Process Monitoring of Demolition Waste Recycling Crushing: Feasibility study
Juvonen, Atte (2023)
2023
Konetekniikan DI-ohjelma - Master's Programme in Mechanical Engineering
Tekniikan ja luonnontieteiden tiedekunta - Faculty of Engineering and Natural Sciences
This publication is copyrighted. You may download, display and print it for Your own personal use. Commercial use is prohibited.
Hyväksymispäivämäärä
2023-06-14
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:tuni-202306096642
https://urn.fi/URN:NBN:fi:tuni-202306096642
Tiivistelmä
Demolition waste is produced when buildings or other infrastructure approach the end of their life and are demolished. To support the objectives of circular economy and financial interests, demolition waste can be processed to produce recycled aggregates and by-products using crushing and screening equipment. Due to the nature and components of the demolition waste, the crushing and screening process is subject to different disturbances, which can compromise the efficiency of operation along with other negative effects.
Directions for future development of the machinery for processing demolition waste have sparked the need to research the possibility of the equipment achieving a level of awareness on the state of the process. Ultimately, the crushing and screening plant could detect the problematic process state in an early phase and avoid serious consequences that might result from for example a total blockage of the machine.
This thesis researches the subject with several methods. A literature review was done to collect information on the subject and to get familiar with different approaches used elsewhere in the industry. An interview study was conducted to gather existing knowledge about the demolition waste crushing process and different failure types that may occur during the process. Information from these two phases were used to build understanding on monitoring the crushing and screening process. Finally, an empirical part of the study was carried out, consisting of a measurement campaign on a real-world process, and a failure case -based analysis for the measurement data.
As a result from the interview study and follow-up analysis, an overview of the demolition waste crushing process and its possible failure modes was formed. The scope of the thesis was limited to track-driven impactor crushing plants, and the results and analysis maintained this focus as well. A total of 11 failure modes were identified, along with their possible causes, root causes and effects on the process.
The measurement campaign was planned with objectives derived from the scope of this thesis as well as interests within a wider scope. Three working days of plant operation were captured using data acquisition equipment and microphones, accelerometers, existing signals on the control system of the plant, as well as additional mechanical sensors. The data was analysed on a failure case -based approach, utilizing all information gathered in the previous phases.
Results of the analysis indicate that detecting anomalies from the crushing process can be done using data from audio-, vibration-, and other domains, but performance greatly depends on the type of process failure and the nature of associated phenomena, also being dependent on correct sensor selection and placement. Different data types were demonstrated to be useful in the analysis, but the overall picture is governed by variability in the process and possible ways it can fail. In the literature, modern deep learning -based methods were suggested as a solution to combat the complexity, but they could not be included in the scope of this work.
Directions for future development of the machinery for processing demolition waste have sparked the need to research the possibility of the equipment achieving a level of awareness on the state of the process. Ultimately, the crushing and screening plant could detect the problematic process state in an early phase and avoid serious consequences that might result from for example a total blockage of the machine.
This thesis researches the subject with several methods. A literature review was done to collect information on the subject and to get familiar with different approaches used elsewhere in the industry. An interview study was conducted to gather existing knowledge about the demolition waste crushing process and different failure types that may occur during the process. Information from these two phases were used to build understanding on monitoring the crushing and screening process. Finally, an empirical part of the study was carried out, consisting of a measurement campaign on a real-world process, and a failure case -based analysis for the measurement data.
As a result from the interview study and follow-up analysis, an overview of the demolition waste crushing process and its possible failure modes was formed. The scope of the thesis was limited to track-driven impactor crushing plants, and the results and analysis maintained this focus as well. A total of 11 failure modes were identified, along with their possible causes, root causes and effects on the process.
The measurement campaign was planned with objectives derived from the scope of this thesis as well as interests within a wider scope. Three working days of plant operation were captured using data acquisition equipment and microphones, accelerometers, existing signals on the control system of the plant, as well as additional mechanical sensors. The data was analysed on a failure case -based approach, utilizing all information gathered in the previous phases.
Results of the analysis indicate that detecting anomalies from the crushing process can be done using data from audio-, vibration-, and other domains, but performance greatly depends on the type of process failure and the nature of associated phenomena, also being dependent on correct sensor selection and placement. Different data types were demonstrated to be useful in the analysis, but the overall picture is governed by variability in the process and possible ways it can fail. In the literature, modern deep learning -based methods were suggested as a solution to combat the complexity, but they could not be included in the scope of this work.