AU2016202377B2

AU2016202377B2 - Improved mining machine and method

Info

Publication number: AU2016202377B2
Application number: AU2016202377A
Authority: AU
Inventors: Mark Thomas DUNN; David William Hainsworth; Chad Owen Hargrave; Jonathon Carey Ralston; David Charles Reid; Peter Bryan REID; Jeremy Patrick Thompson
Original assignee: Commonwealth Scientific and Industrial Research Organization CSIRO
Current assignee: Commonwealth Scientific and Industrial Research Organization CSIRO
Priority date: 2015-05-28
Filing date: 2016-04-14
Publication date: 2020-11-26
Anticipated expiration: 2036-04-14
Also published as: CN106194184A; CN106194184B; AU2016200785B1; AU2016202377A1; AU2016200782B1

Abstract

This invention relates to a mining machine including an extraction device, an at least 2D co ordinate position determining device for generating a future path according to an intended cut profile based on the determined current co-ordinate position of the mining machine as distinct from an expected co-ordinate position; and for generating a seam model of the seam to be cut and/or a cut model of the seam that has been cut. The mining machine further includes sensors to collect characterising data of the seam, said characterising data forming part of the seam model and/or cut model and includes a processor to generate the seam and/or cut model and to control parameters of said mining machine. The processor controls the machine parameters based upon analysis of the seam model and/or cut model to anticipate changes in mining conditions as it progresses along the intended cut profile within the seam model. 1/8 100 106 108 110 114 11 -116 103 101 0120 a 105 112 126 -- 2 102 128 104z 134 Yt -x 136 Fig. 1 200 208 2 06 A(204 1 2 3 4 5 6 7 8 9 10 1'' Fig. 2a 202 250 Fig. 2b x

Description

1/8 100 106 108 110 114 11 -116 103 101 0120

a 105

112 126 -- 2

102 128 104z

134 Yt -x

136 Fig. 1

200 208

2 06 A(204

1 2 3 4 5 6 7 8 9 10 1'' Fig. 2a 202

250

Fig. 2b x

IMPROVED MINING MACHINE AND METHOD

Cross-Reference to Related Applications

The present application claims priority from Australian Provisional Patent Application No 2015901979 filed on 28 May 2015, the content of which is incorporated herein by reference.

FIELD OF THE INVENTION

This invention relates to a mining machine and method whereby a mining machine can be controlled to move across or along a seam containing material to be mined.

DESCRIPTION OF PRIOR ART

In the mining of minerals and coal, processes have been developed which extract these material from a seam. One particular process is referred to as the longwall mining process. In this process, among other components, a movable rail is placed to span across a coal seam. A mining machine is provided with at least one shearing head and the mining machine is moved to traverse along the rail from side-to-side of the seam, and the shearing head or heads are manipulated upwardly and downwardly to shear coal from the face of the seam. Throughout each pass, the rail is moved forwardly toward the seam behind the path of the mining machine. The mining machine is then caused to traverse the seam in the opposite direction in order to repeat the shearing process. During this return traverse the shearing head(s) may also if desired be manipulated upwardly and downwardly to remove further coal from the seam. The process is repeated until all coal in the planned extraction panel is completed.

Thus, by advancing the rail means forwardly towards the seam by a suitable distance after each pass, it is possible to progressively move into the seam with an approximate equal depth of cut with each pass.

However, the seam height, depth and composition is not uniform and, as such, improvements in efficient and effective extraction from the seam is still required, whether the extraction be using longwall mining techniques or other extraction processes in which the spatial position of the extraction device and the seam is important in being able to remove material efficiently and effectively.

Any discussion of documents, acts, materials, devices, articles or the like which has been included in the present specification is not to be taken as an admission that any or all of these matters form part of the prior art base or were common general knowledge in the field relevant to the present disclosure as it existed before the priority date of each claim of this application.

Throughout this specification the word "comprise", or variations such as "comprises" or "comprising", will be understood to imply the inclusion of a stated element, integer or step, or group of elements, integers or steps, but not the exclusion of any other element, integer or step, or group of elements, integers or steps.

STATEMENT OF THE INVENTION

In a first aspect of the present invention, there is provided a mining machine including:

A. an extraction device for removing material from a seam;

B. an at least 2D co-ordinate position determining device for determining the co ordinate position in space of the mining machine, said position determining device generating current co-ordinate position output data signals for:

i. generating a future path of the mining machine determined according to an intended cut profile based on the determined current co-ordinate position of the mining machine as distinct from an expected co-ordinate position; and

ii. generating at least one of a seam model of the seam to be cut and a cut model of the seam that has been cut;

C. one or more sensors, distinct from the at least 2D co-ordinate position determining device, to collect seam characterising data of the seam from which material is about to be extracted or adjacent thereto, said seam characterising data forming part of the at least one of the seam model and cut model, with the co-ordinate position of the seam characterising data determined by reference to the co-ordinate position in space of the mining machine, said sensors providing current seam characterising data output signals therefrom;

D. a processor connected to receive the output data signals from said position determining device and said one or more sensors to generate the seam and/or cut model and further generate output data signals to control one or more parameters of said mining machine, and

E. a memory storage device to store the seam model and/or the cut model,

wherein the processor controls the one or more parameters of said mining machine based upon analysis of the seam model and/or cut model to anticipate changes in mining conditions as said mining machine progresses along the intended cut profile within the seam model.

The current coordinate position may be an absolute coordinate position, and previously acquired surveying and exploratory data, which may be provided with absolute spatial positioning coordinates, may be integrated into the seam model.

In some embodiments, the memory device is accessible to a central memory storage device such that the cut and seam model data may be accessible across multiple mining machines or that the cut an seam model data may be analysed such that processor algorithm may be updated to reflect the variations in the expected and actual performance of the mining machine.

In a preferred embodiment both the seam model and the cut model are analysed by the processor to anticipate changes in mining conditions.

Through the integration of the spatial positioning information from the position determining device combined with the information about the seam characteristics to be mined, the mining machine performance may be improved. In contrast to the use of sensors to reactively control the mining machine, the mining machine of the present invention takes advantage of stored output data contained with the seam model and/or cut model for analysis by the processor to ensure that control signals utilise a greater breadth and depth of information. The mining machine of the present invention is able to more precisely control the extraction from a seam within the boundaries of targeted material being mined. The intended cut profile may be better aligned with the seam model, such that the mining machine can optimise operational settings based upon a better understanding of the location of the mining machine, its components, and the seam, including its varying compositional and physical characteristics.

The co-ordinate position of the seam characterising data is preferably determined using geometric (e.g. triangulation and/or trilateration) techniques and the relative position of the one or more sensors and the at least 2D co-ordinate position determining device. In some embodiments, the one or more sensors and the at least 2D co-ordinate position determining device are clustered within the same approximate location. Range finding sensors, such as 2D or 3D laser or radio frequency-based sensors may be used to provide measurements for determining the co-ordinate position of the seam characterising data. Alternatively, or in additional to, known nominal distances between mining machine components and the seam may be used.

Preferably, the cut model comprises the co-ordinate positions of the seam which has been mined and at least one seam characterising datum tagged (i.e. spatially registering a seam characterising datum) to each position. Preferably, the cut model comprises the co-ordinate positions of the seam which has been mined and at least one mining machine characterising datum tagged to each position.

Through creating a model (preferably a 3D model) of the seam which has been extracted and the interaction between the mining machine and the seam, this seam model information may be used to predict the interaction of the mining machine and the seam within the seam model.

Alternatively, or in addition to, the seam model may comprise the co-ordinate positions of the seam which is about to be mined and at least one seam characterising datum tagged to each position.

The anticipated changes in mining conditions are preferably determined by reference to the cut model, the seam model or a combination thereof.

An advantage of using both the cut and seam model information to determine anticipated changes in mining conditions is that the seam model information may only extend to the surface of the seam (i.e. a subset of the entire seam) to be mined. Therefore a combination of cut and seam model information may be used to extrapolate the anticipated changes in mining conditions to include the volume of seam about to be extracted.

The mining machine may include a longwall miner (including the associated rail, roof supports, drives, conveyor, stage loader and crusher), a continuous miner, a road header; a shuttle car; a flexible conveyor train; a plough, surface miner or any other mining machine with an extraction device to remove material from the seam. The mining machine may operate underground or on the surface.

In some embodiments, the mining machine is a longwall miner and the seam model has been further characterised at the tail and main gateroads. Within this embodiment, interpolation and/or extrapolation may be used between the seam characterising data along the gateroad and the seam characterising data on the seam surface to determine the anticipated changes in mining conditions to include the volume of seam about to be extracted.

Seam characterising data

The seam characterising data may be in the form of radar; optical, thermal, geological, geophysical, radiometric or spectral imaging to determined characterising properties of the seam such as density, size, volume, thermal, compositional and other physical properties of the seam. Variations in these parameters may be used to signal changes in the compositional grade of the seam or a boundary between a target seam and a non-target seam. In the case of a coal seam, the boundary between the coal seam and the non-target seam (e.g. mineral rather than coal) may take the form of density (including void detection), thermal or compositional data.

The seam characterising data may be referenced against a library of calibrated readings, such that accurate interpretation of the data may be achieved.

The seam characterising data is preferably inclusive of the material adjacent the seam to be extracted, as the adjacent material provides information regarding material not to extract (e.g. underburden, interburden, overburden or out of specification target material) and information on the stability of the seam that the mining machine is navigating through.

The seam characterising sensors provide the seam characterising data. A range of sensors may be used including thermal infrared (compositional boundaries), gamma radiation (density, leading to compositional boundaries), radar (material discrimination), geophysical (compositional boundaries and material discrimination) and laser (surface/volumetric analysis).

Mining machine characterising data

The interaction of the mining machine and the seam results in characterising data which provides information relating to resource location, volume, composition, hardness, as well as machine component location, speed and efficiency of the extraction process. This characterising data is preferably derived from the mining machine's onboard sensor and control instrumentation, which monitors the machine settings and performance.

Machine performance data is preferably derived from machine performance sensors or machine settings. The mining machine performance sensors are preferably selected from the group consisting of extraction device power consumption sensor or derivative thereof; noise sensor; vibration sensor; and extraction device location sensor.

In one embodiment, the machine performance data comprises power consumption, or a derivative there of (e.g. current) which is correlated with the seam characterising data derived from a horizon control sensor or seam boundary sensor, such as an IR sensor; gamma sensor and/or a ground penetrating radar sensor. In embodiments in which there are more than one horizon control or seam boundary sensor, the output of these sensors may be also correlated against each other to validate and calibrate the sensor output data via a closed loop calibration process utilising at least two sensors (seam characterising sensor and/or positional sensors).

Mining machine parameters

The ability of the mining machine to adjust its operating parameters to take into account anticipated changes to the mining conditions enables more efficient, effective and safe mining operations.

The one or more parameters of said mining machine are preferably chosen from the group comprising of roof support pressure; roof support height; roof support orientation; mining machine traversal speed; and extraction device location; current; vibration and/or speed.

In a preferred embodiment, the mining machine is a longwall miner, in which the extraction device is a rotatable shearer head mounted upon a moveable carriage; and the one of more sensors comprises a seam boundary detecting sensor. For the purposes of the present invention seam boundary detecting sensor will encompass a key bed package or marker band detecting sensor. The definition of key bed package is provided in US8469455.

Within this embodiment, the mining machine further comprises an actuator for moving said shearer head within a vertical plane towards a seam boundary, wherein the seam boundary detecting sensors provides current seam characterising data output signals to the processor which generates further signals to the actuator to move the extraction device (e.g. shearer head) a distance within a vertical plane towards a seam boundary according to an intended cut profile.

Alternatively, the processor may update the intended cut profile which defines the path of the extraction device, such as a shearer head.

The mining machine preferably further comprises a rail for the mining machine to traverse back and forth across the seam, said sensors collecting characterising data of said seam to generate a seam model from a first traversal to form the basis of the intended cut profile in the second traversal.

Intended cut profile

The intended cut profile comprises a horizontal plane nd a vertical plane, which may be represented in 3D Cartesian coordinates in the (x,y) and (x,z) planes respectively Preferably the intended cut profile in the horizontal plane is a straight line, which may enable optimal coal extraction to be achieved (e.g. for longwall mining a moveable carriage travels in a straight line between the main and tail gateroad). As the rail may not be able to efficiently correct the actual profile to a straight line profile within a single cycle, the intended cut profile may be an intermediate between the actual cut profile and a straight line. Within the vertical plane, the intended cut profile may be straight or may follow the top and/or bottom boundary of the coal seam. Sensor outputs which identify the seam boundaries are preferably entered into a seam model to enable the processor to generate signals to the shearing head actuator to control movement of the shearing head within the identified seam boundaries. This ensures improved extraction efficiencies without the shearing heads transgressing outside the target seam. The intended cut profile may also be used in reference to roadway development by continuous miners and roadheaders.

Position determining device

The at least 2D co-ordinate position determining device preferably comprises an electromagnetic localisation system, such as radio-frequency localisation system, or a global positioning system (GPS) or ground based (e.g. localised) wireless positioning system.

The at least 2D co-ordinate position determining device preferably comprises an optical position determining device. Preferably, the optical position determining device comprises a laser source and a laser sensor. Preferably, the optical position determining device detects registration markers that are located pre-determined positions and determines the at least 2D co-ordinate position of the machine based on the position of the registration markers.

Preferably, the at least 2D coordinate position determining device comprises an inertial navigation system (INS) to determine the co-ordinate position where no registration markers are detected. In one embodiment, the registration markers are located at the gateroads of a longwall mine such that the optical sensor determines the at least 2D co-ordinate position when the longwall mining machine is at or near the gateroads and the INS determines the

2D co-ordinates position when the longwall mining machine is between the gateroads where the registration markers are not detectable by the optical sensor.

Analysis of the seam and/or cut model

In a second aspect of the present invention, there is provided_a mining machine including:

A. an extraction device for removing material from a seam;

i. generating a future path of the mining machine determined according to an intended cut profile based on the determined current (this may be absolute) co-ordinate position of the mining machine as distinct from an expected co ordinate position; and

ii. generating a seam model of the seam to be cut and/or a cut model of the seam that has been cut;

C. one or more sensors, distinct from the at least 2D co-ordinate position determining device, to collect characterising data of the seam from which material is about to be extracted or adjacent thereto, said characterising data forming part of the seam model and/or cut model, with the co-ordinate position of the characterising data determined by reference to the co-ordinate position in space of the mining machine, said sensors providing current seam characterising data output signals therefrom;

D. a processor connected to receive the output data signals from said position determining device and said one or more sensors to generate the seam and/or cut model and further generate output data signals to control one or more parameters of said mining machine,

E. a memory storage device to store the seam model and/or the cut model,

wherein the processor analyses the seam characterising data from the one or more sensors to determine the presence of a sensor inaccuracy and/or imprecision, and in the absence of sensor inaccuracy and/or imprecision, controls the one or more parameters of said mining machine based upon analysis of the seam model and/or cut model to anticipate changes in mining conditions as said mining machine progresses along the intended cut profile within the seam model.

The at least 2D co-ordinate position determining device may determine the absolute co ordinate position in space of the mining machine.

The inaccuracy and/or imprecision of a sensor preferably has a predetermined or designated value(s). The predetermined values may be expressed as a confidence profile from which a confidence value or confidence limit may be derived. This predetermined value is preferable determined from validation trials of the mining control systems within a particular mine and sensor system, utilising for example, spatial and temporal information whilst operating in the specific mining environment.

If the processor determines that a sensor is inaccurate and/or imprecise, the processor preferably signals a level of confidence and/or error to the machine operator.

The output data from a sensor preferably has a confidence profile associated with it. The confidence profile preferably includes statistical information relating to the characteristics of the output data such as the distribution, mean, and standard deviation. The expected value of output data of the sensor will depend upon its confidence profile and the designated confidence value (e.g. a value falling within a 90% confidence limit or 99% confidence limits). The designated confidence value may vary dynamically according to the machine operation being undertaken. For example, a high confidence value/limit may be required for an operation where the consequence of failure is higher as a means of managing potential mining risk (e.g. human injury or death compared with reduced production efficiency or increased machinery wear).

Preferably, the cut or seam model further comprises a confidence profile and confidence value derived from the inaccuracies and imprecision attached to the co-ordinate position data and/or the seam characterising data, or from the confidence profile of the seam and positional sensors. The confidence profile or value preferably comprises a range of seam characterising data and a range of positional data. The confidence value will be dependent upon the level of confidence (e.g. a 99% confidence value provides a broader range of seam characterising data and positional data compared to a 90% confidence value).

The precision associated with a sensor may be adjusted through using smoothing algorithms to filter signal noise. Preferably, the sensor performance is matched to the mining machine dynamics, such that the processor control instructions to a mining machine parameter is matched to the capability of the mining machine.

A goal of the present invention is to incorporate sensor output data into the cut and/or seam model while not incorporating the totality of inaccuracies and imprecision of each of the sensors. To achieve this goal, the processor preferably validates and calibrates the output data from individual or groups of sensors prior to the use of the sensor output data. Through improving the confidence profile of the output data of the sensors, the models which this data forms the basis of is also improved (i.e. the seam model; cut model and intended cut profile), thereby improving the control of the mining machine.

The processor preferably determines that a sensor is inaccurate and/or imprecise when the output data from the sensor conflicts with a validation source. A conflict between the output data and a validation source may occur when the expected value of the sensor output (i.e. confidence value of output data) and validation source do not correlate or overlap or do not correlate or overlap to a predetermined level.

Validation sources may include:

• the expected seam characterising data derived from the cut and/or seam model (including sub models thereof); • output data from a different sensor(s) measuring the same or related parameters, including machine performance sensors; * measurements from manual observations; and/or 0 the confidence profile defining the expected precision and accuracy of the sensor when calibrated or commissioned.

The more validation sources used the less likelihood that there is a false assessment of sensor inaccuracy and/or imprecision. Preferably, at least one validation source is an external reference to avoid potential accumulated errors associated with the use of internal references, such as output data from an associated sensor. In general, the more validation sources used the greater the ability to increase the confidence values of the output data of the sensors (i.e. reduce the expected output range before corrective action is taken address the inaccuracy or imprecision).

If the processor determines that a sensor is inaccurate and/or imprecise; the processor may initiate corrective action. Corrective action preferably comprises calibrating or correcting the sensor data output against a validation source. The corrective action may be selected from one or more of the following actions: a. calibrating the sensor; b. using the output from an alternative or additional sensor to correct or calibrate the output data from the sensor; c. using the seam model and/or cut model, or sub model thereof, to correct or calibrate the output data from the sensor; and/or d. using an extrapolation of the cut model to correct or calibrate the output data from the sensor.

The calibration of the sensors may be carried out in a variety of means including calibration against a reference standard or against positional locations in which accurate survey data is available. In one embodiment, the output data from the sensor is calibrated or corrected by exploratory and/or survey data which preferably forms part of the seam model.

In one embodiment, the calibration of a sensor is carried out relative to another sensor or another validation source. The use of another sensor for calibration may be used particularly if a correlation exists between the output data of the two sensors and one sensor functions as a reference that preferably has a greater accuracy and/or precision than the other sensor.

In an alternative embodiment, the sensor is calibrated using the seam model. The benefit of this approach is that the seam model, or a sub model thereof, is formed from sensor output data and/or prior knowledge (e.g. survey data) which has previously undergone the validation process.

In addition to seam characterising sensor data, the seam model is also preferably derived from prior knowledge acquired from exploratory and/or survey data (e.g. seismic and test drilling data). The exploratory and/or survey seam characterising data is preferably integrated into the seam model with co-ordinate position data (this may be absolute co ordinate position data) associated this seam characterising data.

The processor may use hierarchical relationships between sensors (i.e. one sensor has a subservient relationship to another) or between a sensor and prior acquired knowledge (e.g. survey or exploration data) to enable the processor to preferentially weight output data from one source over another. For example, a sub model of the seam may incorporate survey and/or exploration data. The sensor data may have a subservient relationship with this sub model, if this sub model data is of a higher quality (e.g. increased accuracy and/or precision according to the associated sensor confidence profile). As the spatial sampling density of exploration or survey data is likely to be less than the sensors, the survey or exploration data forming the sub model may be used to calibrate the sensors at positional points where the exploration/survey data are in close proximity to the data from the sensors.

In one embodiment, the calibration of a first sensor is performed relative to a second sensor or other validation source. Preferably, the first sensor and the second sensor have a hierarchical arrangement, with the first sensor subservient or subordinate to the second sensor. The first sensor may comprise a marker band sensor (e.g. IR sensor) to detect a marker band and/or a seam boundary and the second sensor comprises a mining machine power output sensor to measure the power or derivative thereof of the extraction device. The hierarchical arrangement is preferably defined by the respective sensor confidence profiles.

The seam characterising sensor data may be a combination of data from multiple sensors of the same or different type. In addition, the seam characterising sensor data may comprise one or more measurements of the same seam surface or volume taken by the same sensor type at different points in time (e.g. during different traversals across the seam). In a preferred embodiment, the seam characterising sensor data is used (via the seam model) to determine a control system target value-for the mining machine, or control the mining machine itself, in real or near real time. In this embodiment, the seam model derived from prior knowledge and previously acquired sensor data is further validated and/or calibrated with real time data received by the processor. The processor preferably analyses the seam model and produces a real or near real time response through indirect or direct control of one or more mining machine parameters.

In one embodiment the seam model comprises at least one, preferably at least two and more preferably at least three seam sub models, with each sub model derived from a separate seam characterising data source and/or a separate co-ordinate position data source (e.g. a sub model derived from prior knowledge; a sub model derived from acquired seam characterising data; and a sub model derived from real time seam characterising data). Where there are multiple seam characterising sensors, there may be multiple sub models, each associated with one or more seam characterising sensors and/or at least 2D co-ordinate position determining device(s).

In some embodiments, an extrapolation of the cut model is used to correct or calibrate the output data of a sensor. This is particularly preferred when there is a higher degree of uncertainty in the seam model (i.e. seam model has lower confidence value).

In one embodiment, the processor determines that the seam characterising data relates to a calibration target and calibrates the one or more sensors based on a stored sensor reading for that calibration target. The calibration target may be located in relation to a registration marker. This way, seam samples of a known material and with a known characteristic can be placed at the gateroads or equivalent reference points. Since the expected sensor reading is known and stored, processor can subtract any differences from the actual reading, for example, to calibrate the sensor.

Validation of the seam model

The validation and calibration of the seam model is preferably performed by reference to the machine characterising data. While the seam model, like individual sensors, may be validated and calibrated via other sensors or survey data, in a preferred embodiment, the seam model is validated and calibrated by reference to the machine characterising data.

The use of the mining machine characterising data to validate the accuracy and precision of the seam model is preferably achieved through a closed loop calibration. This method is able to improve mining machine performance whilst reducing the need and/or frequency of open loop calibration (i.e. external calibration references) of the sensors. Improved correlation between the cut model (i.e. material from the seam model which has been mined) and associated machine performance data results in increased confidence in the seam model about to be mined.

Preferably, a confidence value, or a derivation thereof, of the machine characterising data and the seam characterising data is used as a feed forward input into seam characterising model about to be mined. The adjustment of the seam model may be to one or more of the seam characterising data sources (e.g. prior knowledge, acquired sensor data and real time sensor data). In one embodiment, the feed forward adjustment adjusts the co-ordinate position data and/or the seam characterising data from the model, or each sub model, such that the error between the different data sources is reduced.

The confidence value or profile of the seam model is preferably transformed to a confidence value or profile of the cut model after the mining machine extracts material from the seam model. The transformation of the cut model confidence profile or value is preferably derived from correlating the mining machine characterising data, which preferably forms part of the cut model, with the seam characterising data or seam model.

Preferably, the confidence value of the cut model is increased relative to the confidence value of the seam model, if the machine characterisation/performance data is better than predicted by the seam model.

Preferably, the processor determines a confidence value associated with the seam model and/or cut model. Preferably, the confidence value comprises statistical uncertainty measures which may be based on a Bayesian Network and/or Bayes Filter.

Controlling the mining machine

The control of the mining machine is achieved through analysis of the seam and/or cut model. This analysis is preferably utilises probability analysis.

The seam model and the mining machine are preferably defined in terms of a confidence profile which defines the uncertainty relating to the efficiency and effectiveness of the mining machine in extracting material from the seam.

The processor preferably comprises a decision making model. The decision making model preferably comprises a decision making risk profile which is preferably used to match the appropriate level of automation of the mining machine to the risk profile of a mining activity.

The decision making model preferably uses the confidence profile of the seam model to determine the level of mining machine automation, such that the decision making model only directly controls one or more parameters of the mining machine when a decision making risk profile is within predetermined limits to ensure mine safety standards are not only maintained butexceeded.

The decision making model preferable utilises machine learning. Preferably processor utilises machine learning..

The processor, through using the decision making model is based on the confidence value of a mining machine activity risk profile and the cut and/or seam model.

Preferably, the mining machine activity risk profile comprises a plurality of mining machine activity confidence values, each confidence value corresponding to a different class of mining machine activity risk profile. For example, the risk profile may be rated upon risk to human life, with mining machine activities which have inherent safety risks requiring increased confidence values.

The decision making risk profile preferably sets out a framework which matches the appropriate level of mining machine automation confidence value with the mining machine activity being controlled, which can range from a manual control mode to a fully autonomous control mode. Preferably, the confidence profile of the seam model is categorised by a first confidence value which, if satisfied, enables the processor to automatically control predetermined activities of the mining machine; and a second confidence value which, if satisfied, enables the processor to provide alerts to the mining machine operators for predetermined activities of the mining machine, wherein the first confidence value is higher (i.e. greater confidence - lower uncertainty) than the second confidence value.

Dependent upon the mining activity risk profile, the processor may trigger an action response plan (TARP) and raise one or more of multiple triggers to initiate or recommend predefined responses, such as machine maintenance, manual/autonomous control or alarms. Preferably the TARP initiates or recommends a plurality of automation levels corresponding to mining activity risk profile and the seam model and mining machine confidence profile.

Mining activities comprising multiple cuts

Removing material from the seam preferably comprises making sequential cuts (i.e., resource extraction) and advancing into the seam after each cut. Preferably, the processor receives the output data signals from the one or more sensors at each of the sequential cuts and determines a correspondence between the output data signals at a first cut and the output signals at a subsequent cut. Preferably, determining the correspondence comprises generating the seam model and/or cut model based on the output data signals at the first cut, determining expected seam characterising data based on the seam model and/or cut model and comparing the expected seam characterising data to the output data signals at the subsequent cut. Preferably, processor determines the presence of sensor inaccuracy and/or imprecision when processor detects a conflict between the expected seam characterising data from one or more validation sources and the output data signals.

Preferably, the one or more sensors comprise a marker band sensor, such as an IR sensor, to collect characterising data indicative of a location of a marker band in the seam. Preferably, generating the seam model and/or cut model comprises determining a coordinate position of the marker band and storing the coordinate position of the marker band in the seam model and/or cut model such that the seam model and/or cut model represents the coordinate position of the marker band.

Preferably, the IR sensor is configured to capture an IR image comprising at least part of the extraction device and the processor determines a location of the extraction device in the IR image. Preferably, the processor determines the presence of sensor inaccuracy and/or imprecision when processor detects a conflict between the location of the extraction device in the IR image and the location of the extraction device provided by machine control data. Preferably, the processor calibrates the IR sensor output data based on the difference between the location of the extraction device in the IR image and the location of the extraction device provided by machine performance data. Preferably, the processor determines the presence of sensor inaccuracy and/or imprecision when processor detects a conflict or potential conflict between one or more of: - expected thermal condition (output data of IR sensor) of the extraction device based on the seam model and/or cut model; - expected thermal condition of the extraction device based on the collected seam characterising data from the one or more sensors; and - the thermal condition of the extraction device based on the IR image of the part of the extraction device.

Preferably, the one or more sensor comprises a GPR sensor to collect characterising data indicative of remaining material above and/or below the seam that has not been extracted (i.e. uncut). Preferably, the processor determines the presence of sensor inaccuracy and/or imprecision when processor detects a conflict between expected seam characterising data based on the seam characterising data from the GPR sensor and the seam characterising data from the IR sensor. Preferably, the processor determines a confidence value of the seam characterising data from the GPR sensor for each of the sequential cuts and determines the absence of sensor inaccuracy and/or imprecision based on the confidence value. Preferably, in the absence of sensor inaccuracy and/or imprecision the processor controls the one or more parameters of said mining machine to adjust, if required, a mining machine parameter (e.g. extraction device position) towards a target value.

The one or more sensors preferably comprise a sensor that collects characterising data of the seam in a direction into the seam including measurements along and perpendicular to the seam boundary. Preferably, the sensor that collects characterising data of the seam in a direction into the seam collects characterising data of the seam behind the surface of the seam to be mined. Preferably, the sensor is a ground penetrating radar (GPR). Preferably, the characterising data collected by the ground penetrating radar comprises multiple time values of reflections within the seam wherein the processor generates the seam model comprising one or more model features that are based on the time values. Preferably, the one or more model features comprise layers of the seam. The GPR preferably collects characterising data above and/or below the machine to be incorporated into the cut model of the seam.

The one or more sensors preferably comprise a sensor that collects the characterising data in a sensing direction and an angle sensor to determine the sensing direction, wherein generating the seam model and/or cut model is based on the sensing direction.

The one or more sensors preferably comprise a sensor that collects the characterising data at a sensing distance and a range finder to determine the sensing distance, wherein generating the seam model and/or cut model is based on the sensing distance.

Preferably, generating the seam model and/or cut model is based on triangulation and/or trilateration of the sensing direction, the sensing distance and the characterising data.

In a third aspect of the present invention there is provided a process for controlling a mining machine of the first and/or aspects of the present invention including the steps of:

A. moving the mining machine relative to the seam; B. generating current co-ordinate position output data signals from the position determining device indicative of the co-ordinate position of the mining machine; C. generating current seam characterising output data signals from the one or more sensors, distinct from the position determining device; D. the processor receiving said output data signals from steps B & C to thereby generate an updated seam model comprising the spatial position of the seam and one or more characteristics of the seam attached or offset therefrom, with the spatial position of the seam and the co-ordinate position of the characteristics of the seam determined by reference to the co-ordinate position in space of the mining machine; E. the processor using seam model and/or cut model data to anticipate required changes to one or more mining machine parameters; F. the processor sending an output signal to adjust one or more of the mining machine settings, parameters and/or providing an alert to an operator to signal what machine settings or parameters should be monitored or changed.

The seam is preferably a coal seam, although it would be appreciated that the process may apply to other seams, including mineral seams.

The processor also preferably receives output data signals from the mining machine relating to one or more settings or performance parameters of the mining machine, said settings or parameters forming part of the cut model, with said cut model comprising the spatial position; one or more characteristics of the seam that has been extracted; and/or one or more characteristics of the mining machine performance characteristics as said mining machine extracted the material from the seam, said seam and mining machine characteristics attached or offset from the spatial position within the cut model.

In a preferred embodiment the volume of material to be mined from the seam (Mvs) is characterised according to the following equation:

M v = f [X (a, b, c...) li, iii.., i, Y (a, b, c...)i, ii, iii.., Z(a, b, c...) i,l, iii..,

where Mv is defined by the 3D spatial matrix X(a, b, c...); Y(a, b, c...); Z(a, b, c...)

a, b, c... zero;

one or more spatial positions are tagged with at least one piece of information (i, ii, iii...) relating to the seam characteristic; if the seam characteristic is 2D surface information of the seam, then the tagged information is 2D unless the information is extrapolated from the one of more of the corresponding cut model material volume previously mined (Mvs -1, Mvs -2, Mvs -3..),

where Mvs is geometrically derivable from the spatial coordinates of the at least 2D position determining device f(X, Y, Z).

Once the volume of material has been mined, the volume of material that was actually mined from the seam model is converted to the cut model (i.e. Mvs -* Mvc). The transformation of seam to the cut model may include a correction of the expected volume to be mined with the actual volume mined as determined from the calculation of the spatial coordinates of the newly exposed seam by the position determining device.

Mvc = Mv + spatial correction from the at least 2D position determining device and preferably tagged information relating to the mining machine's settings and/or performance.

The material volume from the cut model is preferably tagged with the seam characterising information provided from the corresponding seam model. Preferably, additionally information relating to the mining machine's settings and/or performance are tagged to the material volume from the cut model. This additional information is preferably utilised by the processor in anticipating and subsequently controlling the operations of the mining machine, rail or roof supports.

In a fourth aspect of the present invention, there is provided the use of a seam and/or cut model to control one or more parameters of said mining machine based upon analysis of the seam model and/or cut model.

The cut and/or seam model is preferably generated by a processor through receiving output data signals from an at least 2D position determining device and one or more seam characterising sensors.

The co-ordinate position of the seam characterising data is preferably determined by reference to an at least 2D co-ordinate position in space of the mining machine, which is preferably an at least 2D absolute co-ordinate position in space of the mining machine.

The processor preferably further controls the one or more parameters of said mining machine based upon analysis of the seam model and/or cut model to anticipate changes in mining conditions.

The controls of the one or more parameters of the mining machine preferably occurs as the mining machine progresses along the intended cut profile within the seam model.The cut and/or seam model is preferably three dimensional. The cut and/or seam model preferably comprises at least 10, preferably at least 50, more preferably at least 200 and most preferably at least 1000 spatial position values.

In a fifth aspect of the present invention there is provided software that, when executed by a computer, causes the computer to perform the process or use according to the third or fourth aspects of the present invention.

Definitions

Intended cut profile The predetermined path of the extraction device (e.g. shearer) along or across the current mining face based upon the input of the position determining device (e.g. INS). The intended cut profile is preferably derived from seam model. The intended cut profile preferable extends the length of the mining face for longwall mining applications and at least 10 metres and more preferably at least 50 metres in other applications (e.g. roadway development).

Cut model: The at least 2D (preferably 3D) model or map of the spatial coordinates of the mining seam which has been mined. The model may also include characterising seam data and/or mining machine characterising/performance data tagged (spatially registering a seam characterising datum) to the spatial coordinates or offset therefrom. The acquisition of cut model characterisation data and 3D spatial coordinates may be acquired simultaneously (e.g. through the use of photogrammetric methods) or separately (e.g. through the use of fused sensors). The 3D sensing may be performed using a sensor that can discern the seam from the surrounding rock, discern variance in characteristics of the seam or variance in characteristics of the surrounding rock. Determination of the 3D boundaries (i.e. spatial locations) of the seam/s may provide constraints on the seam geometry model that could be reconciled with other sensing data as well as the geological database (e.g. exploration holes). Updating of the seam model based on this 3D data using geological modelling techniques (e.g. explicit or implicit interpolation based methods) may also be used.

The cut model preferably represents at least 50%, more preferably at least 80% and most preferably 100% of the cut seam. Preferably, the cut model encompasses at least 10 metres (and preferably at least 50 metres) of the cut seam behind the mining machine.

Seam model: The at least 2D (preferably 3D) model or map of the mining seam which has yet to be mined. This model is preferably established initially from exploration data and may be refined by extrapolation of as-mined information from the cut model into the yet uncut seam. The seam data may include characterising seam data derived from exploration data such as rock mass defects and faulting structures, composition, hardness, propensity for cave-in, surface and in-seam borehole drilling data, geophysical logging data and 2D and 3D seismic data as well as topographical data of the seam ahead of mining based on seismic signals using the at least 2D co-ordinate position determining device to accurately locate the absolute position of the seismic source. The seam characteristics may be surface characteristics or may be characteristics of a layer within the seam, which may represent thickness of one or more extraction cycles of the mining machine. The model may include characterising seam data and/or machine characterising/performance data tagged to the spatial coordinates or offset therefrom. The model may also comprise spatial and characterising data relating to the material adjacent to the seam to be mined (e.g. underburden, interburden, and/or over-burden). The characterising data is preferably relevant to determining the seam boundary and/or the stability of the seam as it is being mined by the mining machine. Preferably, the seam model represents at least 10 metres (and preferably at least 50 metres) of the seam to be cut in front (i.e. direction of extraction) of the mining machine.

Machine characterising data: data relating to the operational state and performance of the mining machine including machine setting, including control system target settings, positional setting; system performance indicators such as power output or derivative thereof, machine vibration, temperature, and noise.

Interpolation for the purposes of the present invention means the determination of intermediate spatial position along a cut/seam model by prediction.

Extrapolation for the purposes of the present invention means the determination of spatial position along a cut/seam model past the location measured by the positioning determining device by prediction.

Accuracy for the purposes of the present invention means that there is a systematic error or deviation from the true value that is to be determined. It is possible that this systematic error occurs consistently, which would mean that there is high precision but low accuracy.

Precision for the purposes of the present invention means that there is a random error such that the result is inconsistent. In such cases, the determined values may be largely spread but centred on the true value, which means high accuracy but low precision.

Confidence profile for the purposes of the present invention means a statistically derived numerical value derived from the accuracy and precision values and may include distribution, mean and variance.

Confidence value for the purposes of the present invention means a statistically derived numerical value derived from the accuracy and precision values and may include distribution, mean and variance, which has a designated probability limitation. For example, a confidence value may be a 95% confidence limit, which for a normal probability distribution would equate to a mean value +/- 1.96 standard deviations.

Expected value for the purposes of the present invention means the values which fall within a designated confidence value.

Mining activity risk profile for the purposes of the present invention means the inherent risk of a mining activity in terms of the negative impact of each of the identified risks and the likely probability/frequency of the event. The risk profile may be characterised by an impact probability matrix, with risk mitigation actions implemented according to the impact/probability profile. The risk mitigation actions may include decreasing the level of machine automation as the negative impact increases for the same level of probability of an event causing the impact to occur.

Where appropriate in the specification, the term confidence profile may be used to replace confidence value.

Brief description of the figures

Figure 1 is a schematic diagram of a longwall mining machine cutting a coal seam between a main and tail gateroad in an isometric view.

Figure 2a illustrates a cut model.

Figure 2b illustrates the cut model of Fig. 2a in a machine performance dimension.

Figure 3 is a side elevation of mining machine mining a coal seam.

Figure 4 illustrates a hierarchical decision tree model.

Figure 5a & 5b are schematic diagrams of the shearing head progressing towards the main gateroad (5a) and back towards the tail gateroad (5b), with on-board sensors mapping the spatial co-ordinates of a unique coal seam characteristic.

Figure 6 is a schematic diagram of the control input for the shearer head and rail.

Figure 7 is a schematic diagram showing the mining machine's path through the cut seam (dashed lines) and the coal seam surface forming part of the seam model.

Figure 8 is a schematic diagram illustrating the formation of the 3D cut model.

Detailed description of the preferred embodiments

With reference to Figure 1, there is an underground coal deposit 100,comprising a main gateroad 101 and a tail gateroad 102. The gateroads define the target coal seam 103 to be mined. The mining machine 104 progressively traverses the cutting surface 105.

Cut model and seam model

During exploration before the mining operation, the coal seam 103 is characterised based on a drill hole assay 106. The drill hole assay 106 comprises an top layer of coal 108, a marker band 110 and a bottom layer of coal 112. The marker band 110 may be a different type of coal than the top layer 108 or the bottom layer 110, such as a different composition of chemicals, or may be a layer of clay or other sedimentary material. The marker band 110 may be thinner in proportion to the coal layers 108 and 112 than shown in Fig. 1. A first distance 114 is defined between the upper limit of the marker band 110 and the upper limit of the top coal layer 108. Similarly, a second distance 116 is defined between the bottom limit of the marker band 110 and the bottom limit of the bottom coal layer 112.

In some examples, drill hole assay 106 comprises more than one marker band and in those cases, the first distance 114 is defined as the distance between the top limit of the top marker band and the top limit of the top coal layer. Similarly, the second distance 116 is defined as the distance between the bottom limit of the bottom marker band and the bottom limit of the bottom coal layer 116. For simplicity, the following descriptions assumes a single marker band 110 but is applicable to multiple marker bands as well.

Since coal seams are created by sedimentary processes that deposit material equally across an area, it may be assumed that the first distance 114 and the second distance 116 remains constant throughout a significant portion of the coal seam 103. As a result, an example seam model comprises the two distances and may comprise the number of layers, such as:

Seam model: numberOfLayers = 8 distanceFromTopLayer=1000 / in mm distanceFromBottomLayer=2000 / in mm

In some examples, the layers 110, 112 and 110 are not constant over the entire seam 103. Seam 103 may show gradual changes that may be represented by various interpolation models, such as Gaussian mixture models, to represent distances 114 and 116, respectively, across the spatial extent of the seam 103. Seam 103 may also show abrupt changes, such as fault lines. In these cases the model may be separated into multiple sub models that meet at the fault line.

For example, seam 103 may be explored at various points indicated as solid discs in Fig. 1, such as example exploration point 120. At these exploration points drill hole assays, seismic tests and other exploration techniques may be applied to extract the layer information, which can then be interpolated.

Fig. 1 further shows a top distance 122 and a bottom distance 124 at the cut face where the longwall mining machine 104 is about to mine the seam 103 in the current pass. Since the longwall mining machine 104 is equipped with absolute position sensors, the position of the longwall mining machine 104 in relation to seam 103 can be determined. Based on the seam model and the position of machine 104, the distances 122 and 124 can be predicted accurately. Mining machine 104 may then adjust the vertical movement of the shearing heads based on the distances 122 and 124 to extract the maximum amount of coal without mining the strata above or below.

In one example, the vertical movement of the shearing heads is performed in real-time, which may mean that the vertical movement is adjusted upon determining that the longwall mining machine 104 has passed one of the roof supports.

Sensors 126 for determining a coal seam characteristic may be mounted upon the mining machine 104 and/or mounted upon one or more of the roof supports 110. In one embodiment the sensors are mounted on both the mining machine and at least a portion of the roof supports. Within this embodiment, the roof support sensors in front of the advancing mining machine provide characteristics of the coal seam about to be cut. The characteristics may include changes in the composition of the seam, including the grade of the coal or the transition from coal to an inorganic layer. The relative position of the coal characteristics determined by the sensor and the at least 2D determining device 100 may be determined using geometric calculations. This enables the spatial position of the coal characteristics to be accurately known relative to the mining machine. As a result, the mining machine may be able to adjust setting or operating conditions in anticipation in the change in the coal characteristic. For example, a shearer head actuator may be activated to move the shearer head up to follow the change in the coal seam boundary.

For example, sensor 126 may be an infrared camera and a processor 128 mounted on mining machine 104, receives the images and detects the layer structure such as by applying an edge finding algorithm, such as Sobel. The processor 128 may then count the number of pixels at the central pixel column from the bottom row to the lowest edge and from the top row to the top edge. Processor 128 can then convert the pixel numbers into vertical distances based on focal length and chip size. These measurements then determine the top distance 130 and bottom distance 132 of the face that has been mined by machine 104.

A rail 134 guides mining machine 104 along the seam 103 and multiple roof supports 136 move the rail 134 forward as the machine 104 passes.

During each pass of machine 104 across the seam, the sensor 126 determines coal seam characteristics at multiple points along seam 103. Processor 128 can then determine a cut model based on this data.

In one example, machine 104 comprises a laser range finder that measures the distance of the marker band from the sensor 126. This way, based on an absolute position of the machine 104 and the direction of the laser rangefinder, processor 128 can calculate the absolute position of the face imaged by sensor 126.

Fig. 2a illustrates a cut model 200 in relation to the layers 108, 110 and 112 as described above. In this example, the sensor 126 captures eleven measurements along and each measurement is associated with an index along index axis 202. Processor 128 may build the cut model according to the following pseudo-code segment:

cutModel = array( vari=1 while shearingCurrentPass: cutModel.append([i, measuredDistFromTopLayer, measuredDistFromBottomLayer]) i++

end

As can be seen from cut model 200, the seam 103 dips downwardly and is located deeper on the right hand side 204 than on the left hand side 206. However, the machine 104 passed seam 103 along a horizontal plane. As a result, there is a risk that at the right hand side 204 of the seam the machine 204 has mined the roof and wasted coal at the floor. Processor 128 calculates a difference between a top distance 208 of the cut model and a top distance 122 of the seam model. Similarly, processor 128 calculates a difference between a bottom distance 210 of the cut model and a bottom distance 124 of the seam model. Processor 128 then determines a correction of the shearing heads to compensate for the difference.

It is noted that cut model 200 in Fig. 2a may correspond to a part of the mined face. For example, cut model 200 corresponds to the part in Fig. 1 that is shown to be mined during the current pass up to the position of sensor 128 in Fig. 1. In other words, sensor measurement of index=11 is the most recent measurement of sensor 128 at its current position shown in Fig. 1. Based on the seam model and the cut model 200, processor 128 can predict the seam characteristics that lie ahead. As a result, processor 128 can correct the maximum and minimum height of the shearing heads to adapt to the changes in the seam 103. In this case, processor 128 can pre-emptively adjust the shearing head motion to become gradually lower to accommodate the dipping seam 103. As a result, less rock and more coal are mined, which leads to more profitable overall operation.

In one example, processor 128 determines a two-dimensional cut model 200 during each pass and combines the multiple two-dimensional cut models into a single three-dimensional seam model.

In one example, the seam characterising data from each of the different sensor is aggregated into a matrix data structure, where each sensor reading is associated with an X, Y and Z coordinate, such as [(1000, 3000, 2000, 255)] for one IR measurement at location X=1,000mm, Y=3,000mm, Z=2,000mm from a fixed global reference and the pixel value is 255, which corresponds to saturated white in the example of 8-bit resolution. The matrix entry may also comprise further sensor data captured at the same location, such as reflection time '16' for 16ns: [(1000, 3000, 2000, 255, 16)]. Where sensors capture data at locations that are not aligned between the sensors, sensor calibration may be performed as described further below. Sensor calibration may be combined with interpolation to calculate missing values. In one example, the sensor data matrix relates to a grid of constant distances, such that the points of the grid are spaced apart by 100mm in each direction, for example. The sensors are then calibrated against each other to provide sensor data at these canonical grid coordinates. The sensor data stored in that matrix may be the direct sensor reading, such as pixel values or delay of reflections, or may be derived material properties, such as sulphur contents, hardness, electrical permeability, etc.

The data may be represented as characterised by the following equation:

Mv = f [X (a, b, c...) i,li, iii.., Y(a, b, c...)i, ii, iii.., Z(a, b, c...) i,ii, iii..,]

where i, ii, iii relates to the different sensors, and a,b,c relates to the different points in space.

Fig. 2b illustrates a machine performance dimension 250 of cut model 200. In this example, machine 104 measures the rotational speed of the shearing head or the electrical power supplied to the shearing head to maintain constant speed. Light areas of Fig. 2b indicate areas of the face of the seam where the rotational speed was relatively high or power consumption was relatively low. Dark areas indicate areas where the rotational speed was relatively low or power consumption was relatively high.

When Fig. 2b is considered together with Fig. 2a it can be seen that as the seam 103 dips and the shearing head progressively cuts into the roof strata, the corresponding area towards the top right corner of Fig. 2b gets darker, which indicates harder rock. Therefore, processor 128 can consider the performance data of Fig. 2b as a further sensor input and interpret the performance data as relative characterising data of the seam from which material is about to be extracted. As a result, processor 128 can generate the cut model 200 without using sensor 126 but based on only the performance data. Alternatively, processor 128 can combine the measurements from IR sensor 126 and the performance data to obtain a more accurate result than with only a single sensor.

In order to correlate the current position of machine 104 with the coordinates of seam 103, machine 104 may comprise an at least 2D position determining device 140. Processor 128 may update the seam model based on the absolute coordinates provided by the at least 2D position determining device 140 and the measured performance data. The at least 2D position determining device 140 is preferably a 3D position determining device which is preferably inertial navigation device (INS) which preferably comprises gyroscopes and accelerometers.

A preferred sensor 126 for determining a coal seam characteristic is an IR or thermal sensor. Suitable IR or thermal sensors are described in US8622479 or US8469455 which are incorporated herein by reference. The sensors are typically mounted on the moveable carriage. One or more sensors may be used. Preferably a sensor is located at each end of the moveable carriage. The location(s) of the sensor(s) may vary to ensure that that sufficient quality data can be obtained from the coal seam during its mapping in both the trailing and leading positions. Figures 5a and 5b illustrate the sensor 60 positioned at the trailing end of the moveable carriage 60 from which the sensor can scan a thermal image of the freshly cut seam surface. Upon the return path the sensor 60 is now in the leading position and able to rescan the same surface prior to extraction.

Through mapping the thermal image of the cut surface as the shearing heads 50, 55 traverses along the seam face, the sensor is able to send signals to the processor 65 to form a seam model with additional input from the inertial navigation system (INS) 45. Through knowing the distance between the INS 45 and the sensor 60 and the angle at which the sensor is facing the seam wall, and the alignment of the mining machine relative to the coal seam surface, the thermal imaging and spatial positioning data may be combined to form a

3D seam model. The 3D seam model may be used to detect changes in coal grades and/or the boundary of the coal seam with inorganic matter. The processor may reference the seam model to generate or update an intended cut profile to ensure that the shearer heads are maintained within the coal seam.

Upon the return journey (i.e. second transversal, Figure 5b), the sensor may rescan the cut surface and calibrate the intended cut profile through use of identifying features which could be used to spatially match the cut surface on the first traversal. This operation is achieved through the first and second traversal generating a seam model which at least partially overlaps, with the distinguishing seam features generated by the seam characterising sensors being able to be spatially matched by the processor and, if the error between the spatial positions of the identifying feature being outside a predetermined limit, correcting the absolute coordinate position of the intended cut profile. Thus, the seam characterising sensors may be used to generate one or more reference points against which the mining machine can be self calibrated.

In alternative embodiments, sensors orientated to the roof or floor of the seam to capture characterising data relating to the roof loading and stability which may be similarly be integrated into the seam model and used to provide guidance or control to the roof support system as the shearer progresses along the intended cut profile.

In another example, sensors are mounted on roof supports 136 and collect characterising data of the seam from which material is about to be extracted. The sensors on the roof supports are stationary except during the advance phase where each roof support moves the rail forwards. Therefore, the sensors of the roof supports may collect the characterising data once at the start of the current pass, and a second time after the machine 104 has passed to collect characterising data of the newly created face that is about to be mined during the next pass. This way, the sensors create a side view of the seam 103 in front of the machine 104, which allows more accurate control of the shearing heads.

Using Ground Penetrating Radar

In a further example, machine 104 comprises a ground penetrating radar (GPR) system. A purpose-built wideband (900 MHz) bistatic impulse radar may be used to produce a 1-2 ns pulse, which results in high resolution short-range (100 cm) echo data. The radar system may use T = 500 data points acquired at 30 kHz (12-bit ADC) at a rate of 50 Hz. The radar system may be directed at the roof or the floor to measure the distance from the exposed material surface to a subsurface interface boundary before and/or after the machine 104 has passed. Processor 128 can compute the material thickness byd= ,where d is the material thickness, c is the speed of light, T is the measured two-way travel time and E is the dielectric constant of the medium with an example value of 4.5.

Fig. 3 illustrates the coal seam 103 with marker band 110 together with roof strata 302 and floor strata 304. In this example, the objective is to control shearing head 306 vertically such that a constant layer of coal remains under the roof strata 302 and above the floor strata 304. Mining machine 104 comprises a GPR system 304 as describe above. The GPR system 304 transmits a first radar signal 308 towards the roof strata 302. The first radar signal 308 is partly reflected by a first interface 310 between the air in the mined cavity and the coal seam 103. Processor 128 can determine the first distance of the machine 104 from the first interface 310 based on the travel time of the radar signal. This first distance should be equal to the maximum height of the shearing head 306 and processor 128 stores this first distance. Alternatively, machine 104 may comprise a laser range finder to determine the height of the cavity, to identify the first reflection from the first interface and to determine the first distance.

A second reflection can be measured from a second interface 312 between the coal seam 103 and the roof strata 302. Processor 128 determines a second distance to the second interface and can therefore calculate a distance between the first distance and the second distance in order to determine the thickness of the coal cover covering the roof strata 302.

The processor 128 may have stored the vertical offset of the GPR system 304 from the machine 104 and can therefore determine the distance of the roof strata 302 from the top of the marker band 110. As a result, processor 128 can update the top distance 114 of the seam model.

Alternatively, when the machine 104 passes a fault in the seam 103, the distance to the roof strata 302 changes abruptly. When the location of the fault is known accurately from previous measurements, such as exploration measurements or previous passes, processor 128 can update the location of the machine 104 to be at the location of that fault when the fault is detected from the reflected radar signals. During the next pass, processor 128 can then adjust the height of the shearing head 306 in anticipation of that fault to maintain an acceptable layer of coal under the roof strata 302 before and after the fault.

Similarly, GPR system 304 transmits a second radar signal towards the floor strata 304 where it is reflected by a third interface 316 and a fourth interface 318 which allows processor 128 to perform the above calculations with respect to the floor strata 304.

It is noted that mining machine 104 may be a longwall miner comprising a second shearing head (not shown in Fig. 3) or may be a continuous miner having a shearing head 306 in the form of a cutting drum. In the example of a continuous miner, sensor 126 in Fig. 1 may measure characteristics of the coal seam 103 to both sides of machine 104, that is, into the plane of Fig. 3 and out of the plane of Fig. 3.

Sensor calibration

In some examples, the IR sensor 126, the GPR system 304 and the INS system 140 are spaced apart from one another as they are mounted at different locations of machine 104. As a result, their measurements at one particular point in time do not exactly capture the same features. However, the distance between these sensors can be measured when the sensors are mounted and stored by processor 128 on a data store. This way the sensors can be calibrated against each other. In particular, the data from each sensor may need to be associated with an absolute position with respect to a global reference. The INS system 140 can determine this absolute position and processor 128 can add the distance of the sensors from the INS system 140 to determine the absolute position of the sensors. Further, signals from an odometer can be used to align measurements from different sensors. That is, instead of tagging the sensor data with a capture time, the sensor data can be tagged with odometer readings subject the known offset caused by the distance between the sensors.

In yet another example, processor 128 detects features in the characterising data of the seam, such as by detecting an abrupt change in the radar reflection signal from GPR 304. This abrupt change should also be visible in the IR sensor data 126 when the IR sensor 126 passes the same fault. Therefore, by detecting the same fault in the IR sensor data, processor 128 can spatially align the two data streams and can calibrate the two sensors against each other.

Fig. 4 illustrates a hierarchical decision tree model 400 for controlling machine 104 and in particular to control the upper limit of shearing head 306. The decision tree model 400 may be stored on a data store in the form of a nested if-then-else statement or as a state machine. The decision tree 400 starts at an RPM node 402 which relates to the rotational speed of the shearing head 306. If the rotational speed is low, this means the machine 104 may have reached the upper strata 302 and processor 128 reduces 404 the upper limit of shearing head 306. If the rotational speed is normal, processor 128 proceeds to a GPR node 406, which relates to the thickness of coal under the roof strata 302 as described with reference to Fig. 3. If the coal layer as too thin, processor 128 reduces 404 the upper limit.

If the layer is too thick, processor 128 increases 412 the upper limit. If the layer is within the specifications, processor 128 continues to an IR node 410, which is related to the position of the marker band 110 within the mined face of the seam 103. If the marker band 110 is tending upwardly, the slope is considered rising, which causes processor 128 to increase 412 the upper limit. Conversely, if the marker band 110 is tending downwardly, the slope is considered dipping, which causes processor 128 to reduce 404 the upper limit. The decision tree model may be evaluated periodically, such as once a second, or after mining for a certain distance, such as one metre. Similar decision trees may be formulated for the lower limit.

With reference to Figure 6, the processor receives data relating to the absolute spatial positioning of the mining machine as well as data relating to the seam characteristics. The processor integrates the newly-acquired seam characteristics and spatial positioning from the current cut with the existing exploration seam data to form an updated seam model from which an intended cut profile is produced. The seam model and cut model are stored in a memory device, with the processor analysing the seam model and/or cut model in anticipating changes to mining conditions that the mining machine is about to experience and adjusting the controls (including raising alerts) in anticipation of these changes. One or more of the rail moving, roof support and shearer head moving controls are used to guide the mining machine along the intended cut profile. The anticipated changes in the seam conditions may also result in changes to the mining machine operating parameters including roof support pressures and shearer traverse speed. Upon the seam being extracted the seam model is preferably transformed into a cut model which includes the seam characteristics and spatial positioning in addition to data relating to the mining machine performance of the seam which has been extracted. Through analysing the actual interaction of the mining machine with the seam, with the predicted interaction, the seam model may be adjusted or calibrated.

In reference to horizon control in which the shear head is controlled to navigate between defined coal seam boundaries, the intended cut profile may formed by reference to the cut model, the seam model or a combination of both. As illustrated in Figure 7, there is a seam boundary 200 as detected by a thermal IR sensor. The mining machine has made several traversals 210 across the seam. In doing so the intended cut profile has navigated around a discontinuity 220 in the seam boundary. While the current intended cut profile may navigate around the discontinuity in the seam boundary through use of just the seam model, the problem with this approach is that the seam model may only provide a 2D image of the seam boundary, which may result in the 3D portion of seam removed crossing the seam boundary.

To provide a prediction of the location of the seam boundary in 3D (i.e. going into the seam), then there are two possible solutions. Firstly, the sensor could be replaced or supplemented with a sensor which is able to provide data into the seam (e.g. ground penetrating radar may be used to assess the location of the boundary). Secondly, the location of the discontinuity in the 3D cut model may be used to extrapolate the position of the seam boundary into the seam.

When a discontinuity 230 of the seam boundary is detected along the gateroad boundary, interpolation may be used to provide a 3D seam model from which the intended cut profile may be determined.

The formation of a 3D seam model and 3D cut model may be illustrated with the aid of Figure 8. Within a preferred embodiment, an inertial navigational system is mounted on the mining machine and providing output data signals relating to the current 3D spatial position of the mining machine which are received by the processor. Output data signals are also provided by one or more sensors relating to the seam characteristics. The output data signals relating to the seam data may comprise a 2D matrix of output data corresponding to a 2D portion of the seam surface. The sensors may also produce a 3D matrix of output data, e.g. when the sensor is a ground penetrating radar. Each output data signal comprising seam characterising data is tagged to the 3D position of the mining machine which corresponds to when and where the seam characteristics are relative to the mining machine. Geometric calculations are performed to calibrate the 3D position of the 2D matrix of characterising data relative to the position of the mining machine. Preferably an odometer output signal (not shown) is also provided as input to the processor, such that any spatial offset between the mining machine INS location and the seam characteristics can be accounted for, such that the processor can develop a 3D seam model from which the processors can anticipate required changes to the mining machine and either activate these changes directly through sending a signal to an actuator or indirectly through sending a signal to a control alert to alert the operator to monitor and consider changing a particular control.

As the mining machine extracts product from the seam machining machine setting and performance output data signals may also be input into the processor, with the output tagged to spatial location of the mining machine and the seam characteristics of what was extracted. The processor collates this information into a 3D cut model (memory storage device not shown). The 3D cut model may be used by the processors to anticipate which machine settings may be required when assessing the current 3D seam model. Learnings from historical seam and machine data (e.g. 3D cut model) is preferably used to optimise future operational settings of the mining machine. The processor preferably comprises a learning algorithm to enable the mining machine to incorporate the cut model and seam model analysis in expected and actual results from past operational settings.

In one example, the cut model and/or seam model comprise a confidence value. The confidence value may be a single value for the entire model or may comprise one value for each point of the model, such as each point of a grid or each support point of the model. The confidence value may be low at the beginning of the mining process when measurements of the seam are limited. The confidence value may also decrease with the distance from a measurement. For example, the confidence value at the points where drill hole assays 120 are available are high, such as 0.9, while the confidence value at the middle between drill hole assays 120 is low, such as 0.1.

The acquisition of machine performance data, IR image data, GPR data and other seam characterising data from the sensors described herein increases the confidence value at points where the data is measured. In particular, if the sensor data confirms the current seam model, that is, the sensor data is identical to what is predicted by the seam model, the confidence value increases. Processor 128 may calculate the update confidence value and store the updated value associated with the seam model.

In another example, processor 128 calculates the confidence score c according to c=1-exp( - ( 1 / (al*dl+a2*d2+a3*d3), where the dl, d2 and d3 are the differences between the sensor measurements of three sensors and the seam model, respectively and al, a2 and a3 are weights for the respective sensors. This way, if all three sensors, such as IR camera, GPR and power consumption result in the same top distance 122, the confidence value would be c = 1. The weights may be indicative of a distance of the point associated with the confidence score to the location of the measurement.

Fig. 9 illustrates a computer system 900 for controlling one or more parameters of machine 104. The computer system 900 comprises a processor 902, corresponding to processor 128 in Fig. 1, connected to a program memory 904, a data memory 906, a communication port 908 and a user port 910. The communication port may be a CAN bus interface and is connected to GPR system 304, IR sensor 126 and laser range finder 911.

The program memory 904 is a non-transitory computer readable medium, such as a hard drive, a solid state disk or CD-ROM. Software, that is, an executable program stored on program memory 904 causes the processor 902 to perform the method in Fig. 10, that is, receives data, generates the seam and/or cut model and generates the output data signals to control one or more parameters of machine 104. The term "determining a model" refers to calculating one or more values that are indicative of the model. This also applies to related terms.

The processor 902 may then store the model values on data store 906, such as on RAM or a processor register. Processor 902 may also send the determined model values via communication port 908 to a server, such as a mine control server.

The processor 902 may receive data, such as seam characterising data, from data memory 906 as well as from the communications port 908 and the user port 910, which is connected to a display 912 that shows a visual representation 914 of the seam and/or cut model to a user 916, such as a machine operator. In one example, the processor 902 receives seam characterising data from sensors 304, 126 and 911 via communications port 908, such as by using a Wi-Fi network according to IEEE 802.11. The Wi-Fi network may be a decentralised ad-hoc network, such that no dedicated management infrastructure, such as a router, is required or a centralised network with a router or access point managing the network.

In one example, the processor 902 receives and processes the seam characterising data in real time. This means that the processor 902 determines the seam and/or cut model every time seam characterising data is received from one of the sensors 304, 126, 91land completes this calculation before the sensors send the next update.

Although communications port 908 and user port 910 are shown as distinct entities, it is to be understood that any kind of data port may be used to receive data, such as a network connection, a memory interface, a pin of the chip package of processor 902, or logical ports, such as IP sockets or parameters of functions stored on program memory 904 and executed by processor 902. These parameters may be stored on data memory 906 and may be handled by-value or by-reference, that is, as a pointer, in the source code.

The processor 902 may receive data through all these interfaces, which includes memory access of volatile memory, such as cache or RAM, or non-volatile memory, such as an optical disk drive, hard disk drive, storage server or cloud storage. The computer system 900 may further be implemented within a cloud computing environment, such as a managed group of interconnected servers hosting a dynamic number of virtual machines.

It is to be understood that any receiving step may be preceded by the processor 902 determining or computing the data that is later received. For example, the processor 902 determines seam characterising data, such as by pre-filtering the raw sensor data, and stores the seam characterising data in data memory 906, such as RAM or a processor register. The processor 902 then requests the data from the data memory 906, such as by providing a read signal together with a memory address. The data memory 906 provides the data as a voltage signal on a physical bit line and the processor 902 receives the seam characterising data via a memory interface.

It is to be understood that throughout this disclosure unless stated otherwise, nodes, edges, graphs, solutions, variables, cut models, seam models and the like refer to data structures, which are physically stored on data memory 906 or processed by processor 902. Further, for the sake of brevity when reference is made to particular variable names, such as "cut model" or "seam model" this is to be understood to refer to values of variables stored as physical data in computer system 900.

Fig. 10 illustrates a computer implemented method 1000 as performed by processor 802 for controlling one or more parameters of mining machine 104. Fig. 10 is to be understood as a blueprint for the software program and may be implemented step-by-step, such that each step in Fig. 10 is represented by a function in a programming language, such as C++ or Java. The resulting source code is then compiled and stored as computer executable instructions on program memory 904.

Method 1000 commences by processor 802 receiving 1002 output data signals from position determining device 140. Processor 802 further receives 1004 the output data signals from the sensors, such as GPR system 304, IR camera 126 or laser range finder 811. Processor 1006 then generates the seam and/or cut model as described with reference to Figs. 1, 2a and 2b. The generated seam model and/or cut model comprises the spatial position and one or more characteristics of the coal seam attached or offset therefrom.

Based on the seam model and/or cut model data processor 802 anticipates the required changes to one or more mining machine parameters and generates 1008 output data signal to control one or more parameters of the mining machine 104. For example, processor 802 generates signals to control the vertical actuators of the shearing head 306 to maintain a constant coal cover over roof and floor as described above.

Processor 802 may send the output signal to adjust one or more of the mining machine settings, parameters and/or providing an alert to an operator to signal what machine settings or parameters should be monitored or changed.

It will be appreciated by persons skilled in the art that numerous variations and/or modifications may be made to the specific embodiments without departing from the scope as defined in the claims.

The following description provides further details on addressing inaccuracy and/or imprecision. These examples describe a longwall miner performing multiple passes and advancing into the seam after each pass. These passes may also be referred to as sequential cuts, which also applies to continuous miners or roadheaders that move the cutting head up and down to make sequential cuts. Continuous miners/roadheaders also advance into the seam between sequential cuts.

Whilst the following example illustrates the use of the invention to control the vertical positioning of the mining machine with respect to the seam (e.g. horizon control), it will be understood by the person skilled in the art that the invention may also be used for face alignment (lateral positioning) and creep control (longitudinal positioning).

Before mining machine 104 starts the extraction process, that is, before the commencement of the very first pass, processor 128 has little information available to control mining machine 104 (i.e. seam model has a relatively high degree of uncertainty). Therefore, the number of first passes, such as the first five passes, may be operated manually as the available data does not allow safe autonomous operation.

As mining machine 104 traverses the seam during the first pass, the IR sensor 304 collects IR images of the marker band 110. Processor 128 determines the position of the marker band 110 in the IR image. Processor 128 then determines the absolute coordinate position of the marker band based on the absolute coordinate position data from the INS system 140 and the spatial relationship between the IR sensor 126 and INS system 140. Processor 128 further stores the absolute coordinate position of the marker band within the IR image for each location along longitudinal axis 202 as the cut model.

When mining machine 104 traverses the seam during the second pass in the opposite direction, processor 128 can retrieve the corresponding position of marker band 110 from memory and compare the stored position to the position currently measured during the second pass. That is, processor 128 extrapolates the developed cut model to determine an expected position of marker band 110. Ideally, the expected position and the current position should be identical, or within a predetermined tolerance band, as the aim of most extraction operations is to keep the marker band at a constant level. If that is the case, that is, the expected stored position of marker band 110 is equal to the currently measured position, processor 128 determines that the collecting of characterising data of the seam is functioning properly with sufficient accuracy and may switch to autonomous operation. In some examples, it may take several passes for the measurements to settle. The minimum number of passes before commencing autonomous operation may be predetermined, such as five, in order to detect any changes in the position of the marker band 110 over multiple passes, such as in a dipping seam.

If the measured position of marker band 110 does not coincide with the expected position, processor 128 determines that the sensors maybe not working properly. For example, the IR sensor 126 may be pointing in the wrong direction or may be covered or otherwise disturbed. Processor 128 may generate a notification to the operator, such as by raising an alarm signal.

Before the first pass, the seam model may comprise exploration data from drill hole assays. In that case, processor 128 may determine a difference between the position of the marker band 110 in the IR images and the expected position of the marker band from the drill hole assays. Processor 128 may then suggest to the operator a change to follow the marker band as planned and/or adjust the seam model towards the observed position.

In these examples, processor 128 controls mining machine 104 based on sensor data from the IR sensor 104 and the seam model. At the current position of the mining machine, the seam model provides an expected seam characteristic. For example, processor 128 determines from the seam model, such as by interpolation, an expected top distance 114 and bottom distance 116 from marker band 110. In other words, processor 128 infers the seam boundaries and assumes the marker band is at a constant distance from the seam boundaries.

The IR sensor 126 provides a measurement of the top distance 114 and the bottom distance 116 as described above. The model value and the measurement are linked by the absolute coordinate position provided by the INS system 140. When the mining machine 104 passes a position with a high confidence score of the seam model, such as a position where a drill hole assay was extracted, the seam model can be used to calibrate the IR sensor. Vice versa, at positions between drill hole assays, the seam characterising data from the IR sensor 126 can be used to updated the seam model with a higher confidence score.

Further, at positions where the confidence score of the seam model is high, the difference between the seam model and the IR sensor data can be used to determine whether the IR sensor is functioning properly. For example, processor 128 may control mining machine 104 autonomously as long as the difference between the IR sensor data and the seam model is within a predefined threshold, such as 0.1m. Otherwise, processor 128 switches to manual operation or automatically stops mining machine 104 until an operator manually resumes operation.

In yet another example, the IR sensor 126 is configured such that the field of view overlaps with possible positions of the shearing heads. As a result, processor 128 can determine from the IR sensor data the position of the shearing heads or when the position is available in the form of machine performance data (i.e. positional settings of the extraction device) the processor 128 can determine the accuracy of the IR sensor 126.

Processor 128 may further monitor the control input by the operator and apply machine learning methods to learn over time how the operator reacts to changing conditions. More generally, processor 128 performs supervised learning where the learning samples include as features the sensor measurements and as labels the operator input. Over time, processor 128 calculates model parameters, such as the factors of a linear regression or self-organising map. To calculate the model parameters processor 128 minimises the difference between the actual operator input in the training samples and the autonomous control output calculated based on the model parameters and the features of the learning samples.

Fig. 11 illustrates an IR image 1100 provided by IR sensor 126, where higher temperatures are shown as black and lower temperature are shown as white. IR image 1100 clearly shows marker band 110 as a line of warmer pixels 1120 since the material of the marker band is harder than the surrounding material. Additionally, IR image 1100 comprises a first circular shape 1104 and a second circular shape 1106. As the shearing heads extract the material, the friction between the material and the shearing heads causes the shearing heads to heat up to a temperature that is higher than the surrounding material, that is, the wall face. As a result, the IR sensor 126 can detect the shearing heads. The first circular shape 1104 is an image of the leading shearing head while the second circular shape 1106 is an image of the trailing shearing head. Processor 126 may perform pattern recognition algorithms and/or edge detection algorithms to determine the location of the circular shapes 1104 and 1106 within the image. Processor 128 may receive machine control data indicative of the position of the shearing heads and may compare the position of the circular shapes 1104 and 1106 to the machine data.

In the example of Fig. 11, the IR image 1100 has 18 lines and processor 128 stores information that for the highest position of the shearing head the bottom of circular shapes 1104 and 1106 should be in line '4' (starting from '1' at the top). Similarly, in the lowest position of the shearing head, the top of circular shapes 1104 and 1106 should be in line '15'. In the example image of Fig. 11 processor 128 determines that the leading shearing head is in the highest position and the trailing shearing head is in the lowest position.

Processor 128 compares this result to the machine control data, which should also indicate these positions of the shearing heads.

If the shearing head positions from the IR image are different to the machine control data, the processor 128 detects an abnormality from historical sensor output. For example, a rock or a piece of equipment may have fallen on the IR sensor 126 and changed the position or direction of the IR sensor 126. In these cases, processor 128 may raise an alarm, abort autonomous operation or may calibrate the IR sensor 126 by subtracting the measured difference. For example, if processor 128 detects the bottom of leading shearing head at line '6' but the machine control data indicates that the shearing head is at the top and should be in line '4', processor 128 subtracts '2'from each line number of the entire image 1100. This would effectively shift the marker band 1102 upwardly by two pixels. This way, processor 128 reduces inaccuracies of the IR sensor 126.

Further, the IR sensor 126 may be subject to vibrations that occur faster than the movement of the shearing heads. As a result, processor 128 can shift each IR image such that the circular shapes 1104 and 1106 remain at a constant location or change location by a maximum amount of pixels per second between subsequent IR images. This way, processor 128 reduces imprecision of the IR sensor 126. In yet another example, IR image 1100 may comprise more than two temperature values represented as white and black. Instead, a greyscale image may represent a wide range of different temperatures. By comparing pixel values between subsequent images, processor 128 can detect a change in temperature.

Fig. 12 illustrates another IR image 1200 comprising a marker band 1202, a first circular shape 1204 and a second circular shape 1206. In this example scenario, the mining machine 104 extracted material while the marker band 1202, first circular shape 1204 and the second circular shape 1206 were detected at a moderate temperature depicted as lightly shaded. Suddenly, however, the first circular shape 1204 changes to black, which indicates to the processor 128 that the leading shearing head is heating up rapidly. This occurs when the shearing head advances into rock that is harder than the extracted material causing increased friction and more heating.

Processor 128 can now infer that the distance 108 between the marker band and the overburden at the current position is equal to the height of the shearing head above the marker band 110. In other words, machine control data provides the height of the shearing head in relation to the mining machine 104. The INS system 140 provides the absolute coordinate position of the mining machine, which allows processor 128 to determine the absolute coordinate position of the top of the shearing head. Further, the INS data allows the processor 128 to determine the absolute coordinate position of the marker band 110. Finally, processor 128 can subtract the absolute coordinate position of the shearing head from the absolute coordinate position of the marker band 1202 in the IR image to determine the top distance 114. Processor 128 can then update the seam model with the determined distance 114 at the absolute coordinate position and increase the associated confidence value.

In addition to the IR sensor 126, the GPR sensor 304 may provide further reduction of inaccuracies and or imprecision. As described with reference to Fig. 3, the GPR sensor 304 measures reflection times and processor 128 can then determine the thickness of the seam remaining after extraction, which is represented by the distance between the first interface 310 and the second interface 312 in Fig. 3. The target of the extraction process may be a layer of constant thickness to remain under roof 302 and over floor 304 of 0.5m, for example. A lower target thickness may result in higher profitability but also leads to increased risk of advancing into rock which would result in increased wear and higher cost for replacements.

Reducing inaccuracies in the seam model may allow reducing the target thickness without increasing the risk of advancing into rock. To that end, at the beginning of the mining progress, that is, during the first five passes, for example, the target thickness may be conservative, such as 1m since the confidence value at most locations in the seam is low. As the mining machine 104 progresses into the seam, that is, as the roof supports 136 move forwardly, the GPR sensor 304 moves into the mined cavity and can measure the thickness of the remaining layer at each longitudinal coordinate i in Fig. 2a. Processor 128 can then compare the measured thickness to the target thickness, which may be equal to the predicted thickness.

For example, the seam model predicts a top distance 114 of 2m, bottom distance of 3m. Since the confidence value at this stage is low, processor 128 sets a conservative target thickness of 1m. Therefore, maximum shearing head position is set at 1m above the marker band and 2m below the marker band. As the mining machine 104 advances into the seam (in y direction in Fig. 1) GPR sensor 308 measures a roof thickness of 1.5m, which is different to the roof thickness predicted by the model. Therefore, processor 128 further reduces the confidence value of the model, such as by subtracting 0.1 from the current value for each pass where the predicted thickness differs from the measured thickness by more than 5% or subtracts the error in percent from the confidence value. That is, processor 128 subtracts 0.05 from the confidence value for a 5% error.

In this example, the mining machine 104 progresses and at each pass GPR sensor 304 continually measures the same roof thickness of 1.5m, which means processor 128 determines a high consistency of roof thickness. This allows processor 128 to determine the absence of sensor imprecision and to update the seam model by adding 0.5m to the top distance 114 or by adding 10% of the difference to the seam model. With each pass where the predicted thickness is within 5% of the measured thickness, processor 128 increases the confidence score, such as by adding 0.1. Once the confidence score is above a threshold, such as 0.8, processor 128 controls the parameters of mining machine 104. That is, processor 128 gradually increases the upper limit of the shearing heads, such as by adding 5cm at each pass to gradually reach the target of 1m. As the confidence score further increases, processor 128 may even reduce the target below 1m to a minimum of 0.1m, for example.

Processor 128 may perform a feedback control method comprising a proportional (P) component that adjusts the limits of the shearing heads proportionally to the difference between the target and the measured thickness. The feedback control method may further comprise integral (I) and/or differential (D) components to provide a PID control. It is noted that the delay between extracting, advancing and measuring the thickness can be considered a dead time and included into to the PID control. Using feedback control, processor 128 can adjust the limits of the shearing heads to adapt to changing seam conditions, such as a dipping seam.

In another example, the seam model shows a fault line which results in a downwards step of the top distance 114 with a step height of 1m. In anticipation of that step, processor 128 may linearly adjust the target thickness such that the target thickness just before the fault line is 1m more than the target thickness just after the fault line. For example, the respective target thickness values for the three passes before the fault line are 0.1m, 0.4m and 0.7m.

Further, processor 128 may compare the GPR measurements against the cut model. If there is a significant difference, processor 128 updates the cut model. If there is an insignificant or no difference, processor 128 validates the cut model by increasing the confidence score as described above. Processor 128 may extrapolate the cut model to predict the seam model and to control the limits of the shearing heads. For example, each measurement of the GPR sensor 304 of the roof thickness is 0.05m less than the measurement of the previous pass. This indicates that the seam is dipping and processor 128 incorporates the dipping roof into the cut model. By extrapolating the cut model, processor 128 can automatically reduce the upper limit of the shearing heads at each pass by 0.05m independently or in addition to any corrections from the PID control method. This way, processor 128 does not react to observed differences between target and observed thickness but instead pre-emptively corrects any predicted differences. This reduces the control problems associated with large dead times.

In cases where processor 128 uses an aggressively low target thickness, IR sensor 126 can be useful to detect the advancement into rock strata early as explained with reference to Fig. 12. As soon as processor 128 detects a heating of shearing heads as shown by a darker circular shape 1204, processor 128 reduces the confidence value of the seam model at that position and may switch to manual mode or may increase the target thickness to a more conservative value.

The above examples show that processor 128 may reduce the confidence value of the seam model as soon as processor 128 determines sensor inaccuracy and/or imprecision that leads to a discrepancy between the sensor data and the seam model. However, using multiple sensors there may be a discrepancy between different sensors. For example, IR sensor 126 may detect a heated shearing head while machine data does not show increased power consumption or reduced RPM.

In order to make a decision processor 128 may evaluate a Bayesian Network that reflects expert domain knowledge. For example, mining engineers know that an increase in power consumption or decrease of RPM very likely indicates advancement into rock strata. However, the mining engineers also know that the IR sensor 126 is more sensitive to heating but is easily damaged or moved. Therefore, an indication of a shearer head heating from IR sensor 126 indicates that advancement into rock strata even if the machine data shows no anomalies but a 'cold' IR image does not indicate no heating if the machine data shows increased power/decreased RPM. However, any heating or power/RPM change has only little significance if the GPR sensor 304 measures a sufficient thickness during the next passes. These relationships between measurements can be represented by nodes and the expert domain knowledge is represented by probabilities between these nodes. The output node may then be the decision to increase/decrease the target thickness, update the seam model or increase/decrease the confidence value.

In another example, the output node is whether the system is working properly. In this case, the edges in the Bayesian Network represent correlations between different sensors. If the sensors behave differently, that is, their mutual correlations are different to the correlations in the network, the output node indicates that there is a fault in the system. In that case, processor 128 may switch from autonomous to manual operation or may schedule a maintenance procedure. Processor 128 may have access to a trigger action response plan

(TARP) and raise one or more of multiple triggers to initiate predefined responses, such as machine maintenance, manual/autonomous control, alarms, etc.

Fig. 13 illustrates the mining machine of Fig. 3 with further detail. As described above, processor 128 controls the mining machine 104 and in particular, controls the vertical position of shearing head 306. The vertical position of shearing head 306 plus the radius of the shearing head defines vertical position of the upper cutting surface of shearing head 306. Processor 128 may calculate the absolute coordinate position of the upper cutting surface of shearing head 306 based on the absolute coordinate position data received from the INS system 140 and the machine geometry. The inaccuracies and/or imprecision of the absolute coordinate position data from the INS system 140 and potential further variation of the angle sensors and other machine parameters may also lead to inaccuracies and/or imprecision in the vertical position of the upper cutting surface. Fig. 13 illustrates this inaccuracy and/or imprecision in the form of a first statistical distribution 1302.

Similarly, the seam model is subject to inaccuracies and/or imprecision, which is illustrated in Fig. 13 in the form of a second statistical distribution 1304. The distributions 1302 and 1304 may be part of the seam model and/or the cut model as described below.

Figs. 14a to 14d illustrate different examples of the statistical distributions 1302 and 1304. It is noted that the z-coordinate is directed in the vertical direction as indicated in Fig. 1. Therefore, the distribution on the left hand side relates to the shearing head 306 while the distribution on the right hand side relates to the seam boundary as the seam boundary is above the shearing head 306. It is assumed that these examples relate to a point during the extraction process where the shearing head 306 assumes the maximal vertical (z) position before moving down again.

Fig. 14a illustrates an example where the shearing head distribution 1402 and the seam boundary distribution 1404 are spaced apart from each other without any substantial overlap. This relates to a conservative approach where mining the overburden is avoided as much as possible. However, the remaining coal under the overburden, that is in the roof of the mine is not mined which reduces the profitability of the operation.

Fig. 14b illustrates an example where the shearing head 306 is advanced higher and closer to the seam boundary. As a result, shearing head distribution 1402 and seam boundary distribution 1404 overlap, which is indicated by the black area 1405. Therefore, there is a likelihood that the shearing head 306 will advance into the overburden. This example relates to a more aggressive strategy where the increased wear and cost associated with mining rock is accepted for the higher profit associated with mining a higher volume of coal. It is noted that the maximums of distributions 1402 and 1404 indicate the expected values. Therefore, looking at Fig. 14b, if the shearing head 306 is at the expected position and the seam boundary is at the expected height, no rock is mined. However, if the seam boundary is lower than expected (to the left of the maximum of seam boundary distribution 1404) and at the same time, the shearing head 306 is higher than expected (to the right of the shearing head distribution 1402), the vertical shearing head position may be greater than the seam boundary position, which means rock is mined.

Considering the mining machine is operated according to Fig. 14b and the shearing head 306 is moved up and down many times, it is expected that the shearing heard would hit rock in some instances. Mathematically, the probability of hitting rock at a particular position can be determined through existing seam characterising data such as survey logs and/or noting observations when a rock is encountered (or not encountered) at a particular position during mining process. Bayes Filters provide a convenient a framework for sequentially updating the probability density function associated with the probability of hitting rock.

The probability of encountering rock at a particular position may be expressed as a percentage. For example, if the percentage of hitting rock at a particular position is 10%, and if the shearing head is moved to this position 100 times, then it would be expected that rock would be encountered on average 10 times, which can be detected by the machine performance data or the IR images as described with reference to Fig. 12. If the observed percentage does not agree with the calculated percentage, processor 128 can adjust the distributions 1402 and/or 1404.

Fig. 14c illustrates narrowed distributions 1402 and 1404 after no rock was hit for 100 cuts. As can be seen in the figure, processor 128 has reduced the width of the distributions, which relates to an increased confidence value. If the aim is to maintain a constant probability of hitting rock, processor 128 may now adjust the upper limit of shearing head 306 upwardly to move the distributions 1402 and 1404 closer together and create a small overlap of 10%, for example.

In one example, the confidence value may be indicative of the standard deviation of the respective distributions 1402 and 1404. Processor 128 may then calculate a value based on the standard variation that indicates how broad the distribution is. For example, processor 128 calculates a z value that is shifted from the mean by three times the standard deviation. This value may be referred to as 3a value. In other words, the confidence value of the cut model and/or the seam model is indicative of a distance from the expected value. Fig. 14d illustrates the shearing head distribution 1402 with shearing head 3a value 1406 and seam boundary distribution 1404 with seam boundary 3 value 1408. Processor 128 may calculate the 3u value assuming a Gaussian distribution according to

1 - __/1)2 f(xIp,u-)= e 2a.

Other probability density functions may be similarly applied if they more accurately represent the boundary variation. In the case where the shearing head 3u value 1406 is smaller than the seam boundary 3u value 1408, it can be said that the inaccuracies and/or imprecision is negligible in relation to the distance of the shearing head 306 from the seam boundary. Therefore, in cases where the shearing head 3u value 1406 is smaller than the seam boundary 3uvalue 1408 it can be said that there is an absence of inaccuracies and/or imprecision. Vice versa in cases where the shearing head 3u value 1406 is greater than the seam boundary 3a value 1408 it can be said that there is a presence of inaccuracies and/or imprecision.

The aim of the mining operation may be to leave less than a predetermined thickness of coal, such as less than 1 m of coal, under the seam boundary. At the start of the mining operation, there may be significant overlap of the distributions 1402 and 1404 if there is a distance of 1 m between their respective expected values. This is due to the low confidence value in the seam model and the machine coordinates at the beginning of the operation. Processor 128 calculates the corresponding 3a values 1406 and 1408 and determines that the shearing head 3a value 1406 is greater than the seam boundary 3a value 1408. This may indicate to processor 128 that information is not sufficient for autonomous operation and processor 128 may switch to manual operation or stop the machine. As the confidence values improve, shearing head 3a value 1406 may become smaller than the seam boundary 3a value 1408 at which point processor 128 switches to autonomous operation. Processor 128 may also adjust the maximum height of the shearing head 306 such that the 3a values 1406 and 1408 are equal or at a predetermined distance. Using the 3 values instead of calculating the probability of hitting rock simplifies the calculations without significant loss in controlling ability.

It is noted that detecting that the shearing head 306 advanced into the seam boundary allows processor 128 to update the location of the seam boundary in the cut model, that is, shift the mean of the seam boundary distribution 1404 to the current position of the shearing head 306. If advancement into the seam boundary is not detected, it is not known how far the shearing heard 306 is located from the seam boundary. However, not detecting advancement into the seam boundary confirms that the seam boundary is higher than the shearing head 306, which allow processor 128 to increase the confidence value, that is, increase the 3c value 1408 of seam boundary distribution 1404 while keeping the expected value constant thereby narrowing seam boundary distribution 1404.

For example, if the confidence value of the position of an extraction device (e.g. shearing head or cutting drum) relative to the seam boundary indicates that 90% of the time the cutting drum will stay within the seam boundary (i.e. 10% outside the seam boundary) and the machine characterising data confirms that the extraction device operated within the seam boundary 100% of the time, then the confidence value relating to the position of the extraction device relative to the seam boundary may be increased. Thus, the cut model confidence value may be greater than the confidence value of the intended cut profile within seam model and thus extrapolation of the cut model (or other processing thereof) may be used to improve the confidence value of the seam model. A consequence of an improved confidence value of the seam model is a greater opportunity to operate at a higher level of automated mining, which preferably requires that the confidence value of the intended cut profile relative to the seam boundary is above a predetermined level. Alternatively, or in addition to, the intended cut profile may be modified by reducing the predetermined distance between the extraction device and the seam boundary to increase coal extraction efficiency.

Conversely, if the confidence value of the seam model is such that it predicts that the extraction device will be within the seam boundary 90% of the time, but the machine performance data indicates that the cutting drum is outside of the seam boundary 20% of the time, then the seam and/or cut model confidence values may be correspondingly adjusted to reflect this conflict.

In one embodiment, the distance between the extraction device and the seam boundary is controlled by the predetermined confidence value of the extraction device crossing the seam boundary. The above concept of statistical distributions may equally be applied to the calibration of sensors against validation sources. For example, processor 128 may only calibrate a sensor if the confidence value of the calibration source is above a predetermined threshold. In other words, processor 128 may not calibrate a sensor of high confidence value, that is, a narrow distribution, with a validation source with a low confidence value, that is, a broad distribution even in case of a conflict. In particular, even if the two distributions of sensor and validation are identical, a conflicting measurement may be a result of statistical variations and there should be no calibration. Processor 128 may calibrate the sensor only if the conflicting results differ by more the current 3a values.

Some examples herein relate to an at least 2D co-ordinate position determining device for determining the absolute co-ordinate position in space of the mining machine and for generating current absolute co-ordinate position output data signals. However, other examples may comprise an at least 2D co-ordinate position determining device for determining the relative co-ordinate position in space of the mining machine and for generating current relative co-ordinate position output data signals. For example, the mining machine may be reset at a reference point within the mine and then perform dead reckoning using the inertial sensors to calculate the 2D co-ordinate position of the mining machine relative to the reference point.

It should be understood that the techniques of the present disclosure might be implemented using a variety of technologies. For example, the methods described herein may be implemented by a series of computer executable instructions residing on a suitable computer readable medium. Suitable computer readable media may include volatile (e.g. RAM) and/or non-volatile (e.g. ROM, disk) memory, carrier waves and transmission media. Exemplary carrier waves may take the form of electrical, electromagnetic or optical signals conveying digital data steams along a local network or a publically accessible network such as the internet.

It should also be understood that, unless specifically stated otherwise as apparent from the following discussion, it is appreciated that throughout the description, discussions utilizing terms such as "estimating" or "processing" or "computing" or "calculating", "optimizing" or "determining" or "displaying" or "maximising" or the like, refer to the action and processes of a computer system, or similar electronic computing device, that processes and transforms data represented as physical (electronic) quantities within the computer system's registers and memories into other data similarly represented as physical quantities within the computer system memories or registers or other such information storage, transmission or display devices.

The present embodiments are, therefore, to be considered in all respects as illustrative and not restrictive.

Claims

1. A mining machine including:

A. an extraction device for removing material from a seam;

E. a memory storage device to store the seam model and/or the cut model,

2. The mining machine according to claim 1, wherein the processor controls the one or more parameters of the mining machine through generating an alert for an operator to change or monitor said one or more parameters of the mining machine.

3. The mining machine according to claim 1 or 2, wherein the co-ordinate position of the seam characterising data is determined using geometric techniques and the relative position between the one or more sensors and the at least 2D co-ordinate position determining device.

4. The mining machine according to any one of claims 1 to 3, wherein the cut model comprises the co-ordinate positions of the seam which has been mined and at least one seam characterising datum tagged to at least one position and preferably each position.

5. The mining machine according to any one of the preceding claims, wherein the cut model comprises the co-ordinate positions of the seam which has been mined and at least one mining machine characterising datum tagged to at least one position and preferably each position.

6. The mining machine according to any one of the preceding claims, wherein the one or more sensors are located on one or more roof supports.

7. The mining machine according to any one of the preceding claims, wherein the seam model comprises the co-ordinate positions of the seam which is about to be mined and at least one seam characterising datum tagged to at least one position and preferably each position.

8. The mining machine according to any one of the preceding claims, wherein in the anticipated changes in mining conditions are determined by reference to the cut model, the seam model or a combination thereof.

9. The mining machine according to any one of the preceding claims, wherein the one or more parameters of said mining machine are chosen from the group consisting of roof support pressure; mining machine traversal speed; and extraction device location; current; vibration and/or extract device speed.

10. The mining machine according to any one of the preceding claims wherein: a. the extraction device is a shearer head mounted upon a moveable carriage; and b. the one of more sensors comprises a seam characterising sensor;

11. The mining machine according to claim 10, further comprising an actuator for moving said shearer head within a substantially vertical plane towards a seam boundary, wherein the seam boundary detecting sensors provides current seam characterising data output signals to the processor which generates further signals to the actuator to move the shearer head a distance within a substantially vertical plane towards a seam boundary according to an intended cut profile.

12. The mining machine according to any one of the preceding claims, further comprising a rail for the mining machine to traverse back and forth across the seam, said sensors collecting characterising data of said seam to generate a seam model of a first traversal to form the basis of the intended cut profile in a second traversal.

13. The mining machine according to any one of claims 10 to 12, wherein the seam characterising sensor comprises a seam boundary detector and preferably an infrared spectrometer; a gamma ray emission detector or a ground penetrating radar.

14. The mining machine according to claim 12 or 13, wherein the second traversal of the mining machine comprises said sensors collecting characterising data from said seam to generate a seam model which at least partially overlaps the seam model generated in the first transversal, the characterising data of the overlapping section of the seam model used to validate, and correct if necessary, the intended cut profile.

15. The mining machine according to any one of the preceding claims, wherein the processor analyses the seam characterising data from the one or more sensors to determine the presence of at least one of a sensor inaccuracy and imprecision by detecting a discrepancy between the seam characterising data from the one or more sensors and the at least one of the seam model and the cut model, and in the absence of at least one of sensor inaccuracy and imprecision, controls the one or more parameters of said mining machine based upon analysis of the at least one of the seam model and the cut model to anticipate changes in mining conditions as said mining machine progresses along the intended cut profile within the seam model.

16. A process for controlling a mining machine of any one of claims 1 to 15 including the steps of: A. moving the mining machine relative to the seam; B. generating current co-ordinate position output data signals from the position determining device indicative of the co-ordinate position of the mining machine; C. generating current seam characterising output data signals from the one or more sensors, distinct from the position determining device; D. the processor receiving said output data signals from steps B & C to thereby generate a seam model and/or cut model comprising the spatial position of the seam and one or more characteristics of the seam attached or offset therefrom, with the spatial position of the seam and the co-ordinate position of the characteristics of the seam determined by reference to the co-ordinate position in space of the mining machine; E. the processor using seam model and/or cut model data to anticipate required changes to one or more mining machine parameters; F. the processor sending an output signal to adjust one or more of the mining machine settings, parameters and/or providing an alert to an operator to signal what machine settings or parameters should be monitored or changed.

17. The process according to claim 16, wherein the processor also receives output data signals from the mining machine relating to one or more settings or performance parameters of the mining machine, said settings or parameters forming part of the cut model, with said cut model comprising the spatial position; one or more characteristics of the seam that has been extracted; and/or one or more characteristics of the mining machine performance characteristics as said mining machine extracted the material from the seam, said seam and mining machine characteristics attached or offset from the spatial position within the cut model.

18. Use of a seam model and cut model to control one or more parameters of a mining machine based upon analysis of the seam model and/or cut model, wherein the cut model and the seam model is generated by a processor through receiving output data signals from an at least 2D position determining device and one or more seam characterising sensors.

19. Use of claim 18, wherein the processor controls the one or more parameters of said mining machine based upon analysis of the seam model and/or cut model to anticipate changes in mining conditions.

20. Use of claim 18 or 19, wherein the cut and/or seam model comprises a co-ordinate position of the seam characterising data determined by reference to an at least 2D co-ordinate position in space of the mining machine.

21. Software that, when executed by a computer, causes the computer to perform the process or use of any one of claims 16 to 20.