Software engineering
ELRA & Zühlke: Competitive edge through an innovative user experience
Learn more
Homepage zuehlke.com
Industrial sector
Kanadevia Inova aims to reduce downtime & costs with AI
Zühlke and Kanadevia Inova (former Hitachi Zosen Inova) developed and operationalised an AI-based computer vision detection system to optimise the waste management process of bulky waste.
The project at a glance
Kanadevia Inova is a global Greentech company operating in Waste-to-Energy (WtE) and renewable gas sectors. Kanadevia Inova develops projects with its clients, drawing on its experience as a general engineering, procurement, and construction contractor, to deliver complex turnkey plants and system solutions for thermal and biological WtE recovery, gas upgrading, and power to gas plants.
One of the main challenges in WtE plants are bulky waste items because they often lead to disruption in the feeding process for combustion. This can lead to severe consequences such as unplanned maintenance and downtime, fluctuation in steam condition, plant unavailability, and risk to staff safety.
“Zühlke's collaborative approach and cutting-edge solutions will enable us to tackle a critical challenge in bunker management, with the potential for smoother operations and reduced downtime.”
With a multidisciplinary team, including data scientists, Kanadevia Inova developed a computer vision algorithm that helps plant operators to detect and handle bulky items, and improve plant operation and maintenance.
Multidisciplinary know-how leads to a machine learning-based solution
Zühlke joined the project after an initial proof of concept but before an existing minimum viable product (MVP) was created. A detection algorithm had to be developed to meet the required performance criteria and be fully deployable as a pilot product.
Combining data science, data engineering, UI/UX, cloud platform technologies, software architecture and development, and Machine Learning Operations (MLOps) know-how, Zühlke supported the Kanadevia Inova team in developing a data processing pipeline and a computer vision detection algorithm for bulky waste items.
“With Zühlke's expertise in data science and MLOps, we were able to transform our concept into a reliable, operational AI system, enhancing bunker management safety and efficiency.”
The Zühlke specialists developed a web application through which plant operators receive alarms (including key information) of detected bulky waste items from the algorithm and provide feedback.
Zühlke also conceptualised and implemented an initial MLOps system with which the detection algorithm can be adapted and retrained for each plant individually and semi-automatically, accounting for the specific needs of each deployment site. This leads to significant time and cost savings for future implementation in other plants.
Pilot demonstration on customer plants
The image processing and bulky waste detection algorithm was deployed in two pilot WtE plants, one is operated by AVAG Umwelt AG in the city of Thun (CH) and the other is operated by Renergia Zentralschweiz in the city of Perlen (CH).
The waste tipping bay gates are continuously monitored by a set of specific cameras and the recorded video stream is fed into the deployed detection pipeline. Newly arriving objects are analysed and potential bulky waste items are identified near real-time. Additionally, plant operators are notified about bulky waste items through the provided web application. That means that both pilot customers have the opportunity to test the live data throughout the pilot development process and challenge the result with the aimed value for the daily operation.
“Zühlke played an essential role in this development. Their comprehensive support from data processing to machine learning and MLOps brought our AI project to life. The initial detection performance goals were met and a better than-targeted failure rate was achieved.”
By continuously developing the system together with its customers, Kanadevia Inova will be able to add value, including more reliable and safe processes, less downtime, and significant cost savings.
