bakers and operators on recommendations is stored in the database, creating a continuous improvement loop reducing the load on scarce personnel.
We’re very proud that the solution we built with our customer, has won the prestigious Las Vegas IBIE Best in Baking award.