煤礦機械遙感、制導(dǎo)與遙操作研究進(jìn)展外文文獻(xiàn)翻譯、中英文翻譯、外文翻譯
煤礦機械遙感、制導(dǎo)與遙操作研究進(jìn)展外文文獻(xiàn)翻譯、中英文翻譯、外文翻譯,煤礦,機械,遙感,制導(dǎo),操作,研究進(jìn)展,外文,文獻(xiàn),翻譯,中英文
英文原文
Recent Advances in Remote Coal Mining Machine Sensing, Guidance, and Teleoperation
J. C. Ralston, D. W. Hainsworth, D. C. Reid, D. L. Anderson, and R. J. McPhee
CSIRO Exploration and Mining
PO Box 883, Kenmore Q 4069,Australia. j.ralston@cat.csiro.au
Abstract
This paper presents some recent applications of sensing, guidance and telerobotic technology in the coal mining industry. Of special interest is the development of semi or fully autonomous systems to provide remote guidance and communications for coal mining equipment. We consider the use of radar and inertial based sensors in an attempt to solve the horizontal and lateral guidance problems associated with mining equipment automation. We also describe a novel teleoperated robot vehicle with unique communications capabilities, called the Numbat, which is used in underground mine safety and reconnaissance missions.
1.0 Introduction
Australia has excellent coal resources with a good long-term potential for wealth generation. However the Australian mining industry is currently facing a number of unique economic challenges, as well as expectations to meet increased safety requirements and strict environmental mandates. A real need therefore exists to find methods to increase coal production and improve mining efficiency while ensuring the safety of mine personnel. To this end, the CSIRO Mine Equipment Automation group is involved in developing and applying modern automation systems to coal mining machinery. Mining automation technology has significant potential to provide more accurate mining methods, incorporate sensing to improve productivity and minimise equipment damage, and to increase personnel safety by removing them from dangerous machinery and environments1,2.
An area in which modern automation technology can provide significant productivity and safety improvements is in the area of remote machine guidance. This requires the development of automated systems that are capable of sensing the material to be mined as well as guiding the mining machine to a designated target heading. These two underlying guidance scenarios are known as the horizon and lateral control problems3,4. Any practical automated system also requires integrated communication capabilities in order to transfer information on the status of the underground environment and mining equipment in real-time.Section 2 introduces ground penetrating radar (GPR) as a coal-thickness sensor for application in the horizon control problem and describes its application to a longwall coal mining operation. In Section 3, an internal navigation system (INS) is presented as a solution for the lateral control problem. Section 4 presents a novel teleoperated field robot, called the Numbat, which is used for underground mine safety and reconnaissance activities. Numbat’s real-time communication capabilities are used to convey information as to the condition of an underground mine in situations where it could be too hazardous for manual exploration.
2.0 Ground Penetrating Radar for Horizon Control
This section describes the development of a GPR system for measuring coal thickness in coal mining operations. Although GPR has significant potential for depth measurement, the raw signals are frequently complicated and so cannot be readily interpreted by mining personnel. We show how real-time digital signal processing plays a key role in transforming the raw radar signals into a form that can be readily understood. We also indicate some of the unique challenges encountered when implementing a radar processing system in a harsh coal mining environment.
2.1 Coal Seam Horizon Sensing
In an underground coal mining operation, there is an optimal roof and floor coal thickness that is required to provide sufficient structural support while maximising product extraction and avoiding unnecessary product waste5. If the remnant coal is too thick, permanently unrecoverable coal is left. If the layer remaining in the roof is too thin, it can greatly increase the risk of roof fall. Mining hard rock instead of coal also greatly increases the risk of machine damage. Figure 1 shows a cross sectional view of a typical underground mining operation, where the goal is to keep the mining machine in the coal seam by following an optimal mining horizon. In order to realise this objective, a method is required to determine the depth of coal.
Figure 1: The horizon control problem. Here the underlying objective is to keep the mining machine in the coal seam to in order to maximise coal recovery. The black bands represent coal and the textured zones represent tuff.
2.2 Existing Coal Depth Sensors
There are two main classes of coal depth sensing systems: reactive and predictive3. Reactive coal depth sensors are based on detecting changes in operational characteristics when a coal-rock interface is encountered. Reactive methods include monitoring cutting drum current, sensing changes in machine vibration signatures, and the use of instrumented cutting picks. The underlying limitation with reactive methods is that they rely on the actual penetration of the coal- rock interface, in which case it is too late to prevent product dilution and machine damage. Unlike reactive methods, predictive sensors attempt to estimate the distance from the machine to the coal-rock interface. The most common predictive method uses a gamma radiation sensor to estimate the coal depth by exploiting the radioactivity of host strata typically found with coal deposits. The natural gamma sensors can work well in specific situations but require up to 30 seconds to obtain a reliable depth measurement. Such a delay can impact severely on coal mining operations. This limitation motivated the search for alternative predictive sensing methods.
2.2 GPR as a Depth Sensor
GPR has found special application for subsurface characterisation in civil engineering, ordinance detection, and geotechnical fields6,7,8. The central attraction of GPR is that it is non-invasive and non-destructive, and can provide instantaneous imaging of subsurface features. The basic principle of GPR involves transmitting energy into the ground and then measuring the reflections arising from the interface of materials with different dielectric values. The magnitude of the received reflection is dependent on the ground conductivity and permittivity, the size and shape of the target, and the difference between dielectric constants at a boundary. Voids, cavities, and other dielectric interfaces represent discontinuities that can give rise to pulse echoes. The geological features typically found in coal-bearing strata are particularly amenable to radar imaging. This is because coal has a relatively low conductivity and high dielectric constant with respect to its host strata.
To date GPR has not been widely used the coal mining industry for coal depth measurement. This is largely because of the practical issues associated with making sensitive electronic equipment suitable for a rugged and hazardous coal mining environment and because of the complex nature of the radar signals returns from a GPR system. A key requirement in the successful implementation of a coal measurement system is the use of signal processing to transform the raw radar data into a form that can immediately utilised by non-expert personnel. Most commercial GPR systems are not suited to underground coal mining and do not provide automated real-time processing capabilities, and consequently rely on trained operators to manually interpret the raw radar data in an off-line capacity9.
Figure 2 shows typical raw GPR output with the sensor directed into free space. The main peak represents an antenna-air coupling characteristic. Clearly additional processing of the raw radar data is required in order to provide coal depth estimates for horizon control.
Figure 2: Snapshot of raw GPR signal showing radar-air coupling characteristic.
2.3 GPR Signal Model and Processing
GPR-based coal thickness sensors rely on measuring the propagation delay from the sensor to the coal-rock interface propagation delay and knowledge of the local coal geology. We first derive a model for the received radar signal by considering the physical arrangement of the imaging scenario. The resulting set of equations lead to the associated processing tasks.
2.3.1 Radar Signal Model
The received radar signal can be modelled as2where W0(t) is the radar-air coupling pulse, W(m) is a lumped impulse response of the radar, M specifies the temporal support of the (two-sided) wavelet, and Nk(t) is an independent noise process for time t=0,…,T-1, and radar stretch index k=0,1,…. The sequence Nk(t) includes sensor and timing jitter noise. The sequence Rk(t) is given by where ak (n) is the magnitude of the echo and tk(n) is the pulse time delay for n=0,…,pk-1. The signal Rk(t) is analogous to the reflectivity series found in seismic imaging10. Note also that this is an approximation since the pulse shape can vary depending on the material being imaged.
Equation (1) represents the convolution of the basic pulse wavelet with the (scaled) reflection coefficients. In a given radar signal observation Zk(t), there may be on pk interfaces present. The underlying goal is to estimate tk(n) and pk, and in particular the first reflection, from observations of Zk(t) for all k. Figure 3 shows typical raw radar data, Zk(t), arranged as a sonar- style plot for t = 0,…,479 and k = 0,…, 999.
Figure 3: Array of raw GPR data. The vertical axis represents time delay and the horizontal axis shows adjacent radar data while the sensor is held in a stationary position.
2.3.2 Radar Signal Processing
Figure 4 shows the four basic stages involved in processing the raw radar data: pre-processing, filtering, detection, and estimation. This includes time-varying gain for path loss compensation, suppression of radar-air coupling characteristic, wavelet deconvolution, delay and displacement omain filtering, short-term ensemble averaging, energy detection and peak location.
Figure 4: Block diagram showing processing applied to the raw GPR data to provide coal depth estimates.
Preprocessing Stage
There are two preprocessing steps used. The first step is to apply an exponential gain to the received signal to compensate for the radar signal attenuation characteristic through earth material. This time-varying gain considerably extends the dynamic range of the system and has the effect of removing the time dependency of ak(t) in Equation (2). The second preprocessing step is to remove the transmitter-air/receiver coupling characteristic. An estimate of W0(t) in Equation (1) is made during initial system calibration by computing an ensemble average of Zk(t) over k assuming statistical stationarity of the raw returns over a nominated calibration interval.
Filtering Stage
Since the movement of the mining machine is very slow with respect GPR sampling rate, Doppler effects are negligible and so is a good approximation for n = 10 – 20 adjacent raw returns. This property permits lowpass filtering in the displacement domain, which serves to suppress artifacts arising from radar timing jitter. Lowpass filtering in the time/delay domain also proves particularly effective in suppressing the effects of sensor noise and machine vibration. To improve the detection power, the radar data is deconvolved with a regularised inverse11 obtained from a suitable estimate of the basic radar wavelet W(t). The small phase shift introduced for causality reasons is accounted for in the processed data.
Detection Stage
The processed data is passed through a detection stage to determine potential dielectric interfaces. A time-domain sliding window energy detector is implemented at each instant k. The output of the energy detector is then compared to a threshold previously determined empirically from field tests and calibration. If the detector output exceeds the given threshold, an estimate of the average two-way travel time is then made by computing the median of the resulting set of (sequentially ranked) indices within the domain of the energy window. An estimate of pk is obtained by noting the number of distinct sets of peaks in a given realisation. This strategy represents a simple yet effective method for detecting the coal-tuff interface in an automatic manner, as well as being computationally efficient to facilitate real-time implementation.
Depth Estimation Stage
Given that the first echo represents the two way travel time for the coal-rock interface, an estimate of the coal depth is given by for n=0,…,pk-1, where ec ? 4.5 for coal. The processed radar data leads to a greatly simplified representation and thus acceptance of the technology to mining personnel. The data can also be meaningfully employed in more tracking and control schemes. It is important to note, however, that the radar data is subject to a wide variety of aberrations from the basic radar model: Nonlinearities, multiple reflections, heterogeneity variation, moisture variation, sensor and mining machine vibration, operator misuse, and sensor placement all contribute to measurement error2.
2.4 GPR Longwall Application
The GPR unit was applied to an underground coal mining machine known as a longwall shearer. The primary purpose of our tests was to establish whether the radar unit could be used as a sensor for horizon control in a coal mining scenario. To this end the GPR-based sensor was mounted on a longwall shearer at Dartbrook, NSW, Australia. The particular underground scenario had a problem that offered a long term potential for an automatic coal thickness measurement system. The floor of the seam being extracted consisted of a weathered clay layer with a high ash content (~80%). There was real benefit in cutting as close to the floor as possible as the coal immediately above the tuff layer was of high quality. However, if the tuff layer was mined, the ash content of the product increased dramatically.
The GPR processing system as a whole consists of three main components: the wideband (800 MHz) bistatic impulse radar, the data processing unit, and the visualisation system2. This antenna configuration produces a 5 ns data stretch, which results in high-resolution short-range (150 cm) echo data12. The complete GPR system was first assembled and tested on the surface. It was then transported underground and mounted on the longwall shearer as seen in Figure 5. The terms of the exemption for the operation of the GPR underground specified that the equipment could only be operated during maintenance periods.
Figure 5: The GPR processing and display unit mounted on the longwall shearer.
2.4.1 Test Procedure
Testing involved making coal thickness measurements in the area between the horizontal push rams of adjacent roof supports. The longwall shearer was moved to a position on the face where the remnant coal thickness exceeded the radar’s nominal penetration depth. The radar unit was placed directly on the surface of the coal and the returns recorded. Figure 6 shows GPR echoes from an undisturbed region where the coal-rock interface was out of the GPR range (greater than 300 mm).
Figure 6: Raw radar output display in the case where the coal-rock interface is out of the range of the radar. No reflection is observed.
A test bench was constructed to manually establish a local floor horizon within the range of the instrument. Figure 7 shows the raw radar data at 100 mm coal thickness with the horizontal axis representing the coal-rock interface. The coal-tuff interface can be clearly seen in where the coal thickness is within the range of the sensor.
Figure 7: Output from the GPR display unit in the presence of a coal-rock interface approximately 100 mm below the surface.
Considerations for Underground Use
Special design and construction considerations were necessary in order to make the radar processing system suitable for use in the harsh coal mining environment. In addition, the system need operate in the presence of potentially explosive gases. The signal processing module and display screen are housed in a ruggedised flameproof enclosure with a viewable window as shown in Figure 8. Of special interest is the use of a non-metallic flameproof enclosure for the radar transmitter-receiver assembly. Clearly this was necessary as the use of a metallic enclosure would not permit transmission of the radar signals. The radar enclosure is made of a ruggedised carbon-based alloy of a sufficiently low dielectric constant so that the radar signals are not adversely attenuated, and is interconnected with the main flameproof via an armoured cable.
Figure 8: Flameproof enclosure to house the signal processing and
display components of the GPR system.
The GPR system must operate with as little operator intervention as possible (ideally none), be physically robust, comply to strict intrinsically safety requirements, and facilitate remote software maintenance and data retrieval. As a result, the operator “interface” is limited to four ruggedised push buttons to select calibration and other control.
2.6 GPR Summary
GPR has the potential to solve an important coal depth estimation problem in the mining industry. A GPR processing system for coal depth measurement has been designed, built, and evaluated on a coal mining machine. The first generation radar measurement system developed provides data acquisition, processing, and visualisation of coal thickness estimates in real-time, as well as remote data communication facilities. Experiments conducted show that there exists a positive correlation between the coal-rock interface and known geology. The continued development of radar signal processing techniques is the key to improving the reliability and utility of the measurement system.
3.0 Inertial Navigation Systems for Highwall Mining
Highwall mining is an important method of coal mining in which a remotely controlled mining machine is driven into a coal seam that has been exposed by previous open cut operation. A continuous haulage system is use to carry the coal from the miner to an open-air installation for stockpiling and transport. The highwall mining process forms a series of nominally parallel, unsupported drives. Figure 9 shows a highwall mining machine in operation.
Figure 9: A highwall mining machine creating a drive into an exposed coal seam.
3.1 The Need for Lateral Guidance
It is vital that the coal pillars remaining between adjacent drives are capable of supporting the overburden structure. This mining method is totally reliant on effective remote control. Straight, parallel openings at the tightest separation consistent with geotechnical design can only be achieved if the mining machine’s position and heading can be determined and controlled remotely. Figure 10 shows a plan view highlighting the lateral guidance problem associated with highwall coal mining.
A number of methods have been proposed for lateral guidance of highwall mining machines. Laser surveying offers the potential for accurate longwall machine position determination but site- dependent issues such as visibility and line of sight mean that a general approach is not possible. Some opto/mechanical borehole-type surveying methods have been considered but also suffer from practical implementation difficulties because of the need for high accuracy angle measurements in a very inhospitable environment. Special purpose triangulation systems using radio propagation cannot achieve the required machine positional accuracy using the low frequencies required to penetrate the overburden. The limitations associated with the existing methods strongly suggest the need for a positional sensing system that requires only a minimal external infrastructure.
20m
500m
Figure 10: The lateral control scenario encountered in highwall mining. The goal is to maintain a targ
收藏