WO2016011484A1 - Assessment method and system - Google Patents

Abstract

Contemplated is an assessment method that comprises the use of an information theory entropy objective function with various constraints to estimate at least one characteristic of a mineral particle stream of a mineral processing apparatus. Also contemplated is an assessment method for a time varying process in which the fundamental process information is not directly observable that comprises the use of an information theory entropy objective function with various constraints to estimate at least one characteristic of the fundamental process information which is not directly observable, wherein at least some of the constraints are provided by measuring a characteristic of the process, which is related to the characteristic to be estimated, at a number of different times.

Description

ASSESSMENT METHOD AND SYSTEM

FIELD

The present disclosure relates to an assessment method and system and especially, but not exclusively, to an assessment method for processing of mineral ores into concentrated ore-rich streams.

DEFINITION

In the specification the term "comprising" shall be understood to have a broad meaning similar to the term "including" and will be understood to imply the inclusion of a stated integer or step or group of integers or steps but not the exclusion of any other integer or step or group of integers or steps. This definition also applies to variations on the term "comprising" such as "comprise" and "comprises".

BACKGROUND

Processing of mineral matter usually consists of numerous operational units contained and linked within a specific mineral processing plant. The number of operational units can be one hundred or more, with each unit having a specific function and purpose. These operational units are generally involved in two main processes: 'comminution' (grinding) and 'separation'. Processing of mineral matter by comminution and separation has the purpose of releasing and separating the valuable ore from the naturally occurring non- valuable mineral matter associated with the ore.

Separation itself is generally divided into classification and concentration. Classification means to separate particles according to size, whilst concentration means to create two products - one with a higher grade of the mineral of interest, and one with a lower grade of mineral of interest. Typically, after successive concentrating of streams, a saleable product is created.

The assessment of mineral distributions in mineral processing streams is important in order to allow control of the operational units to improve performance. A number of approaches have been applied.

In one approach, if the mineral particles are binary (meaning that that they consist of only two minerals) then some testing which differentiates particles with different proportions can be used. For example, float-sink tests can be used. While this is common practice for coal, it is much more difficult for iron ore due to its high density. Optical methods have been used for diamond processing. A binary model simulation approach may be used: complex mineral ores processing system designs may ignore the fact that particles are multimineral and instead use a binary model simulation even if it is inappropriate due to the multimineral nature of the particles.

In another approach, key information such as assays and size distribution are measured and quantitative interpretations are made from these measurements. This approach relies on experts to make these interpretations. However there is often a lack of quantitative data to support or refute the interpretations and resulting recommendations regarding plant operation, and this limits the ability to act decisively.

A further approach is simply not to perform assessment at all during plant operation but to rely upon laboratory test work performed prior to or during the design of the plant. The laboratory results and their interpretation are the basis of plant design criteria and hence there is limited opportunity for process improvement once the plant is commissioned and operational.

A further approach is to perform detailed mineralogical analysis of the mineral streams. This approach is limited because it is expensive and there are often significant time delays between the events when samples of particles are sent for analysis and when analysis is returned.

The present inventor has discerned that there is scope for an improved assessment method, or at least one which provides a useful alternative to those described above, and consequently scope to provide an improved or alternative method of controlling operational units in minerals processing.

SUMMARY

According to a first aspect of the present disclosure there is provided an assessment method comprising use of an information theory entropy objective function with various constraints to estimate at least one characteristic of a mineral particle stream of a mineral processing apparatus.

In an embodiment, at least one of the constraints comprises a characteristic of the mineral processing apparatus.

In an embodiment, at least one of the constraints comprises a relationship between feed and product streams of at least one operation unit of the mineral processing apparatus. In an embodiment, at least one of the constraints comprises a plant data relating to the mineral processing apparatus and/or a plant of which the mineral processing apparatus forms a part.

In an embodiment, at least one of the constraints comprises data relating to the number of feed and product streams of the mineral processing apparatus.

In an embodiment, at least one of the constraints comprises a related characteristic of a different stream of the apparatus.

In an embodiment, at least one of the constraints comprises a measurement of the characteristic.

In an embodiment, at least one of the constraints comprises mass balance data relating to the mineral processing apparatus.

In an embodiment, at least one of the constraints comprises a partition function of the mineral processing apparatus.

In an embodiment, at least one of the constraints comprises data relating to the operation of the mineral processing apparatus at a different time.

In an embodiment, at least one of the constraints comprises data relating to bulk information about a mineral at least some of which is included in the mineral particle stream.

In an embodiment, at least one of the constraints comprises data relating to the operation of the mineral processing apparatus at a plurality of different times.

In an embodiment, at least one of the constraints comprises data relating to the mineral processing apparatus and/or a plant of which the mineral processing apparatus forms a part, derived from a previously performed apparatus audit or plant audit.

In an embodiment, at least one of the constraints comprises data relating to the mineral processing apparatus and/or a plant of which the mineral processing apparatus forms a part, derived from a number of previously performed apparatus audits or plant audits.

In an embodiment, at least one of the constraints comprises data relating to the mineral processing apparatus and/or a plant of which the mineral processing apparatus forms a part, derived from two, and only two, previously performed apparatus or plant audits. In an embodiment, at least one of the constraints comprises data relating to the mineral processing apparatus and/or a plant of which the mineral processing apparatus forms a part, derived from at least three previously performed apparatus or plant audits.

In an embodiment, at least one of the constraints comprises data regarding the operation of the mineral processing apparatus on an input feed of different composition to the input feed of the mineral particle stream of which a characteristic is estimated.

In this embodiment the method may estimate said at least one characteristic by combining information from different feed ores, rather than treating each ore independently.

In an embodiment, at least one characteristic estimated by the method comprises a distribution of mineral particles in the mineral particle stream.

In an embodiment, at least one characteristic estimated by the method comprises a density distribution of mineral particles in the mineral particle stream.

In an embodiment, at least one characteristic estimated by the method comprises a multimineral particle distribution of mineral particles in the mineral particle stream.

In an embodiment, at least one characteristic estimated by the method comprises a multimineral particle density distribution of mineral particles in the mineral particle stream.

In an embodiment, at least one characteristic estimated by the method comprises a size distribution of mineral particles in the mineral particle stream.

In an embodiment, at least one characteristic estimated by the method comprises an assay within size-classes of mineral particles in the mineral particle stream.

In an embodiment, at least one characteristic estimated by the method comprises average density within size classes of mineral particles in the mineral particle stream.

In an embodiment, at least one characteristic estimated by the method comprises average density within size classes of mineral particles in the mineral particle stream estimated using size distribution and bulk average density within streams as constraints.

In an embodiment, at least one characteristic estimated by the method comprises average mineral composition of mineral particles in the mineral particle stream.

In an embodiment, at least one characteristic estimated by the method comprises average mineral composition of mineral particles in the mineral particle stream estimated using bulk assays and/or basic mineralogical information as constraints. In an embodiment, at least one characteristic estimated by the method comprises multimineral particle composition of mineral particles in the mineral particle stream.

In an embodiment, at least one characteristic estimated by the method comprises multimineral particle composition of mineral particles in the mineral particle stream estimated using solid flow, size information, assays within sizes and limited mineralogical information as constraints.

In an embodiment, at least one characteristic estimated by the method comprises a comminution breakage function of the mineral process apparatus.

In an embodiment, at least one characteristic estimated by the method comprises a comminution breakage function of the mineral process apparatus estimated using product and feed size distribution data estimated using at least one previous apparatus or plant audit as a constraint.

In an embodiment the method comprises taking a characteristic related to the operation of a minerals processing unit which acts on the product stream as a prior for use in providing a constraint in the information theory entropy objective function.

In an embodiment the method comprises taking a characteristic related to the operation of a minerals processing unit which acts on the product stream as an ignorant prior.

In an embodiment the method comprises taking a partition curve related to the operation of a minerals processing unit which acts on the product stream as a prior for use in providing a constraint in the information theory entropy objective function.

In an embodiment the method comprises taking a partition curve related to the operation of a minerals processing unit which acts on the product stream as an ignorant prior.

In an embodiment the method comprises taking consistency of operation of the minerals processing unit over time as a constraint.

In an embodiment the method comprises acquiring data from an earlier audit to be used as the basis of a constraint, and taking at least part of the acquired data as a prior for use in providing a constraint in the information theory entropy objective function.

In an embodiment the method comprises adjusting the prior to be consistent with corresponding more current measured data of the mineral particle stream

In an embodiment the method comprises calculating expressions for the acquired data and for the more current measured data which describe a characteristic to be estimated.

In an embodiment the method comprises forcing equality of the calculated expressions. In an embodiment the method comprises using the calculated expressions to provide two estimates for the characteristic to be estimated.

In an embodiment the method comprises repeating the steps of:

-adjusting the prior to be consistent with corresponding more current measured data of the mineral particle stream;

-calculating expressions for the acquired data and for the more current measured data which describe a characteristic to be estimated;

-forcing equality of the calculated expressions; and

-using the calculated expressions to provide two estimates for the characteristic to be estimated;

until convergence of the two estimates is achieved.

In an embodiment calculating expressions for the acquired data and for the more current measured data comprises calculating partition curves for the mineral processing apparatus for the earlier audit and for the more current measured data.

In an embodiment the acquired data and the more current measured data relate to particle distributions for the mineral processing apparatus.

In an embodiment using the calculated expressions to provide two estimates for the characteristic to be estimated comprises applying these expressions to at least one mineral feed to be processed by the mineral processing apparatus to provide the mineral particle stream.

In an embodiment the method comprises use of at least one estimated characteristic of a mineral particle stream of a mineral processing apparatus, which has been estimated by use of an information theory entropy objective function with various constraints, as a constraint in an information theory entropy objective function, to thereby estimate at least one further characteristic of the mineral particle stream.

In an embodiment the method comprises use of at least two estimated characteristics of a mineral particle stream of a mineral processing apparatus, which have been estimated by use of an information theory entropy objective function with various constraints, as a constraints in an information theory entropy objective function, to thereby estimate at least one further characteristic of the mineral particle stream.

In an embodiment the estimated average density of particles in the mineral particle stream is used as a constraint. In an embodiment the estimated size distribution of particles in the mineral particle stream is used as a constraint.

In an embodiment the method is implemented by a computer.

According to a second aspect of the present disclosure there is provided a system for assessing a mineral particle stream of a mineral processing apparatus comprising processing apparatus programmed with instructions for processing the acquired data to implement an assessment method comprising use of an information theory entropy objective function with various constraints to estimate at least one characteristic of a mineral particle stream of a mineral processing apparatus.

According to a third aspect of the present disclosure there is provided a system for assessing a mineral particle stream of a mineral processing apparatus comprising:

-data acquisition apparatus for acquiring data for use in an assessment method;

-processing apparatus programmed with instructions for processing the acquired data to implement an assessment method comprising use of an information theory entropy objective function with various constraints to estimate at least one characteristic of a mineral particle stream of a mineral processing apparatus; and

-data output apparatus for output of data relating to the estimate of the at least one characteristic.

According to a fourth aspect of the present disclosure there is provided a method of providing improved operation of mineral processing apparatus comprising using the method in accordance with the first aspect, determining how the estimated characteristic can be changed to provide improved performance by changing operation of the mineral processing apparatus, and implementing changes in operation to provide improved mineral processing apparatus operation.

In an embodiment implementing changes in operation comprises implementing changes in operation of the mineral processing apparatus comprising the mineral particle stream of which the characteristic was estimated.

In an embodiment implementing changes in operation comprises implementing changes to a mineral processing operating unit which is not directly connected to the mineral particle stream of which the characteristic was estimated. An example of this is where the mineral particle stream of which the characteristic was estimated is a product stream of a separating apparatus, and changes are implemented in comminution of the feed to the separating apparatus.

In an embodiment implementing changes in operation comprises providing new mineral processing apparatus in which changes are implemented.

According to a fifth aspect of the present disclosure there is provided a physical medium comprising computer executable instructions for operating a computer to perform a method in accordance with the first aspect.

In the above, mineral processing apparatus may, unless logic or context determine otherwise, relate to a single operating unit, to a plurality of related operating units or to a mineral processing plant.

According to a sixth aspect of the present disclosure there is provided an assessment method for a time varying process in which the fundamental process information is not directly observable comprising use of an information theory entropy objective function with various constraints to estimate at least one characteristic of the fundamental process information which is not directly observable, wherein at least some of the constraints are provided by measuring a characteristic of the process, which is related to the characteristic to be estimated, at a number of different times.

It will be appreciated that functions, characteristics and other features set out in relation to embodiments of the first aspect are applicable to the other aspects.

BRIEF DESCRIPTION OF THE DRAWINGS

Embodiments will be described below, in detail, with reference to accompanying drawings. The primary purpose of this detailed description is to instruct persons having an interest in the subject matter of the invention how to carry the invention into practical effect. However, it is to be clearly understood that the specific nature of this detailed description does not supersede the generality of the preceding broad description in relation to the monopoly claimed. In the accompanying drawings:

Figure 1 is a schematic illustration of particles being concentrated by flotation;

Figure 2 is a schematic illustration of particles being broken into smaller particles by comminution;

Figure 3 is a schematic illustration of a binary particle;

Figure 4 is a schematic illustration of a multimineral particle; Figure 5 is a schematic illustration of an example of a simple mineral processing circuit with two separation streams (in this case cyclones);

Figure 6 is a diagrammatic illustration of a basic inference approach;

Figure 7 is a diagrammatic illustration of an inference approach using variable feeds assessed at different plant audits;

Figure 8 is a diagrammatic illustration of estimation of multimineral particle distributions;

Figure 9 is a diagrammatic illustration of an example of the calculation stages used in order to estimate the information about the mineral processing plant useful for detailed understanding of the plant's performance;

Figure 10 is a condensed representation of Figure 9, but wherein three elements of the calculation of Figure 9 are illustrated as a single block;

Figure 1 1 a is an example of two natural ore feed density distribution curves plotted on a graph of density versus cumulative proportion;

Figure 1 1 b shows back-calculated feed density distribution curves, compared to the original density distribution curves shown in Figure 1 1 a;

Figure 12 shows an actual partition curve for an example separator unit and an estimated partition curve calculated from output flow rates by an implementation of the method of the present disclosure;

Figure 13a illustrates schematically an overview of an embodiment of software for use in performing an assessment method in accordance with the present disclosure;

Figure 13b illustrates schematically and in further detail an embodiment of a main processing part of the software of Figure 13a;

Figure 14 illustrates schematically an embodiment of a system for implementing an assessment method in accordance with the present disclosure; and

Figures 15a to 15e are sequential pages of a printout of computer code, provided by way of an example of code useful for the implementation of a method in accordance with the present disclosure, with explanatory comments.

GLOSSARY AND DEFINITIONS

In order to aid clarity in the present disclosure it is useful to provide a glossary of terms used herein. Mineral processing terms

'Mineral' - in mineral processing all matter that constitutes ore or mineral ore is regarded as a mineral with an intrinsic value. Mineral that does not have an intrinsic value is often described as 'gangue'.

'Mineralogy' - is the subject study of minerals. Mineralogy generally implies that a detailed knowledge of the minerals is provided by advanced microscopic methods.

'Process Mineralogy' - is the subject study of mineral processing from the viewpoint of mineralogy. It needs to be understood that not all mineral process designers necessarily consider mineral, but may limit their investigation of a process to elements only.

'Elements' - are the same as chemical elements. Minerals are therefore made up of elements.

'Binary particle' - a particle that consists of only two minerals is called binary.

'Liberated' - a particle that consists of only one mineral - generally produced by the breakage of particles. The process by which particles become liberated is called liberation.

'Composite particle' - a particle consisting of more than one mineral.

'Multimineral particle' - a particle that consists of numerous minerals is called multimineral particles; whilst a multimineral particle may be binary; binary particle would generally not be called binary. The binary structure is generally the current convention for many simulation models in mineral processing. The word 'multimineral' is used instead of 'composite' to indicate that the modelling structuring facilitates potentially more than two minerals.

'Simulation model' - a simulation model is a computer emulation of an actual process.

'Assays' - a sampling test to determine the average composition of an element.

'Element to Mineral Conversion' - often the information from the elements is used to calculate mineral composition. This is called an element to mineral conversion.

'Units and streams' - a mineral processing plant is made up of units and streams. Units perform some processing task. Streams carry matter, such as ore, from one unit to the next.

'Plant Audits'- also called 'Plant survey' is the process of data collection and analysis to provide some understanding on the efficiency of a plant's performance. The wording is generally non-distinct. However a plant survey is preferentially used where data is collected from a larger number of streams with the purpose of deeper understanding - particularly the behaviour of some units. A full 'Plant survey' would therefore mean the collection and analysis of data sufficient to understand the unit behaviour of all units.

'Timed Audits' - For timed audits, where time normally indicates that a different feed mineral ore is being processed; then time is not a specific variable.

'Partition curve' - a 'partition' has both physical and mathematical meaning; and implies the way particles behave differently at units due to physical variability. For example a density partition curve describes the way in which heavier particles might go to the 'heavies' stream preferentially to lighter material. The word 'partition' is mainly used in density separation and size separation processes. It is seldom used for flotation. A density partition curve is therefore a graphic representation describing the differential behaviour of heavier particles to lighter particles.

Washability' - a washability curve generally refers to the density distribution of the mineral ore feed going into a unit.

'Comminution' - breakage, particularly of ore material.

'Feed' - naturally occurring mineral ore input to a processing unit.

'Products' - ore coming out of a unit. Whilst 'product' is generally used for the valuable ore product it can be used for both. More distinct terminology is concentrate (largely valuable processed ore) and tails (largely nonvaluable ore).

'Heavies' or 'Underflow' - the bottom discharge from a wet cyclone.

'Lights' the lightest density discharge from the top of the cyclone.

'Middlings' the medium density to discharge. Normally produced as the light heavies or heavy lights. With respect to mineral of interest, heavies are generally composite particles.

'Control System' - generally refers to electronically manipulated systems for interpreting and changing processes to achieve a desired outcome.

'Flotation' - a mineral processing separation process utilising different material densities

Mathematical definitions

The word 'inference' implies estimation of a variable that is unmeasured. Here 'inference' is categorised into 'deductive inference' and 'plausible inference'. 'Deductive inference'. Deductive inference is when the measured variable is estimated with high accuracy. For example if A=100 and A+B=120; then it is logical and standard to estimate B as 20.

'Plausible inference' - plausible inference is when an estimate of a variable has a high level of uncertainty; i.e. it is not known with certainty the values of the variable but an estimated value will be used for practical purposes.

'Information theory' (also called principle of maximum entropy) provides the mathematical framework to infer information. Information theory is well-known in electrical engineering and theoretical physics. It is not widely known in mineral processing. In the context of the present disclosure information theory is used to infer multimineral particle distributions (inclusive of binary particle distributions).

'Prior'- a prior is a prior belief about variables. For example, if one is given a die, a standard 'prior' is that the die is unbiased. A prior is different to an 'assumption', which is generally fixed; whereas a prior is a starting belief with expectation that the estimated variable will deviate from the prior if information proves so. In the context of the present disclosure, in the absence of information to the contrary, there is the prior belief that for a particular unit, this unit performs the same for different plant surveys (with respect to ore multimineral particles). Therefore there are two variable classes being estimated - the ore multimineral particle distributions and the way these particles are processed in units.

'Entropy' - a parameter used for the purpose of evaluating P (unit behaviour) for a single audit and single unit given by a mathematical formula: e.g. for a die:

Entropy = p. 1η(/ )

'Well-posed' - suppose there are N variables and N_M measured variables and N_c constraints. Then if N <= N_M +N_C then (subject to certain mathematical conditions) all variables can be estimated. This problem is called well-posed.

'ill-posed' - for the example above if N > N_M +N_C then the problem is called 'illposed' and the variables cannot be reliably estimated with high certainty.

'Markov Chain Monte Carlo' - also abbreviated to MCMC. A mathematical probabilistic approach that allows efficient computational modelling of multi-dimensional information - in this case multimineral particles. 'Hidden Markov Models' - a mathematical framework developed for the analysis of variables that cannot be directly observed. The theory of Hidden Markov Models is generally limited to time series analysis whereas the disclosure herein also includes spatial (stream locations) information.

'Distribution' - herein a distribution means a statistical distribution. The phrase 'density distribution' often used in statistics is herein avoided because of ambiguity with physical density.

DETAILED DESCRIPTION OF EMBODIMENTS

With reference to the accompanying drawings embodiments of assessment methods in accordance with the present disclosure will be described.

As set out in the background section above, two main processes in minerals processing plants are comminution and separation. The examples provided herein relate mainly to separation, but it will be appreciated that the methods described are applicable to other processes in mineral processing, including comminution.

The present disclosure relates to an assessment method which can estimate unknowns, and in particular examples, unknowns relating to mineral particles in a mineral processing plant.

There are two classes of unknowns that it is desirable to estimate. These are: first, the ore characteristics, and in a particular example the proportions of different minerals in multimineral particle types; and, second, the way in which these particles are processed at units.

In separation processes, it is desirable to identify the probability that a given particle type goes to a given product stream; for example, in a density separation unit, the probability that a particle of a particular size and density will go to the 'heavies'.

For comminution the probability model is more complex. When a particle is broken it generates smaller particles with a varying distribution of particle types. This is now the probability distribution being estimated.

Figure 1 is a representation of mineral-bearing particles being concentrated by flotation. In Figure 1 a number of different binary particles, designated by reference numerals 1 to 6, each comprises at least one of two minerals. The first mineral, designated 7, is a mineral which is valuable, represented by the darker colour. The second mineral, designated 8, is a mineral which is not valuable. In this example the first mineral 7 is denser than the second mineral 8. The particles 1 to 6 are fed by a feed 9 into a flotation separator 10, which has a concentrate stream output 1 1 and a lights stream output 12. Other factors being equal, particles which have a higher proportion of the first mineral have a higher probability of going to the concentrate (underflow) stream output 1 1 than do particles with a lower proportion of the first mineral. The probability that a particle of a particular type will go to the concentrate stream output 1 1 is one of the mathematical probabilities capable of being estimated using the method of the present disclosure. The particles 1 to 6 are shown on the left hand side of Figure 1 , representing the state prior to passage through the flotation separator 10, and on the right hand side of Figure 1 , representing the state after passage through the flotation separator 10.

Figure 2 is a representation of mineral-bearing particles being broken into smaller particles by comminution. As illustrated in Figure 2 larger particles 21 , 22, 23 (shown on the left hand side of Figure 2) are passed via a feed 24 into a comminution device, for example ball mill 25, are broken into smaller particles, and exit comminution device via output stream 26. The smaller particles formed by breakage of the larger particles 21 , 22, 23 are shown on the right hand side of Figure 2, and collectively designated 27. The probability that a larger particle of a particular type will form a smaller particle of a particular type as shown Figure 2 is another mathematical probability capable of being estimated using a method in accordance with the present disclosure.

It is important to appreciate the difference between binary particles and multimineral particles.

There are often considered to be five main derivative branches in mineral processing, namely: coal; iron ore; diamonds; gold (and other precious metals); and sulphide minerals. Whilst there are exceptions, coal, iron ore and diamonds are often considered relatively simple to model because the particles to be processed are generally binary, that is, a combination of a specific valuable mineral and a non-valuable mineral. Figure 3 shows a representation of a binary particle 30 in which the lighter coloured mineral 31 can be considered valuable and the darker coloured mineral 32 can be considered non-valuable. Hence for iron ore, the valuable mineral is the iron ore (hematite) and the non-valuable mineral is host rock.

In contrast, gold and other precious minerals, and sulphide minerals, are generally considered complex and a multimineral representation is more appropriate. Figure 4 shows a representation of a multimineral particle 40. For sulphide minerals in particular, it is well known that particles to be processed may comprise many minerals. For example, even a moderately copper bearing ore could consist of: non-sulphide gangue (NSG), in multimineral particle 40 represented by the lightest region 41 ; pyrite (non-valuable), in multimineral particle 40 represented by the larger grey region 42, and copper-bearing mineral (valuable). Further, the copper-bearing mineral could consist of different sub- minerals such as chalcopyrite, represented by a black region 43 in multimineral particle 40 and copper-oxide represented by the smaller grey region 44 in multimineral particle 40. The floatability of the particle (or probability of going to the concentrate) is affected not by the composition and proportion of just one mineral but by the composition and proportion of a number of minerals.

The different regions 41 to 44 in multimineral particle 40 represent different types of mineral (recognising that all matter in mineral processing is generally classified as a mineral). Multimineral particle distributions can be determined via mineralogy. This involves taking samples of particles and subjecting them to image analysis methods. This approach is costly and takes a long time. Hence it is seldom used, although it is not uncommon to use mineralogical analysis if particular operational units require detailed investigation.

However, mineralogical analysis is often applied to ore samples prior to processing. These samples are used to provide broad understanding of the ore characteristics, and not specifically used for plant operational improvement. Therefore it is believed that in mineral processing in many instances there is sufficient information to convert elemental information to mineral information (sometimes referred to as element to mineral conversion).

As an aside, it is important to understand that the difference between minerals and elements is very important from a processing viewpoint. For example, copper oxide and copper sulphide are both copper bearing minerals, but copper sulphide is floatable and copper oxide is not. Indeed the problem is even more pronounced because different types of copper sulphide minerals (i.e. chalcocite, chalcopyrite) may float at different rates. Therefore, from a processing viewpoint (if one wants to understand reasons for an inefficiency) it is important to identify the mineral that copper is actually contained in.

As mineralogical assays are generally available, information obtained from such assays can be used to estimate average mineral compositions; but assays cannot conventionally be used to estimate the detailed multimineral particle distribution. The problem of estimating the multimineral particle distribution from assays (and hence average mineral compositions) is an ill-posed problem. Information theory provides a basis to estimate the multimineral particle distributions. While application of information theory can estimate these distributions the inferred estimates contain some level of uncertainty. This uncertainty can be reduced by using more 'information'. There are two relevant classes of Information:

1 . Measurements; and

2. Knowledge of the process.

Some basic knowledge of the process is the connection between streams and units.

For example, consider the case where there is a single operating unit which has one feed stream and two product streams. For a simple estimation of solid flow for all three streams only two solid stream measurements are needed (because it is known that the solid flow through the two product streams is equal to the solid flow through the feed stream). This is because the relationship between the streams and the unit (three streams and one unit) is one constraint. There is therefore sufficient information to 'deduce' the solid flow of the third stream, without uncertainty or the need to minimise the uncertainty of the calculated value. This may be considered an example of deductive inference.

This example of an approach of 'estimating' variables, when there is more information than the number of variables, is called mass balancing; and is also referred to as material mass balancing, mineral mass balancing and data reconciliation.

More generally, if we have N unknown variables, and N_M measured variables and N_c constraints, then if:

N <= N_M + Nc

then (subject to some mathematical conditions) all variables can be calculated. This problem is called well-posed.

However, if:

N > N_M + Nc

then the problem is called 'ill-posed' and the unknown variables cannot be estimated by orthodox approaches. The estimation of such ill-posed problems is referred to herein as 'inference' whist the calculation of 'well-posed' problems may be referred to herein as mass balancing. An example of an ill-posed problem in the above example would be if the solid flow in only one stream, for example the feed stream, were known. In this case the solid flow of the product streams could not be deduced. An inference of the solid flow of the product streams could be made as say half the feed solid flow, but such an inference would contain considerable uncertainty to the extent that the inference would be considered unreliable.

The problem of estimating the particle mineral distribution is more complex, at least in part because the particle mineral distribution is a distribution rather than a defined number of variables.

For example, Figure 5 illustrates a mineral processing apparatus 50, with a feed stream 51 entering a first cyclone separation unit 52, which outputs a first lights stream 53 and a first heavies (concentrates) stream 54. The first heavies stream 54 enters a second cyclone separation unit 55 which outputs a second lights output stream 56 and a second heavies stream 57. The second lights stream 56 is combined with the first lights stream 53 and directed to a lights product output 58. The second heavies product stream is fed to a heavies product output 59.

It will be appreciated that Figure 5 is intended to be a simple example of a minerals processing apparatus. Depending on the ore, the heavies may be valuable product, as would be the case for iron ore (iron being dense). However for coal, 'lights' are the product. The second lights output stream (which could be considered a 'middlings' stream) is illustrated as being combined with the first lights output stream, but could alternatively be comminuted further, reseparated, further treated in any desired manner, stored as a product, or any combination of these options. However, irrespective of the specific mineral, it is desirable to ascertain the density distribution of the streams and the separation characteristics of the separation device. This separation characteristic is called a 'partition curve'.

Thus Figure 5 thus illustrates a system comprising two separation units and five streams, for which it is desired to estimate the density distributions.

Referring again to Figure 5 by way of example, if the solid flows and average density at all streams are known then there are five streams, ten measurements (being two measurements for each stream) and two units.

If the density distribution is split into N particles (or components), there are:

5^*N unknowns

10 measurements (average density and solid flow for each of the five streams)

2^*N Unit constraints (mass balance).

Therefore the problem is well-posed if: 5N < 10 + 2N

So N=3 is the maximum number of components for the problem to be well-posed.

This approach of limiting the number of components to be well-posed is a common technique in mineral processing.

Clearly if there are more than 3 components or particle types the problem is ill-posed. It needs to be emphasised that limiting the number of components has not decreased uncertainty of the original problem (to understand the complete density distribution). All that has been done is that uncertainty has been removed (even though it still exists) from the mathematical formulation by reducing the amount of information that it is attempted to know. In other words the mathematical formulation is a simplification of the reality. Therefore the method of the present disclosure adopts 'inference' methods via information theory to estimate multimineral particle distributions using information theory as illustrated Figures 6 and 7.

Figure 6 illustrates schematically the use of an inference model 61 to use plant data 62 to infer stream data 63 and unit performance 64.

It is believed that information theory has not previously been used for inference of multimineral particle distributions. In this context the word 'inference' implies 'plausible inference' as distinct from deductive inference. Here, 'plausible inference' means an inference of a measured variable by minimising the uncertainty of the calculated value, in contrast to deductive inference meaning estimating a variable by using logical deduction without the requirement for minimisation of uncertainty.

However, it is highly desirable to be able to provide a reduced level of uncertainty in the inferred estimates of multimineral particle distributions. An explanation of how this can be achieved follows.

Simply put, more information is required in order to reduce the uncertainty.

This additional information can be obtained from earlier plant audits.

The two things that may change for different plant audits, are:

the feed (the multimineral particle distribution); and

the plant unit processing capability. Whilst recognising that the feed may change between different plant audits, it is a reasonable fundamental assumption that the behaviour of the units will not change as significantly.

Thus a particle with particular properties going through a unit may well behave similarly to a particle with the same properties going through the same unit at a different time. That is, similar particles are likely to behave similarly in a given unit, even at different times.

This provides information that can be used to decrease the uncertainty of the inferred results.

The partition curve is here used as a general term to represent how particles are separated in a unit.

Suppose we have only three streams: feed, denoted by subscript F; 'heavies' product, denoted by subscript H; and tail.

The distribution of particles in each stream is represented by f. The solid flow is represented by S. Thus S_F f_{F i} is the mass flow of the i^th particle type in the feed stream.

The probability or proportion of these particles that to go the 'heavies' (underflow) is: p _ ^ ¹¹ f^H->

' ^~ S, f, _.,

In the context of mineral processing the partition curve is used for density distribution:

Thus for the case of two cyclones, but where there is a single 'heavies' output, a simplistic description of the partition curve is: s_F fAP)

Where r is the partition curve; S_H is the solid flow of the heavies product; S_F is the solid flow of the feed and _,/^'(/?) is the density distribution. The 'prior' is therefore that the partition curve does not change. If the prior were indeed 'fixed' in practice, then for the case where we have two different feeds (each for a different audit) then the problem now is:

10^*N unknowns

20 Measurements (density)

4^*N Mass Balance Unit constraints

2^*N Partition curve constraints

This means that the problem is well-posed if:

ION < 20 + 4N + 2N which is true for N<5.

So N can now be 5 and the problem is well-posed whereas without the use of the 'prior' that the partition curve does not change, N was 3.

This means that, in practise, the uncertainty has now significantly decreased. Further, and possibly more importantly, the more times the plant is audited the more the uncertainty decreases. This may be regarded as providing a choice or tension, between conducting (and/or integrating information from) a greater number of audits and reducing the level of uncertainty. Alternatively, this may be regarded as providing a compounding benefit: the more information from different audits is integrated, the more uncertainty is reduced; and the uncertainty can become so low that the number of measurements can be confidently reduced.

Figure 7 illustrates schematically the integration of information from different plant audits into an assessment method of the type illustrated by Figure 6. More specifically, Figure 7 illustrates schematically the use of an inference model 71 , based on similar unit performance, which uses first, second and third sets of plant data 72, 73, 74 to infer first, second and third sets of stream data and unit performance data 75, 76, 77.

As will be appreciated from the above discussion, the first, second and third sets of plant data 72, 73, 74 relate to measurements from a single plant taken at respective first, second and third audits (or more generally at respective first, second and third times). The first, second and third sets of inferred stream data and unit performance data 75, 76, 77 are sets of data inferred (calculated) at different times. It is important to note that this analysis does not rely on the ore feed being constant between audits - the partition curve can be used as a prior even when the feed ores are different. A mathematical basis for use by the inference model 71 , and based on information theory, will now be described.

The entropy for the purpose of evaluating P (unit behaviour) for a single audit and single unit is given by: J = -F_{k i} ' P_k H^Pu ) (1 )

Where the prime ' indicates the variable is non-differentiable.

As a first step, an initial estimate of particle distributions can be obtained by solving for the entropy (eqn.1 ), subject to constraints, to give feed distribution and unit probability (unit behaviour).

As a second step, overall unit behaviour can be estimated as follows.

If there are numerous audits, then the entropy above is summed with respect to time.

Furthermore each timed probability distribution becomes a prior and we aim to find some average.

Hence over time:

The mathematical solution of this equation is the overall probabilistic distribution; and is denoted as Pk,i^*.

As a third step, stream particle distributions are re-estimated, as follows.

Equation (1 ) is used together with the entropy of the feed, and the overall unit behaviour is included as a prior.

Steps 2 and 3 are successively repeated until convergence.

The relationship between information theory and entropy is here clarified.

Entropy is a well-known term in thermodynamics. Various mathematical theorists were able to derive the mathematical formulation of entropy as:

-∑ !^η(/ ) Shannon in the 1930s independently derived the above equation for signal processing; and his theory was described as 'a Theory of Communication'. For a paper on the foundations of information theory see, for example, Shannon CE Ά mathematical theory of communication' The Bell System Technical Journal, Vol. 27, pp. 379^123, 623-656, July, October, 1948 (the entire contents of which is incorporated herein by reference).

As the method became more widely used, particularly to other scientific areas, for example 'genetic modelling' the phrase communication theory changed to information theory, as it provided the basis for estimating variables and reducing uncertainty by including information.

Information theory is a sub-branch of the mathematical subject 'inverse problems', and is distinguished from other inverse methods by its general use of the entropy formulation. From an academic viewpoint, there are various modifications of the entropy approach, rather than adhering to the strict thermodynamic definition; and consequently various academic theorists (particularly Jaynes) have suggested that entropy when used by information theorists should be distinguished from thermodynamic entropy, for example by use of the phrases 'probability-entropy' or 'information-entropy'. No such change in definition has yet been generally accepted. What this means in practice is that researchers who are introduced to probability-entropy (for information theory) are likely to consider it necessary to explore the application of entropy to thermodynamics and this is quite difficult to master, and actually not necessary. For a general review of information theory see, E.T Jaynes, Probability theory as logic in: P.F Fougere (Ed.), Proceedings, Maximum Entropy and Bayesian Methods, Kluwer Academic Publishers (1990), also available in a revised and extended form at http://bayes.wustl.edu/etj/node1 .html, the entire contents of which is incorporated herein by reference.

With respect to modifications of entropy the most significant modification is the Kullback- Liebler divergence where the prior was introduced into the formulation. See, for example, Kullback, S., Leibler, R.A. (1951 ). "On Information and Sufficiency". Annals of Mathematical Statistics 22 (1 ): 79-86. doi:10.1214/aoms/1 177729694. MR 39968, the entire contents of which is incorporated herein by reference. Equation (3), above, is therefore an example of the Kullback-Liebler divergence.

An assessment method in accordance with the present disclosure is particularly applicable for dealing with multimineral particles. The multimineral extension of the approach using information theory is necessarily complex; yet it needs to be recognised that modern mathematical approaches have gone a long way to provide methods to deal with these extensions.

Mathematical approaches relevant to the problem of estimating multimineral distributions (or generally multi-dimensional problems) are Markov Chain Monte Carlo and Principal Component Analysis. In the paper 'Assessment of stereological equations for mineral processing' (Gay S.L., Latti D. 2010, IMPC 2010. Vol. 1 : 181 -190), the entire contents of which are incorporated herein by reference, the present inventor described methods by which Markov Chain Monte Carlo (MCMC) can be used for a single stream. The mathematics for dealing with multiple streams simultaneously (satisfying mass balance constraints and unit priors) is a mathematical extension of this approach. The method is based on there being two unknowns: the properties (multimineral composition) of individual particles and the distribution of these particles for each stream. The individual particles are determined by the MCMC approach described in the abovementioned paper (Gay S.L., Latti D. 2010, IMPC 2010. Vol. 1 : 181 -190). The actual distribution for each stream is determined by the information theory approach. What this means in practise is that the assessment can be started with any set of multimineral particle types (for example 200). This is called a seed set.

It will be appreciated that implementation of the described methodology will involve large amounts of data, and calculation can only be practicably accomplished using a computer. However, with suitable software implementation this methodology provides a way to easily monitor a process plant by use of information theory and including 'mass balancing' and 'inference' stages to the monitoring of a process plant. It will further be appreciated that accurate monitoring of a process plant makes a huge contribution to the implementation of process improvement.

In broad terms, for complex ores, an assessment method in accordance with the present disclosure can be used to provide an (accurately) estimated detailed description of the ore characteristics at each stream and hence a detailed understanding of how processing units are performing with respect to the ore. An approach to improving a mineral processing plant including use of the described methodology consists of the following nine stages, as set out in Table 1 below.

3 Mass balancing electronically manipulated control systems

4 Inference

5 Define operational parameters

6 Define Operation Models

7 Fit Operation Models

8 Simulation

9 Optimisation

The stages set out in Table 1 represent a process by which the mineral processing in a plant can be understood in detail, and in which improvement of plants would be greatly facilitated.

In comparing current practise with use of the nine stages set out in Table 1 , it is found that key steps are often missing, meaning that current practice does not allow full analytical process improvement. In particular the present disclosure facilitates the fourth step; 'inference'. That is, the described methodology provides the means and opportunity to improve plant mineral processing operations by including all nine stages in the mineral processing improvement process. (It should be appreciated that in this context the word 'inference' implies 'plausible inference', which is distinct from deductive inference.)

While understanding of mineral processing in a plant could, in theory, be achieved without the inference stage, this would require considerably more data to be provided (more samples and sample analysis) and make other stages in the nine-stage process more difficult. Accordingly, use of the inference stage is considered to make a high level of understanding of mineral processing in a plant practicable and economically viable, whereas in the absence of the inference stage such understanding would be impracticable, economically unviable, or at least considerably more costly.

Mineral processors will sample a plant. This means generally finding size distributions, and assays (elemental compositions) for each size at various streams. They do not have to sample these variables at all streams, but a sufficient subset to be able to estimate the variables at all streams. This estimation method is not generally called 'inference' but 'mass balancing'.

A common method of controlling mineral processing plants is by use of electronically manipulated control systems. Control systems generally are based on the idea of maintaining stability of plant operations, rather than specifically changing operating conditions over the long term to improve operation of the mineral processing plant. Some control systems measure size distributions and assays. In such cases this information can be used with an assessment method in accordance with the present disclosure to understand unit operations and stream ore characteristics to improve plant performance.

Therefore control systems can be linked with assessment in accordance with the present disclosure to ascertain (and implement) optimal or improved operating conditions.

To clarify the need for inference, refer again to Figure 4. The general approach to simulation is to ignore that particles are multimineral, and instead to treat them as binary in the simulation. However, in a multimineral particle the floatability of a mineral on the particle is affected not only by the presence of the mineral of interest but also the presence of other minerals. By ignoring the presence of other minerals, the ore is characterised by (for example) flotation tests, which are required every time there is a substantial change in the ore properties.

An approach which uses the mineral associations, by using a multimineral particle model, can provide a simulation model which is a better reflection of reality and which is both more accurate and requires far less recalibration.

Similarly, an approach which only measures average density (for each size for each stream) is unable to resolve the density distribution; and thereby cannot estimate the partition curves for cyclones or other gravity separation devices. Any attempt at simulation without this fundamental information would be unreliable.

Figure 8 illustrates schematically how the described methodology allows accurate monitoring of the performance of mineral processing plants by time-successive sampling with inference methods because it allows determination of the 'real' data structure. This may be regarded as allowing the modelling framework to match reality, at least more closely than approaches which apply simplified models (such as assuming a binary particle model for multimineral particles, or not modelling the particle distributions at all). As illustrated in Figure 8 by block 81 , the operating parameters of unit model 82 are considered 'fixed' providing the 'prior' which allows accurate inference of multimineral particle distributions which could not otherwise be accurately inferred. As with some other measurement/modelling methodologies, the present methodology relies upon input of observable ore properties, block 83, and provides output of observable ore properties, block 84. However, the inference methodology effectively allows input of the 'hidden' or 'unobservable' ore properties, corresponding to multimineral particle distributions, block 85, and the unit model provides output related to the 'hidden' ore properties (corresponding to multimineral particle distributions), block 86, from which the output of observable ore properties, block 84 are provided.

The word 'hidden' in this context has specific mathematical meaning; for example embodiments of the method outlined herein would be considered related to 'Hidden Markov Models' - another sub-branch of probability theory. That is, 'hidden' means unobserved.

Figure 9 illustrates schematically an assessment method in accordance with the present disclosure in which a number of calculation stages are performed in order to estimate information about the mineral processing plant useful for detailed understanding of the plant's performance. In this embodiment plant observable data, block 91 , and plant flow sheet data, block 92, are processed using information theory, block 93. Hidden plant data, in this example multimineral particle distributions, are then inferred, or estimated, block 94, but with a relatively high level of uncertainty. Hidden information, for example previous multimineral particle distribution data, from earlier times (for the same plant and operational units) is then integrated, block 95, providing additional constraints and allowing the hidden plant data (multimineral particle distribution data) to be estimated at a lower level of uncertainty, block 96.

It will be appreciated that knowledge of the particle distribution in each stream provides a direct knowledge and understanding of how well each unit is performing. This can also provide the basis (data structure) for simulating the operation of the mineral processing plant: for example to perform 'what if scenarios (such as simulating what would happen one unit processed the particles more efficiently, in terms of overall plant performance (and profitability). Simulation can also be used to assess whether introducing a new unit is of benefit.

Using multimineral particle distributions is necessary (or at least highly desirable) in order to obtain a good understanding of mineral processing plant operation, because in many cases ore particles are in reality complex, so that using a simpler model, such as a binary particle model, to represent the ore particles and their distributions will result in less detail and accuracy, and hence a poorer understanding and reduced capability of improving unit and plant performance.

Although such multimineral distributions can be estimated via mineralogical analysis, the present methodology allows estimation by analysis of conventional plant data, which will normally be considerably quicker and more cost effective. This conventional information includes assays within each size class at streams throughout the plant, and basic information required for assay to mineral conversion. While the use of information theory implies some uncertainty, this uncertainty can be reduced by integrating information from successive audits, or plant surveys (of the same plant). As mentioned above, the basis of this integration is the normally justified assumption that each operational unit's behaviour (with respect to multimineral particles) can be treated as being the substantially the same over plant audits at different times, for the basis of creating a prior. It should be understood that use of this behaviour similarity as a prior does not imply equality. That is, it is not assumed that the unit operations remain the same, but rather that information from successive plant surveys can be used to substantially improve the certainty of the measured particle distributions.

As previously mentioned, uncertainty is reduced by the integration of information from successive plant audits. This reduction in uncertainty can also allow a decrease in the number of required measurements per audit. This can lead to a desired level and certainty of information being made available without the need for detailed sampling strategies. For example, it may well be possible to use only bulk assays and size distributions rather than needing to use assays within sizes. In addition this latter capability means that even for simple ores, the described methodology can be used to substantially reduce the cost of sample analysis and audits. For example it is possible that this methodology can be used for coal by measuring only average density in each size-class - rather than performing detailed and costly float-sink tests.

Figure 10 illustrates schematically the method of Figure 9, in a summarised form to place emphasis on the input of plant observable data, block 101 , and plant flow sheet data, block 102, and output of an estimate of hidden plant data at relatively low level of uncertainty, block 106. Thus the three intermediate calculation steps (blocks 93, 94 and 95) of Figure 9 are shown as a single processing step, block 103, in Figure 10.

Figure 10 is a general representation of the methodology, which has applicability to various input and output combinations. The following scenarios, numbered 1 to 7, outline various input and output parameter combinations achievable with and by the use of the described methodology, and believed unattainable by other known approaches. In all of the below scenarios it is presumed that the plant configuration (the relationships between streams and units) is known, and that information is available from different plant audits.

Scenario 1: Estimation of size distributions

Input: Solid flows

Average sizes

Output:

Size distributions

Scenario 2: Estimation of assays with sizes

Input:

Solid Flows

Size distributions

Bulk assays

Output:

Assays within size-classes

Scenario 3: Estimation of average density with size

Input:

Solid Flows=

Size Information

Bulk average density

Output:

average density within size-classes

Scenario 4: Estimation of average mineral compositions at streams

Input:

Solid Flows

Assays

Basic mineral information of plant feed (average mineral composition) Relationship between minerals and elements

Bulk Assays for each stream

Output:

Bulk mineral compositions Scenario 5: Estimation of average mineral composition within size-classes at streams Input:

Solid Flows

Size Distributions

Assays

Basic mineral information of plant feed (average mineral composition for each size class)

Relationship between minerals and elements

Assays within Size-classes for each stream

Output:

Average Mineral Composition for each size for each stream

Scenario 6: Estimate multimineral compositions

Input:

Solid Flows

Size Information

Assays within sizes (either determine from scenario 1 or measured)

Limited mineralogical information (relationship between assays and minerals, mineral densities)

Output:

Multimineral particle information

Elemental compositions within particle types

Behaviour of units with respect to multimineral particles

Density distributions

Scenario 7: Estimate density distributions

Input:

Solid Flows

Size Information Average densities within sizes (either determine from scenario 2 or measured) Output:

Density distribution of particles

Understanding of how density separation units are partitioning particles according to density.

Scenario 8: Estimate Stream information in detail if feed is known in detail and only solid flows are measured in other streams

Input:

Solid Flows at all streams other than feed,

Feed information in detail

Output:

Stream detailed information

In some laboratory tests, a pilot plant, which is a smaller scale representation of an industrial-scale plant, is used. Similarly some laboratory tests require use of multiple units (or repeated use of the same unit). The described methodology may be analogously used in such laboratory tests. The described methodology thus has other uses, in addition to use in actual mineral processing plants.

Similarly some laboratory tests may consist of a single unit. Scenario 8 in particular can be used to infer product stream information in detail if product solid flow is known and feed information is known in detail.

The general approach described can also be used to identify the 'breakage function' for comminution. Again there are many variations. However, if feed and product size distributions are measured over repeated audits it is possible to estimate the breakage function: the proportion of particles of a particular size in the feed that originally belonged to a specific coarser size-class in the feed.

The described methodology was initially designed for multimineral particles rather than binary particles; however methods described in this specification can be applied to binary particles. Furthermore, even though iron ore, coal, and diamonds are often considered 'simple' there are enough exceptions that in many cases the multimineral methodology can beneficially be used to allow the ore to be modelled in greater depth and therefore more accurately than using a binary particle approach. In an embodiment for estimating mineral particle distributions in each stream of a mineral processing plant, the described methodology uses an information theory entropy objective function with various constraints (measurements and mass balance constraints) to make the estimation with a low level of uncertainty.

This may be used to add an 'inference' stage (from information theory) to estimate multimineral particle distributions in mineral processing, which may be used as Step 4 in the multiple stage process outlined in Table 1 .

In practice two or more scenarios (for example of scenarios 1 to 8 set out above, or other scenarios utilising similar principles) may be integrated. For example some streams may have assay with size, others may have just bulk assays or size distributions, and others just solid flow. Yet the nett result is that the various scenarios may be used together (synergistically) to capture detailed information in all streams.

An embodiment of an assessment method will be described by an application to the parameters for the above mentioned Scenario 4: Estimation of average mineral compositions at streams:

Input:

Solid Flows

Assays

Basic mineral information of plant feed (average mineral composition)

Relationship between minerals and elements

Bulk Assays for each stream

Output:

Bulk mineral compositions

Validating a mineral processing process can be done, where the assays are known in advance, for two feed mineral ores as represented in Figure 1 1 a.

The two feed mineral ores (shown on the graph of Figure 1 1 a as Feed A, designated 1 101 , and Feed B, designated 1 102) are processed via two separation units (as shown in Figure 5).

The partition curves are also known (for the purpose of validation) and the distribution characteristics (partition curves) are shown in Table 2. Table 2: Partition curves of the units used

for the example

Density Mid Value Partitionl Partition2

1 .35 0.40 0.22

1 .45 0.50 0.32

1 .55 0.60 0.44

1 .65 0.69 0.56

1 .75 0.77 0.68

1 .9 0.86 0.82

2.15 0.94 0.94

2.45 0.98 0.99

After applying the partition curves to the feed mineral ore a simple 'simulation' of the process application is performed; and the average densities are calculated.

Simulation of process applications and of minerals processing plants is known per se, and may be regarded as a collation of unit models. It is believed that currently such simulations in the mineral processing industry do not use multimineral particle models.

If the average density at all streams (and the feed flow rate) is known then fundamental mass balancing properties can be used to determine the flow rates at all streams (refer Table 3). Mass balancing is well known per se and its application, based on the present disclosure, will be appreciated by the skilled addressee. There have been previous suggestions to use information theory for mass balancing, at least in chemical engineering, see for example, CM. Crowe (1996). Formulation of linear data reconciliation using information theory. Chemical Engineering Science, 51 (12), p. 3359-3366, the entire contents of which is incorporated herein by reference. In this paper the author proposed use of information theory in mass balancing applied to chemical engineering; the paper regards the distribution of chemicals as unknown; and aims to determine this distribution for the purpose of calculating standard deviations. The mass balancing calculation only requires the top line of data (average density) to back-calculate the density distribution curves of the original feed mineral ore. This is then done for the example as described above and the results shown in Figure 1 1 b. It is evident that there is good correlation between the original feed data (FeedA designated 1 101 and FeedB designated 1 102) and estimated feed data (FeedAEst, designated 1 161 , and FeedBEst, designated 1 162).

It should be appreciated that this back-calculation was based on only two different audits and two separation units. In reality most plants will have more plant audits, and far more separation units rendering the method effective for back-calculating the multimineral particle distribution.

It will be appreciated that if only average density (for each size) and flow rates were measured, then it would not be possible to estimate the average density distribution nor the partition curve. Use of inference methods allows these estimates. In the above example coal is the desired product. It is generally considered that the maximum and minimum of the densities of ore are known - i.e. between 1 .2 (pure coal) and 2.6 (host rock). An ignorant prior approach is taken, starting with a uniform distribution between 1 .2 and 2.6 for each stream as priors, and this is modified to be consistent with the measured average density. Typically this distribution is truncated exponential (either negative or positive exponential) as derived immediately from information theory (Kullbach-Liebler divergence). It is noteworthy that even for this simple sub-problem information theory is not well-known in the mineral processing context. Estimating various exponential distributions at all streams is inconsistent due to mass balance issues (what goes into a unit must equal what comes out). So the estimated distributions are further adjusted to be consistent. This is basically an information theory entropy objective function with various constraints (measurements and mass balance constraints). However, without use of further constraints, use of this approach may be of limited value because of the uncertainty of the calculated distributions. Therefore a further prior is introduced: that the partition curve from one audit to the next is the same. A second set of densities is measured for each stream, and a further set of density distributions is measured that is consistent with the first, yet because of anural uncertainty, the estimated partition curves are not the same. These partition curves are now adjusted to be the same, and this partition curve is applied to the calculated feeds to estimate the new stream density distributions. Once again these are adjusted to be consistent with the average densities and the process iterates until convergence. In an embodiment, a method in accordance with the present disclosure comprises:

Step 1 . Take uniform density distributions for each stream as a prior.

Step 2. Adjust these density distributions using information theory to be consistent with known average densities, and known mass balance constraints.

Step 3. Calculate the partition curves for the two cyclones (or other processing units) over the two audits.

Step 4. Force equality of the cyclones (or other processing units) between audits, i.e.:

Cyclone 1 (Auditl ) =Cyclone1 (Audit2);

Cyclone 2 (Auditl ) =Cyclone2 (Audit2).

Step 5. Use partition curves applied to two feeds to provide updated density distributions for the various downstream streams.

Step 6. Repeat Steps 2-5 until convergence of results is achieved.

A further example of implementation of a method in accordance with the present disclosure is that is that with the prerequisite of detailed knowledge of the feed (which can be inferred from limited measurements in key streams) then it is possible to estimate the stream properties if only solid flow is measured.

Solid flows are normally measured with water flows; so the general term 'flow' is used here to encompass both solid flow and water flow.

In this example it is therefore possible to limit sampling to key streams and perform only flow measurement on other streams. This is sufficient to identify stream properties at these streams. This example may correspond to scenario 8, set out above.

By way of further explanation a detailed example is provided below.

This example relates to assessment of feeds from, and therefore operation of, a separation unit, for example a dense medium cyclone.

Two different feeds are input to the unit (e.g. cyclone) at different times. One feed is clearly 'heavier' than the other. Consequently, for the heavier feed there will be a greater flow rate at the heavies output.

The unit separation characteristics can therefore be determined from the flow rates of the heavies output.

This implementation uses a slightly complex version of the preceding maths. In this case the focus is two on major steps, taking the partition curve as ignorant prior.

Step 1 is to adjust the partition curve, given the particle density at the heavies outputs is unmeasured. Step 1 is substantially the same as previously outlined. The constraint is that the flow rates must correspond to the actual flow rates.

Step 2 is to estimate the partition curve given the calculated average density from step 1 and calculated size distribution in step 1 .

It might initially appear that step 2 could be redundant: however, this is not the case. It is in fact a subtle variation. It turns out that the partition curve estimated from step 2 is more density-dependent and size-dependent than that estimated in step 1 , and is therefore a more accurate representation of the real partition curve.

If a large number of different feeds were sampled then it could be expected that step 2 would not be required. However, step 2 is required when the number of feed is small; and is therefore consistent with the objective of providing accurate estimates with a decreased number of samples.

This implementation, using two sampled feeds, has been successfully tested. Figure 12 shows the actual partition curve 1201 compared with the estimated partition curve 1261 . As can be seen, the estimated partition curve 1261 is nearly identical to the actual partition curve 1201 .

Table 4, below, shows the various product stream cumulative density distributions.

Figures 13a and 13b illustrate an embodiment of software 1300 for performing an inference-based assessment method for estimating density distribution. These drawings also illustrate the flow of information in a method for estimating density distribution. The method is applied to each size-class independently.

The software 1300 comprises a first calculation module 1302, which receives data (block 1304) relating to average densities for each size, for each stream, for different audits, for example from a suitable database. The first calculation module 1302 estimates density distributions, using information theory, for each size for each stream, to provide initial estimated density distributions for each size, for each stream, for each audit (block 1306). The initial estimated density distributions (block 1306) are input (arrow 1308) to a main calculation unit 1350, which performs an iterative procedure providing iterated estimates of density distributions for each size, for each stream, for each audit (block 1306). The average densities (block 1304) are also used (arrow 1312) in each iteration. On each iteration new density distributions are calculated (that is, the estimates are updated). On convergence of the iterative procedure unit partition curves are also outputted (block 1314). The calculation procedures use similarity of the operational units (over different audits) to estimate density distributions; yet this is a prior and not an assumption; so the estimation of the final unit properties may indeed be different across different audits. The only input required is the average density (for each size) for each stream for each audit which, using information theory, is converted to initial density distributions (for each size, stream, audit). The main calculation unit 1350 is illustrated in more detail Figure 13b.

As illustrated in Figure 13b, the main calculation unit 1350 comprises a second calculation module 1352, which receives the initial estimated density distributions for each size, for each stream, for each audit (block 1306). The second calculation module 1352 performs mass balancing (optionally using information theory) to reestimate density distributions for each size for each stream. The mass-balance consistent density distributions for each size for each stream for each audit (block 1354) are received by a third calculation module 1356, which estimates unit behaviour for each unit for each size for each audit and identifies partition curves for each unit for each audit (block 1358). A fourth calculation module 1360 integrates the audit data for each unit and size, and calculates average partition curves for each unit and size over all audits (block 1362). A fifth, simulation, calculation module 1364 calculates a simulation for each set of audit data, using the partition curves and feed data to reestimate density distributions for each size for each stream. The simulation estimated density distributions (block 1366) and (as foreshadowed above) the average densities for each size, for each stream, for different audits (block 1304) are used by a sixth calculation module 1368 which calculates, using information theory, adjusted average densities for each size and stream and audit, independently. These adjusted average densities provide the next iteration of density distributions (block 1310) which is then used by the second calculation module 1352 as a basis for calculation of the subsequent iteration.

It will be appreciated that in Figures 13a and 13b the rectangular shapes represent data. The density distribution data at block 1306 and 1310 represent input and output; with output becoming input to the next iteration. The updated estimation of the partition curves 1358 is output when the iterative process is complete. For the final iteration, the process stops after calculating these unit partition curves.

Hence similarity of units over audits is used in the iterative process but not for final results.

One example embodiment of a hardware system for implementing the software, illustrated in Figure 14, comprises: at least one data collection and/or storage device 1401 for allowing input of data upon which the estimated mineral stream calculations are performed; a data processing apparatus 1402 for performing a method in accordance with the present disclosure; and at least one data output device 1403 for output of data relating to the estimated characteristics of the mineral stream.

The methodology uses information theory to estimate multimineral distributions. The simulation system uses this estimation to accurately simulate a mineral processing plant. This provides a more accurate simulation than known simulation systems which cannot resolve the multimineral information in a cost-efficient manner and therefore use simplified data structures, which provide a less accurate reflection of reality.

A simulation may be specifically designed for improving existing mineral processing operations, by identifying operational parameters and how changing these operational parameters affects the unit characteristics. The objective, in this case, is to identify whether operational changes will lead to improved performance, and consequent improved profit.

A separate application may act as a master software system, utilising optimisation strategies to identify the operational parameters that lead to the best profit. The simulation system may act as a client of the master system.

A database, for example an Access (Microsoft™) database, may be used to store data in an organised manner. This allows organised data and automated data retrieval, which assists considerably in implementing the calculation procedure. The calculation procedure is fully automated, and integrates data from successive audits via the Access database.

Figures 15a to 15e are sequential pages of a printout, provided as an example of code useful for the implementation described above with reference to the results illustrated in Figure 12 and Table 4. For this example it is presumed that particles are single-sized (for easier validation). The code is displayed in an upright font, set to the left of each figure. Comments to assist understanding of various lines and sections of code are provided in a bold italic font, generally positioned further to the right of each figure than the code.

The present disclosure relates largely to an inference (information theory) based assessment method. However, it will be appreciated that in minerals processing accurate assessment is an important basis process improvement. Many examples of how accurate assessment of particle stream characteristics, especially but not exclusively multimineral particle distributions, can lead to detection of sub-optimal processing steps in a mineral processing plant, and how improvements to the processing could be implemented to provide process improvements will be evident to the skilled addressee. By way of example, some specific examples of how accurate assessment can be used to implement process improvement are set out in Table 5, below, although many other examples will be apparent to the skilled addressee.

Notably, if all unit operations and feeds are known in detail the methodology in accordance with the present disclosure will allow operation of a mineral processing plant to be accurately simulated. This enables proposed process improvements to be simulated prior to implementation, and their effects, including effects on downstream units and processes, to be accurately predicted. This can reduce trial and error in implementation, and facilitate optimisation of plant operation.

The inference method allows for increasing process efficiency by using plant surveys over time. This capability has a wide range of applications. For example, if the mineralogy is known then the procedure can be conducted using mineral compositions rather than elemental compositions. The composition of mineral with respect to elements therefore provides a constraint for the mathematical solution.

The approach can be extended to density distributions. Particle properties are now density rather than elemental composition. The method will generally use as input average densities for each size-class.

The elemental compositions (assays) for each size-class provide constraints to the problem. The method could be applied to overall assays only. In this case it would not be possible to distinguish any variation with size. However the approach could be applied to bulk assays and size distributions; this would have less constraints, and therefore estimates of particle distributions would be less certain, but may well still be practical. Assessment of the accuracy/cost tradeoff, with regard to the needs of the mineral processing industry has not been completed, but it is apparent that if the particle distribution for each size can be estimated, it is also possible to estimate the assays in each size.

Hence by using this approach, the elements within sizes can be estimated from bulk assays and size distributions.

For timed audits, where time normally indicates that a different feed mineral ore is being processed, then time is not a specific variable.

With reference to Table 1 , an alternative way of viewing the larger process of which the assessment methods disclosed herein may form part is set out in Table 6.

2 Data Collection

3 Data Collation

4 Define operating variables

5 Perform mass balance

6 Perform inference

Perform model fitting (relate unit

7

behaviour to operating conditions)

8 Simulation

9 Optimisation

The objective discussed most in the present disclosure is to estimate stream properties or characteristics in detail (step 4 of Table 1 , and step 6 of Table 6). However, it should be appreciated that this serves as a means to identify how the ore properties are being processed by each unit, and therefore as a means to assess unit (and plant) operation.

The present disclosure has largely focused on the issue of presuming that unit behaviour is the 'same' from one audit to the next. It needs to be emphasised that this patent has focused on the benefits if unit behaviour is non-varying. However, this is not an assumption of practical implementation. In practical implementation unit behaviour may be adjustable according to various operational variables. In practise these operational variables will be varied; as will the unit behaviour.

The prior that a unit behaves the same at different times is not an assumption. If the unit does indeed behave the same then the method in accordance with the present disclosure will estimate that indeed the unit behaves the same. However, if the unit behaviour varies, then it is not possible for the solution to be that the unit behaviour has stayed the same; rather the results will indicate that the unit response behaviour has indeed changed. However there is some possibility that the 'full' adjustment of the unit behaviour may not be adequately reflected in the analysis. For example the analysis may be 90% correlated with the actual variation.

However there is sufficient correlation that further analysis (the model fitting stage, stage 7 as shown in Table 6) provides a sufficient estimate of plant behaviour to varying operating conditions (via the simulation, stage 8 and optimisation stage, stage 9, of Table 6)

Once the plant has been improved by varying the operational conditions, further analysis is made and the models are improved further. Of course, the same procedures as those set out herein can be applied to laboratory tests. For example, if particles are subject to laboratory flotation, and the flotation tests for one sample compared to another are exactly the same then the process can be applied to identify the unit properties, and feed properties (in terms of multimineral particles).

The present disclosure describes an assessment method for estimating the particle distributions of a mineral ore stream from general survey data. This capability is currently unavailable to the mining industry, rendering current interpretation of plant data difficult.

The above description is set out with reference to the processing of mineral ores into concentrated ore-rich streams. However, it will be appreciated that an assessment method in accordance with the present disclosure can be extended to other fields of process modelling or engineering, especially where there is a time varying process and the fundamental process information is not directly observable.

Reference to previously published articles or documents in this specification should not be taken as an acknowledgment or admission that the contents of the referenced articles or documents form part of the common general knowledge in the field of mineral processing in Australia or in any other country.

Modifications and improvements may be incorporated without departing from the scope of the disclosure hereof.

Claims

THE CLAIMS DEFINING THE INVENTION ARE AS FOLLOWS

1 . An assessment method comprising use of an information theory entropy objective function with various constraints to estimate at least one characteristic of a mineral particle stream of a mineral processing apparatus.

2. An assessment method according claim 1 , wherein at least one of the constraints comprises a characteristic of the mineral processing apparatus.

3. An assessment method according to claim 2, wherein at least one of the constraints comprises a relationship between feed and product streams of at least one operation unit of the mineral processing apparatus.

4. An assessment method according to any preceding claim, wherein at least one of the constraints comprises plant data relating to the mineral processing apparatus and/or a plant of which the mineral processing apparatus forms a part.

5. An assessment method according to claim 2, or any preceding claim when in accordance with claim 2, wherein at least one of the constraints comprises data relating to the number of feed and product streams of the mineral processing apparatus.

6. An assessment method according to claim 2, or any preceding claim when in accordance with claim 2, wherein at least one of the constraints comprises a related characteristic of a different stream of the apparatus.

7. An assessment method according to claim 2, or any preceding claim when in accordance with claim 2, wherein at least one of the constraints comprises a measurement of the characteristic.

8. An assessment method according to claim 2, or any preceding claim when in accordance with claim 2, wherein at least one of the constraints comprises mass balance data relating to the mineral processing apparatus.

9. An assessment method according to claim 2, or any preceding claim when in accordance with claim 2, wherein at least one of the constraints comprises a partition function of the mineral processing apparatus.

10. An assessment method according to claim 2, or any preceding claim when in accordance with claim 2, wherein at least one of the constraints comprises data relating to the operation of the mineral processing apparatus at a different time.

1 1 An assessment method according to any preceding claim, wherein at least one of the constraints comprises data relating to bulk information about a mineral at least some of which is included in the mineral particle stream.

12. An assessment method according to claim 10, wherein at least one of the constraints comprises data relating to the operation of the mineral processing apparatus at a plurality of different times.

13. An assessment method according to any preceding claim, wherein at least one of the constraints comprises data relating to the mineral processing apparatus and/or a plant of which the mineral processing apparatus forms a part, derived from a previously performed apparatus audit or plant audit.

14. An assessment method according to claim 13, wherein at least one of the constraints comprises data relating to the mineral processing apparatus and/or a plant of which the mineral processing apparatus forms a part, derived from a number of previously performed apparatus audits or plant audits.

15. An assessment method according to claim 14, wherein at least one of the constraints comprises data relating to the mineral processing apparatus and/or a plant of which the mineral processing apparatus forms a part, derived from two, and only two, previously performed apparatus or plant audits.

16. An assessment method according to claim 14, wherein at least one of the constraints comprises data relating to the mineral processing apparatus and/or a plant of which the mineral processing apparatus forms a part, derived from at least three previously performed apparatus or plant audits.

17. An assessment method according to any preceding claim, wherein the mineral particle stream of which a characteristic is estimated is an input feed of a mineral processing apparatus, and wherein at least one of the constraints comprises data regarding the operation of the mineral processing apparatus on an input feed of different composition to the input feed of the mineral particle stream of which a characteristic is estimated.

18. An assessment method according to claim 17, wherein the method estimates said characteristic by combining information from different feed ores, rather than treating each ore independently.

19. An assessment method according to any preceding claim, wherein at least one characteristic estimated by the method comprises a distribution of mineral particles in the mineral particle stream.

20. An assessment method according to claim 19, wherein at least one characteristic estimated by the method comprises a density distribution of mineral particles in the mineral particle stream.

21 . An assessment method according to claim 19, wherein at least one characteristic estimated by the method comprises a multimineral particle distribution of mineral particles in the mineral particle stream.

22. An assessment method according to claim 21 , wherein at least one characteristic estimated by the method comprises a multimineral particle density distribution of mineral particles in the mineral particle stream.

23. An assessment method according to any preceding claim, wherein at least one characteristic estimated by the method comprises a size distribution of mineral particles in the mineral particle stream.

24. An assessment method according to any preceding claim, wherein at least one characteristic estimated by the method comprises an assay within size-classes of mineral particles in the mineral particle stream.

25. An assessment method according to claim 24, wherein at least one characteristic estimated by the method comprises average density within size classes of mineral particles in the mineral particle stream.

26. An assessment method according to claim 25, wherein at least one characteristic estimated by the method comprises average density within size classes of mineral particles in the mineral particle stream estimated using size distribution and bulk average density within streams as constraints.

27. An assessment method according to any preceding claim, wherein at least one characteristic estimated by the method comprises average mineral composition of mineral particles in the mineral particle stream.

28. An assessment method according to claim 27, wherein at least one characteristic estimated by the method comprises average mineral composition of mineral particles in the mineral particle stream estimated using bulk assays and/or basic mineralogical information as constraints.

29. An assessment method according to any preceding claim, wherein at least one characteristic estimated by the method comprises multimineral particle composition of mineral particles in the mineral particle stream.

30. An assessment method according to claim 29, wherein at least one characteristic estimated by the method comprises multimineral particle composition of mineral particles in the mineral particle stream estimated using solid flow, size information, assays within sizes and limited mineralogical information as constraints.

31 . An assessment method according to any preceding claim, wherein at least one characteristic estimated by the method comprises a comminution breakage function of the mineral process apparatus.

32. An assessment method according to claim 2, or any preceding claim when in accordance with claim 2, wherein the method comprises taking a characteristic related to the operation of a minerals processing unit which acts on the product stream as a prior for use in providing a constraint in the information theory entropy objective function.

33. An assessment method according to claim 32, wherein the method comprises taking the characteristic related to the operation of the minerals processing unit as an ignorant prior.

34. An assessment method according to claim 32, wherein the method comprises taking a partition curve related to the operation of the minerals processing unit for use in providing a constraint in the information theory entropy objective function.

35. An assessment method according to claim 34, wherein the method comprises taking the partition curve as an ignorant prior.

36. An assessment method according to claim 2, or any preceding claim when in accordance with claim 2, wherein the method comprises taking consistency of operation of a minerals processing unit over time as a constraint.

37. An assessment method according to claim any of claims 14 to 16, wherein the method comprises acquiring data from an earlier audit to be used as the basis of a constraint, and taking at least part of the acquired data as a prior for use in providing a constraint in the information theory entropy objective function.

38. An assessment method according to claim 37, wherein the method comprises adjusting the prior to be consistent with corresponding more current measured data of the mineral particle stream.

39. An assessment method according to claim 38, wherein the method comprises calculating expressions for the acquired data and for the more current measured data which describe a characteristic to be estimated.

40. An assessment method according to claim 39, wherein the method comprises forcing equality of the calculated expressions.

41 . An assessment method according to either of claims 39 or 40, wherein the method comprises using the calculated expressions to provide two estimates for the characteristic to be estimated.

42. An assessment method according to claim 41 , wherein the method comprises repeating the steps of:

-adjusting the prior to be consistent with corresponding more current measured data of the mineral particle stream;

-calculating expressions for the acquired data and for the more current measured data which describe a characteristic to be estimated;

-forcing equality of the calculated expressions; and

-using the calculated expressions to provide two estimates for the characteristic to be estimated;

until convergence of the two estimates is achieved.

43. An assessment method according to any of claims 39 to 42, wherein calculating expressions for the acquired data and for the more current measured data comprises calculating partition curves for the mineral processing apparatus for the earlier audit and for the more current measured data.

44. An assessment method according to any of claims 41 to 43, wherein the acquired data and the more current measured data relate to particle distributions for the mineral processing apparatus.

45. An assessment method according to any of claims 41 to 44, wherein using the calculated expressions to provide two estimates for the characteristic to be estimated comprises applying these expressions to at least one mineral feed to be processed by the mineral processing apparatus to provide the mineral particle stream.

46. An assessment method according to any preceding claim, wherein the method comprises use of at least one estimated characteristic of a mineral particle stream of a mineral processing apparatus, which has been estimated by use of an information theory entropy objective function with various constraints, as a constraint in an information theory entropy objective function, to thereby estimate at least one further characteristic of the mineral particle stream.

47. An assessment method according to claim 46, wherein the method comprises use of at least two estimated characteristics of a mineral particle stream of a mineral processing apparatus, which have been estimated by use of an information theory entropy objective function with various constraints, as constraints in an information theory entropy objective function, to thereby estimate at least one further characteristic of the mineral particle stream.

48. An assessment method according to either of claims 46 or 47, wherein the estimated average density of particles in the mineral particle stream, having been estimated by use of an information theory entropy objective function with various constraints, is used as a constraint.

49. An assessment method according to any preceding claim, wherein the estimated size distribution of particles in the mineral particle stream, having been estimated by use of an information theory entropy objective function with various constraints, is used as a constraint.

50. An assessment method according to any preceding claim, wherein the method is implemented by a computer.

51 . A system for assessing a mineral particle stream of a mineral processing apparatus comprising processing apparatus programmed with instructions for processing the acquired data to implement an assessment method comprising use of an information theory entropy objective function with various constraints to estimate at least one characteristic of a mineral particle stream of a mineral processing apparatus.

52. A system for assessing a mineral particle stream of a mineral processing apparatus comprising:

-data acquisition apparatus for acquiring data for use in an assessment method;

-processing apparatus programmed with instructions for processing the acquired data to implement an assessment method comprising use of an information theory entropy objective function with various constraints to estimate at least one characteristic of a mineral particle stream of a mineral processing apparatus; and -data output apparatus for output of data relating to the estimate of the at least one characteristic.

53. A system according to either of claims 51 or 52, wherein the assessment method is in accordance with any of claims 1 to 50.

54. A method of providing improved operation of mineral processing apparatus comprising using of a method in accordance with any of claims 1 to 50, determining how the estimated characteristic can be changed to provide improved performance by changing operation of the mineral processing apparatus, and implementing changes in operation to provide improved mineral processing apparatus operation.

55. A method according to claim 54, wherein implementing changes in operation comprises implementing changes in operation of the mineral processing apparatus comprising the mineral particle stream of which the characteristic was estimated.

56. A method according to either of claims 54 or 55, wherein implementing changes in operation comprises implementing changes to a mineral processing operating unit which is not directly connected to the mineral particle stream of which the characteristic was estimated.

57. A method according to claim 56, wherein the mineral particle stream of which the characteristic was estimated is a product stream of a separating apparatus, and the method comprises implemented changes in comminution of the feed to the separating apparatus.

58. A method according to any of claims 54 to 57, wherein implementing changes in operation comprises providing new mineral processing apparatus in which changes are implemented.

59. A physical medium comprising computer executable instructions for operating a computer to perform a method in accordance with claim 50.

60. An assessment method for a time varying process in which the fundamental process information is not directly observable comprising use of an information theory entropy objective function with various constraints to estimate at least one characteristic of the fundamental process information which is not directly observable, wherein at least some of the constraints are provided by measuring a characteristic of the process, which is related to the characteristic to be estimated, at a number of different times.