Inter Connect Point (ICP) is your enabling partner in deploying industrial-level sensing and cutting-edge tools for monitoring and management of manufacturing production lines, and both fixed and on-the-move or even remote assets for enterprises, from full visibility to complete optimization of entire value-chains leading you towards improved service and product quality, ensuring stable and predictable returns on investment.
Our experience lies in agro-industry process automation and smart city solution design, development, project implementations, and rare earth mineral exploration. These include precision and analytics for data and controls for food processing, transport fleet, traffic, parking, water and electricity distribution, garbage disposal management, and smart home roll-outs.
Case Study: Rwanda Mountain Tea Limited, Nyabihu, Rwanda
The Industrial Internet of Things (IIoT) and Machine Learning – “Leveraging the Industrial Internet of Things (IIoT) to improve quality control and revenues in the Tea Value-chain”
In many African countries, black tea processing involves human sensory perception to determine optimum production levels of elements that are emitted during the manufacturing process to produce the desired quality. However, there are limits to reliance on human senses because these vary from individual to individual and might be affected by the individual’s current mood. These variabilities cause inconsistencies in the quality of tea produced, leading to poorer grades which reduces the revenues of tea manufacturers and their suppliers. Currently, an experienced person is relied upon to determine when various critical tea processing stages like withering, fermentation, and drying should start or stop based on their judgment and intuition developed over time. The tea expert physically monitors the green leaf as it goes through various stages of production using their physical senses like eyes and nose while being assisted by basic rudimentary measurement devices. Other researchers have attempted to apply methods like infrared bolometry and spectroscopy to determine the ratio composition of elements emitted during fermentation but these methods are expensive. It, therefore, becomes paramount that efficient and low-cost methods like the one being proposed are embraced to address inconsistencies found during tea fermentation especially for tea factories located in developing countries. The ability to solve these challenges means helping tea factories to supplement the human expert evaluation with appropriate and researched low-cost technologies. In addition, the research has the potential to transfer knowledge and solve similar challenges that are existing in other African agroindustries such as coffee and cocoa production. This research project has the potential to also address the challenges of food security and therefore lead to the improvement of livelihoods. When manufacturer revenues improve and remain consistent, the farming communities (out-growers) that supply these factories also benefit by getting improved and consistent earnings from their tea.
Project Description:
This solution explored novel ways of offering an integrated Industrial Internet of Things (IIoT) hardware, software, and analytics solution to supplement human expert evaluation in order to improve tea production, using Rwanda Mountain Tea Ltd. in western Uganda as a case study. The proposed approach is the use of the electronic nose (eNose) sensors and machine learning techniques to drive consistent controls through triggered responses and pattern identification from the collected data during tea processing. Rwanda Mountain Tea Ltd was proposed to be our first research partner in the implementation of the IIoT solution. The factory has 8 sister tea factories Rwanda and is frequently increasing supplier farmers practicing both large-scale and small-scale farming. The factory has contributed significant tax revenues, served multiple smallholder farmers and provided direct jobs. With improved processes leveraging advanced technology, these earnings further improve.
Design:
An integrated Industrial Internet of Things (IIoT) and machine learning approach has been proposed to supplement the processing of black tea at Kayonza Growers Tea Factory in western Uganda. Later, the project designs and findings will be scaled to other tea factories in Uganda and East Africa and be used to suggest how the knowledge gained can be transferred to make improvements in other dynamic African agro industries such as coffee and cocoa.
The approach supplements the human-sensory based perception in tea processing. The approach proposes the use of customized low power end-node sensor devices and machine learning techniques to improve controls/supervision and consistently determine optimum levels of environmental elements like temperature and humidity, and other high-level elements that are emitted during green leaf withering, fermentation, drying, and grading of black tea. During tea production, different gases that include oxygen, carbon dioxide, and hydrogen are emitted as high-level elements. These gases are considered to form an integrated electronic nose (eNose) sensor array that will be used to track and send data.
Temperature and humidity sensors are integrated to establish moisture content variations during the tea processing. The data from the sensors and other devices is sent over an encrypted network to a data analytics platform for analysis and visualization using the LoRa® wireless modulation for seamless long-range communications. By tracking the best tea produced (based on what price it attracts during periodic auctions), thresholds are set and optimized over time to best manage the tea production processes.
Activities:
- Interviews with internal factory staff are carried out to understand the current tea production process. This is a key fact-finding mission to validate assumptions about the needs or problems before execution. Additionally, this ensures minimizing possible waste to achieve a lean approach by going out with an intent to learn and devising the best means to achieve the solution.
- A customized low power end-node sensor device to form the electronic nose (eNose) sensor array is designed. Specifically, this chamber is composed of carbon dioxide, oxygen, and hydrogen sensors which sense the high level emitted elements during tea processing. In addition to these major elements, we explore other possible elements that are key to tea processing and include them in the design.
- Unsupervised and supervised machine learning techniques like principal component analysis (PCA), k-means clustering, correlation, and regression are used to identify patterns and predict optimum thresholds observed from both the high-level elements emitted during tea processing and environmental elements like temperature and humidity respectively.
- A cloud-based analytics platform extracts sensor data via an Application Program Interface (API) and relays it for real-time monitoring through visualizations.
Outputs:
Deployment generated significant data that is available for research upon request. A peer-reviewed research article was published in a credible conference and journal (IEEE AIIoT) at: https://ieeexplore.ieee.org/document/10385114
Impact:
This solution has generated novel approaches to tea manufacturing that is setting of minimum processing standard agreements.