Mining, one of the world's oldest industries, has historically played a significant role in social and economic development. However, as near-surface mineral deposits are progressively depleted, the industry faces the challenge of exploring and extracting resources at increasingly greater depths. In recent years, automation has emerged as a key trend in mining. Much like the development of self-driving cars, robotic lawnmowers, and other technologies designed to perform tasks deemed too hazardous or monotonous for humans, the mining industry is advancing toward greater automation. Some companies envision a future in which no human workers are required underground, thus marking the beginning of a transition to fully autonomous mining operations. This vision includes fully automated mines with processes monitored and optimized to achieve productivity and safety objectives. Significant progress has already been made, such as the Aitik mine operated by Boliden became the first in Europe to introduce self-driving trucks equipped with a variety of sensors to operate safely and avoid harming people or animals. An important concern in mining is the safety of humans and equipment in the mine operation to maintain productivity. The major safety concerns constitute fires and leakage of gases. Despite the advancements, fires remain a persistent issue in mining operations, occurring approximately once a week in Swedish mines, with similar frequencies reported globally. Notably, 80 % of these fires originate from mining vehicles and 80 % of those fires is caused by hydraulic oil leakage. Smoke from fires poses severe risks, including potential injuries, fatalities, and costly production halts. Additionally, explosive and toxic gases can delay operations. The current "fire alarm" in underground mines often relies on manual detection by miners. Developing reliable fire detection systems is a critical step toward realizing the vision of autonomous mining. Early detection of faulty equipment such as overheated cables or motors, is essential to prevent fires from escalating. Similarly, identifying oil leaks in motors or hydraulic systems can mitigate fire risks.
Various sensors, particularly gas sensors, for early detection of gases are proposed for installation on mining machines based on the results of rigorous on-site and offsite experiments. Among the sensor technologies the photoionization detector (PID) technique, based on Albert Einstein’s Nobel Prize-winning work in Physics (1921), stood out. In PID sensors, organic molecules are excited by ultraviolet (UV) light. These molecules then pass through an electric field, generating an electrical pulse upon striking an electrode, which allows for quantification. Various PID sensors with sensitivities ranging from parts per million (ppm) to parts per billion (ppb) have been tested. Notably, the ppb-sensitive sensors demonstrated promising results in detecting gases emitted at early stages, such as from overheated cables.
Mounting gas sensors on mining machines offer several advantages, such as proximity to fire sources, adaptability to changes in mining areas, simplified maintenance and calibration, and integrated data logging that facilitates correlation with machine data.
To summarize, this thesis proposes a comprehensive framework utilizing Bayesian Network creating an expert system architecture with maintenance-on-demand capabilities. This approach aims to enable predictive maintenance and enhance safety by identifying potential hazards before they escalate.
In addition, other tools were experimented with, for instance drones providing visual feedback when miners are no longer working underground, or for measuring gases. Furthermore, new technologies such as augmented reality (AR) were experimented with, and the conclusion is that AR can play a role in improving human interaction in the troubleshooting process when support is needed to solve a problem and thus contributes to achieving the net goal of reducing greenhouse gas emissions.