PRODUCT QUALITY ESTIMATION METHOD AND SYSTEM

TECHNICAL FIELD

This invention relates to a system and method for estimating the quality of a product. In particular the present invention may be used to estimate the quality of a product after the expiry of a period of time, or at a future time. Reference throughout this specification will also be made to the present invention being used to estimate product lifetime for perishable products (and in particular raw food products), but those skilled in the art should appreciate that other applications are also envisioned.

BACKGROUND ART

The spoilage of perishable goods is an area of concern for producers and consumers alike. In the case of perishable food products producers need to ensure that food products have not degraded in quality over time and preferably are safe for consumers to eat. Such perishable goods are normally marked with a use by, best before or shelf life date to ensure that they are not consumed outside of a timeframe in which quality can be guaranteed with confidence.

The assessment of effective shelf life is an important area for producers who are interested in transporting perishable goods over comparatively long transit times. Transit times eat into the shelf life of the goods, which are normally subjected to quality testing and food safety assessments on arrival at their destination.

Producers and distributors of such perishable goods need to ensure that the quality of the goods do not degrade significantly during transit. If such degradation is likely then there is the option for the producer to release the product onto an alternative market, as opposed to a geographically remote market. However, in some instances this decision to divert must be made while the goods are in transit

to allow enough time to reach the alternative market with shelf life time remaining. At present such shipments are inspected at their destination, and if they have degraded to an unacceptable level, these goods are only then shipped on at this time.

In the case of perishable food products, mathematical modelling techniques have previously been used to assess the size of micro-organism populations present after a period of time. These existing modelling techniques generally use microorganism plate counts to model the growth characteristics of organisms on growth medium in controlled laboratory conditions. The environment in which the culture medium is grown is controlled to allow for the modelling of growth characteristics of micro-organisms to be tracked and understood. Such laboratory based studies control the variables normally which effect the growth rate of such organisms and are generally completed in relation to a single target organism only.

These organism growth models can accurately predict the growth of a single organism on a broth culture within a controlled laboratory environment. However, there are additional variables at play when the actual conditions of a perishable food product are considered within its own production and distribution line.

Such existing modelling techniques do not take into account the composition of organism populations co-existing on a product. With variations in the seasons of the year, the composition of these populations can vary which in turn can affect the overall count of organisms present and the population size of particular types of harmful organisms which can have an effect on the resulting quality of the food product.

Furthermore the production and distribution lines of such perishable products may be subject to unique environmental variables or factors which are not normally present within a laboratory environment. In the case of meat processing lines

particular processing sites can have unique local micro-organism populations with varying compositions, driven by the humidity, sunlight hours or local flora and fauna present within the region involved. Although such local factors may not necessarily contribute to a significant variation from laboratory standard conditions, these factors may still have an effect on the accuracy of shelf life projections made using such existing techniques.

The same issues are also present with respect to distribution lines for perishable products. Variations in the local environments of vehicles used to transport such goods in addition to the transit routes taken can contribute to providing a unique local environment for the products and their associated micro-organism populations.

An improved product quality estimation method, and a system for executing such a method would be of advantage over the prior art. In particular, a quality assessment method and system which could be tailored to take into account unique local environmental variables to improve the accuracy of quality predictions would be of advantage.

All references, including any patents or patent applications cited in this specification are hereby incorporated by reference. No admission is made that any reference constitutes prior art. The discussion of the references states what their authors assert, and the applicants reserve the right to challenge the accuracy and pertinency of the cited documents. It will be clearly understood that, although a number of prior art publications are referred to herein, this reference does not constitute an admission that any of these documents form part of the common general knowledge in the art, in New Zealand or in any other country.

It is acknowledged that the term 'comprise' may, under varying jurisdictions, be attributed with either an exclusive or an inclusive meaning. For the purpose of this

specification, and unless otherwise noted, the term 'comprise' shall have an inclusive meaning - i.e. that it will be taken to mean an inclusion of not only the listed components it directly references, but also other non-specified components or elements. This rationale will also be used when the term 'comprised' or 'comprising' is used in relation to one or more steps in a method or process.

It is an object of the present invention to address the foregoing problems or at least to provide the public with a useful choice.

Further aspects and advantages of the present invention will become apparent from the ensuing description which is given by way of example only.

DISCLOSURE OF INVENTION

According to one aspect of the present invention there is provided a method of estimating the quality of a product at a set time characterised by the steps of;

(i) defining a model for the change in quality of a product over time, and

(ii) calculating at least one quality estimate distribution for the product at a set time using the model defined, and

(iii) recording at least one corresponding quality measurement distribution at said set time, and

(iv) determining if the model is valid based on a difference between at least one quality estimate distribution and it's associated quality measurement distribution, and

(v) reconfiguring the model using at least one recorded quality measurement distribution if the model is determined to be invalid.

According to a further aspect of the present invention there is provided a method of

estimating the quality of a product substantially as described above further characterised by the additional subsequent step of;

(vi) calculating at least one quality estimate distribution for the product at a set time using the reconfigured model.

According to yet another aspect of the present invention there is provided a method of estimating the quality of a product substantially as described above further characterised by the additional subsequent step of;

vii) repeating steps (ii) through (v) as further quality estimate distributions and quality measurement distributions are calculated and recorded.

According to a further aspect of the present invention there is provided computer executable instructions adapted to implement a method of estimating the quality of a product at a set time, said instructions being configured to execute the steps of;

(i) defining a model for the change in quality of a product over time, and

(ii) calculating at least one quality estimate distribution for the product at a set time using the model defined, and

(iii) receiving at least one corresponding quality measurement distribution at said set time, and

(iv) determining if the model is valid based on a difference between at least one quality estimate and it's associated quality measurement distribution, and

(v) reconfiguring the model using at least one recorded quality measurement distribution if the model is determined to be invalid.

According to a further aspect of the present invention there is provided computer executable instructions substantially as described above, said instructions being

configured to execute the additional step of;

(vi) repeating steps (ii) through (v) as further quality estimates and quality measurement distributions are calculated and received.

According to yet another aspect of the present invention there is provided a method of estimating the quality of a product at a set time characterised by the steps of;

(i) obtaining an initial quality description of the product and initialising a model for the change in quality of the product using said initial quality description, and

(ii) subsequently recording over time a plurality of initial quality measurements, and

(iii) comparing the initial quality description with said plurality of subsequently recorded initial quality measurements to determine if said initial quality description is valid, and

(iv) obtaining a new quality description and re-initialising said model if the initial quality description is deemed to be invalid, and

(v) calculating at least one quality estimate distribution for the product at a set time using the model, and

(vi) recording at least one corresponding quality measurement distribution at said set time, and

(vii) determining if the model is valid based on a difference between at least one quality estimate and it's associated quality measurement distribution, and

(viii) reconfiguring the model using at least one recorded quality measurement distribution if the model is determined to be invalid.

According to yet another aspect of the present invention there is provided a method of estimating the quality of a product, wherein said method includes the further step of repeating steps (ii) through (iv) over time as subsequent additional initial quality measurements are recorded.

According to yet another aspect of the present invention there is provided a method of estimating the quality of a product, said method being further characterised by the additional subsequent step of repeating steps (v) through (viii) as further quality estimates and quality measurement distributions are calculated and recorded.

According to yet another aspect of the present invention there is provided computer executable instructions adapted to implement a method of estimating the quality of a product at a set time, said instructions being configured to execute the steps of;

(i) obtaining an initial quality description of the product and initialising a model for the change in quality of the product using said initial quality description, and

(ii) subsequently receiving over time a plurality of initial quality measurements, and

(iii) comparing the initial quality description with said plurality of subsequently received initial quality measurements to determine if said initial quality description is valid, and

(iv) obtaining a new quality description and re-initialising said model if the initial quality description is deemed to be invalid, and

(v) calculating at least one quality estimate distribution for the product at a set time using the model defined, and

(vi) receiving at least one corresponding quality measurement distribution at said set time, and

(vii) determining if the model is valid based on a difference between at least one quality estimate and it's associated quality measurement distribution, and

(viii) reconfiguring the model using at least one recorded quality measurement distribution if the model is determined to be invalid.

The present invention is adapted to provide a method of estimating the quality of the product at a set time. Preferably this method may be implemented by a system or apparatus formed by a computer system or systems programmed with appropriate executable instructions. The present invention may therefore encompass a method of estimation, computer executable instructions adapted to complete such a method, and a computer system loaded with such computer executable instructions. In the main reference throughout this specification will however be made to the present invention being provided through a method of estimating quality, but again those skilled in the art should appreciate that other aspects of the present invention are also within its scope.

The present invention may preferably be used to estimate the quality of a product after the expiry of a set period of time. Preferably a product with which the present invention may be used may consist of a product which degrades in quality over time.

In a further preferred embodiment the present invention may be used to estimate the quality of a food product after a set period of time. Food products are well known for their perishable characteristics and are also transported over

comparatively long transit times, and require monitoring for any degradation in quality of the product.

Perishable foods, as discussed throughout this specification, should be taken to encompass any food or beverage which may degrade in quality over time. However those skilled in the art should appreciate that other types of perishable products may also be considered through use of the present invention. For example in one alternative embodiment the viability of agricultural seed (which again degrades with time) may also be considered using the present invention, and reference to food products only in isolation throughout this specification should in no way be seen as limiting.

Reference in general throughout this specification will also be made to the present invention being used to estimate the quality of a product. Products referred to throughout this specification can encompass both a single instance of an article, or a collection of products or articles making up a single shipment. Those skilled in the art should appreciate that references to a product in general can also encompass an entire shipment of individual product articles through to a single instance of such an article. The present invention may perform equally well in assessing quality of a single product, or a shipment of articles.

In a preferred embodiment the characteristic associated with the quality of a product may be defined through numeric population values of micro-organisms found on a surface of a food product. Such population counts used to indicate quality may be general in nature and span a count of all micro-organisms - irrespective of type. Alternatively such counts may be targeted to consider only micro-organisms known to have a serious effect on the health of a person ingesting food products. Such population counts may provide a direct measurable indicator as to the quality of a product, with a rise in population levels over time above a certain threshold indicating that the food product is spoiled.

Reference throughout this specification will also be made to the quality of a product being estimated by population counts of microbes living on a food product. However those skilled in the art should appreciate that other parameters or variables which can be used to indicate the quality of a perishable product may also be employed in conjunction with the present invention, and reference to the above only throughout this specification should in no way be seen as limiting.

However, in alternative embodiments chemical compound or composition tests may be used to, for example, determine the level of browning of red meat through oxidative damage, ethylene levels associated with the ripening of fruit, enzyme activity levels, or the monitoring of environmental factors which may control the germination rates of seeds. Those skilled in the art should appreciate that population counts of microbes as discussed throughout this specification should in no way be seen as limiting. The present invention may also be used in conjunction with a number of different forms of perishable products.

A preferred embodiment may employ distributions of population data, both in the definition of the model to be employed, and also in the resulting output provided by such a model. Distributions of results can be considered to obtain a clear view of the current state or quality of a product as illustrated by the completion of a number of sample measurements. The overall shape of such a distribution when plotted to show values measured or estimated against the number of time these values occur gives an indication of the amounts of a particular product or group of products which may be in danger of degradation. Such information can be employed to make pre-emptive decisions regarding the markets to which such products may be directed depending on the transit times required and also the quality acceptance standards required for each market.

Preferably the present invention is adapted to estimate the quality of a product through the use of a modelling technique. A model of the growth characteristics

and expected microbe population levels may be used to estimate or predict population levels at a future or set time. Such a predictive model may receive as an input an actual measurement value taken at an initial time and be subsequently used to estimate population levels at a further future time.

In a further preferred embodiment the present invention may employ a stochastic population model. Stochastic population models are well known in the mathematical modelling field and can provide output frequency distributions for a future time period when supplied with an initial input frequency distribution and environmental parameter data.

Those skilled in the art should appreciate that such existing mathematical models may receive as inputs environmental parameters recorded during the time which expires between the initial population measurement and the subsequent calculation of a population estimate. For example, in one instance a temperature logging apparatus may travel with a food product and record variations in the temperature of the environment in which the product is housed.

Preferably the first step in the method of estimation provided may be to define such a model which may predict the expected change in quality of a food product over a known period of time. In this respect the present invention may build on existing prior art quality estimation techniques which again use such models to make a basic estimation of quality. The present invention may initially start with such laboratory database models, and as discussed below can employ further improvements to assist in the provision of more accurate estimates of quality.

According to one further aspect of the present invention there is provided a method of defining a model for the change in quality of a product over time characterised by the steps of;

i) obtaining an initial quality description of a product and initialising a model for the change in quality of the product using said initial quality description, and

ii) subsequently recording over time a plurality of initial quality measurements, and

iii) comparing the initial quality description with said plurality of subsequently recorded initial quality measurements to determine if said initial quality description is valid, and

iv) obtaining a new quality description and re-initialising said model if the initial quality description is determined to be invalid.

According to a further aspect of the present invention there is provided a method of defining a model substantially as described above further characterised by the additional subsequent step of

v) repeating steps (ii) through (iv) over time as subsequent additional initial quality measurements are made.

According to a further aspect of the present invention there is provided computer executable instructions adapted to implement a method of defining a model for the change in quality of a product over time characterised by the steps of;

(i) obtaining an initial quality description of a product and initialising a model for the change in quality of the product using said initial quality description, and

(ii) subsequently receiving over time a plurality of initial quality measurements, and

(iii) comparing the initial quality description with said plurality of subsequently recorded initial quality measurements to determine if said initial quality description is valid, and

(iv) obtaining a new quality description and re-initialising said model if the initial quality description is deemed to be invalid.

Preferably the present invention facilitates the provision of both a method of defining a model for the change in quality of a product, and also the utilisation of such a model with the potential to re-configure this model if required. The initial definition of such a model may be completed to preferably reflect the initial conditions of a product or the conditions present in a specific production facility for a product. Through subsequent use of this model and through feedback provided with respect to the output of the model this further reconfiguration process may also be completed.

Preferably such a model may be defined through obtaining an initial quality description of a product. This initial quality description may preferably indicate the starting condition expected of a product produced by a specific production facility. This predefined starting point in turn allows the model to estimate changes to be experienced over time in such a product based on the time frames involved and the environment in which the product is located.

In a preferred embodiment an initial quality description of a product may be composed from or incorporate a distribution of micro-organisms population values.

For example, in one embodiment a plurality of samples may be taken from the product where measurements of population levels are to provide the distribution required to give an initial quality description. In other embodiments such an initial quality description may be provided by an assumed population distribution for the product type in question.

In yet other embodiments this initial quality description distribution may be estimated through the output of a model provided in accordance with the present invention. In such estimation focused applications, the output of a model provided in accordance with the present invention may in turn be used to set up or provide the initial quality description for a further model for the change in quality of a product over a future time period.

In yet other embodiments a combination of estimated and measured population distributions may be employed to provide such an initial quality description. For example, in some instances a limited number of actual population measurement values may be available and therefore may provide only a sparse distribution of results. Such a sparse set of measured values may be combined with a model estimated distribution to provide the initial quality description required.

Preferably the present invention may also allow for the subsequent recording over time of a plurality of initial quality measurements once a model has been initialised with an initial quality description. These initial quality measurements may preferably be made from or by a plurality of micro-organism population count values which reflect a distribution of such values recorded at specific points in time. Preferably a number of such population value distributions may be recorded sequentially over time to monitor the actual initial quality description of a product produced by a specific production facility. Such subsequent initial quality measurements may be made to determine whether the initial quality description used to initialise the model is in fact representative of the product to be modelled.

To achieve this objective a comparison may be made between the initial quality description used to initialise the model and the plurality of subsequently recorded initial quality measurements. A number of these initial quality measurements are used in such a comparison to ensure that random errors or noise present in such measurements do not result in the initial quality description being discarded

unnecessarily. Conversely, if a plurality of such initial quality measurements reflect a consistent trend, bias or offset between measurements and the initial quality description, the situation will reflect that the initial quality description will need to be modified.

In a further preferred embodiment this comparison between the initial quality description and subsequently recorded initial quality measurements may be completed through a cumulative sum function. In such embodiments a pair of positive and negative threshold drift levels may be defined for the initial quality description, and numeric differences between the initial quality description and subsequent initial quality measurements can be recorded and summed consecutively as each new initial quality measurement becomes available. This cumulative sum function will then provide a result which will trend towards and eventually exceed one of the threshold drift levels if the initial quality description used does not effectively represent the current initial state of the product involved.

Reference throughout this specification will also be made to such a comparison being made through a cumulative sum or CUSUM function. However, those skilled in the art should appreciate that a range of quality assurance techniques may be employed in alternative embodiments to complete such a comparison. Standard quality assurance statistical techniques (such as for example, a T-Test) which can compare and determine whether two population distributions are similar or are not may also be employed.

The above approach may be used in the definition of a model to ensure that the initial description of a product stays valid over time. Through repetitively comparing this initial description with actual initial measurements made, confirmation may be obtained that the model initialised still closely tracks real world conditions associated with the initial state of the product.

Preferably if this comparison between the initial quality description and the subsequently recorded initial quality measurements determines the description is invalid a replacement description may be obtained and the model re-initialised. Essentially the same process may be followed to obtain a new quality description as that considered with respect to obtaining the original quality description, although any original assumptions made could be reconsidered in light of an original quality description being discarded.

In a further preferred embodiment one or more of the more recent initial quality measurements made may be employed to provide the new quality description required to re-initialise the model. These more recent initial quality measurements may in such instances provide a more current and accurate description of the initial condition of a product to be modelled in conjunction with the present invention.

Preferably once this model is defined it can be used to calculate one or more quality estimates at a set future time. These quality estimates may be generated from the model defined through its initial quality description associated with an initial time prior to the set future time of interest. The model may then be used to provide a quality estimate population distribution for this future set time.

The present invention may also employ a recording of a corresponding quality measurement made at the set time for which a quality estimate was calculated. Such quality measurements may be recorded through a plurality of directly measured population levels of micro-organisms present on a food product to provide a corresponding distribution to that estimated by the model.

In a preferred embodiment a quality estimate distribution may be calculated for each and every corresponding quality measurement distribution recorded and available for use with the present invention. In the case of food products being monitored, the limiting factor involved may be the availability of actual quality

measurement recordings, as opposed to the ability to calculate quality estimates from the model defined.

In a further preferred embodiment paired sets of quality estimate and quality measurement distributions may be collected in relation to a number of individual food products sourced from the same production line and/or distribution channel.

These paired measurements and estimates can be taken for the same set time, preferably after the expiry of a transit time for a product to its eventual market.

Preferably a plurality of paired sets of quality estimate and quality measurement distributions may be obtained over numerous shipments and production batches to assess whether the original current model defined is valid based on actual measured population values.

In a preferred embodiment a determination may be made as to the validity of the initial model used based on investigating any differences between the values of at least one paired set of quality estimate and measurement value distributions. Differences in these paired sets of results can indicate failure of the initial model to accurately or correctly track the growth characteristics of the micro-organism populations present.

Those skilled in the art should appreciate that whether the model is defined as valid or invalid will depend on the particular circumstances and application in which the present invention is employed. For example, in some instances a model may be deemed to be invalid simply through having a less accurate set of predictions or estimates than that which could be provided through reconfiguring the model as discussed below. In most instances the end users tolerance for accuracy will in turn affect the determination as to whether the model used is deemed to be valid. Furthermore, such validity thresholds will also determine the frequency at which a model is reconfigured again depending on the requirements of the end user. The present invention may allow for the rapid reconfiguration of the model in some

instances, whereas in others the model may be reconfigured periodically to track gradual changes in environmental conditions.

In a preferred embodiment a collection of paired estimate and measurement frequency distributions may be assessed to determine whether the model involved is invalid. In a further preferred embodiment assessment or updating technology may be based on Bayesian statistical methods such as a particle filter.

For example, in a preferred embodiment an algorithm may be used, such as described in detail in the publication; Kotecha and Djuric 2003. Gaussian Particle Filtering. IEEE transactions on signal processing 51 : (10), 2592 - 2601.

Those skilled in the art should appreciate that a number of Bayesian statistical techniques may be employed to provide a statistical measure as to how well two distributions match with one another, and therefore an estimate as to whether the model involved is invalid.

Preferably, if the initially model defined is deemed to be invalid then a reconfiguration process may be executed on the model. This reconfiguration process may use as an input at least one of the recorded quality measurement distributions, and aim to reconfigure or modify the structure and operation of the model to improve the accuracy of its estimates in future.

In a preferred embodiment the model defined may be reconfigured through the application of a mathematical filter to its initial output. Preferably in such embodiments Bayesian probability techniques may be employed to provide such a filter, and therefore the reconfigured model required.

For example in some preferred embodiments a grid filter or a particle filter may be employed to provide the reconfigured model. A grid filter may be employed in instances where only a relatively small number of parameters are provided as

environmental variable inputs to the model, thereby allowing the reconfigured mode! to execute without a heavy computational load placed on the computer systems required to run same. Conversely the particle filter discussed above may be employed in relation to models with a relatively large number of inputs but will in turn place an increased computational load on the computer system used. Those skilled in the art should appreciate that an appropriate filter may be selected for the particular application in which the present invention is to be used.

Preferably once a model defined as invalid is reconfigured, it may subsequently be used to provide estimates of population count distributions at set times in the future. Furthermore, the same process as discussed above may be repeated as further future quality measurement values and their associated distributions are obtained to ensure that the model remains valid, and that it keeps pace with any fluctuations or changes in the environment in which a product is produced or transported.

This technique may be used to reconfigure or update such an initial model with feedback from actual measurements taken from a production line or distribution channel. In effect in preferred embodiments the present invention may operate as an iterative, repetitive process to allow the model provided to evolve over time to match and model the current environment of a food product.

Such a periodic or potentially frequent reconfiguration process can allow for the provision of a model which takes into account recent changes in the environment of a product, or alternatively gradual seasonal environmental trends. The technique used to assess the validity of the model can therefore be tailored to particular applications to detect such gradual changes or trends in addition to significant immediate variations between measured and estimated quality value distributions. This approach can be used to detect significant quality problems which may have recently appeared in a products production line or distribution

chain.

The present invention may also be used in the case of food products to target or identify population level distributions for specific target organisms. Such target organisms may have a detrimental effect on the quality and potential safety of a food product and are therefore of interest to both consumers and food regulation authorities. In such instances an initial species specific model may be employed to target such an organism, with the reconfiguration process discussed above allowing the initial specific model to evolve. The evolving model can take into account the effects caused by local environmental conditions and in particular composition changes in the make-up of other organisms which the target organism shares the food product with.

Alternatively, the present invention may be provide broad estimates of general population count distributions, irrespective of particular target species if required.

The present invention may also be used in a procedure analysis role to assess whether any recent changes in the operation of a production line or distribution chain have an effect on the expected quality of a product. If recent changes have occurred and an immediate significant variation between estimated and measurement quality is detected, such procedures may be investigated further to determine their effect on quality. The same approach may also be taken to analyse new procedures to determine whether they are improving the quality of the resulting product.

The present invention may also provide additional advantages in terms of tracking and tracing the origin of particular goods which are susceptible to degradation over time. Preferably both quality estimate distributions provided in accordance with the present invention, and their associated quality measurement distributions may be recorded in relation to particular products or shipments of products to be made

available to the end users or any other interested parties. Such information can clearly illustrate quality focused characteristics of the product to reassure consumers.

The present invention may also be employed as a decision making aid to determine which markets perishable products should be directed to. In some instances telemetry systems may be provided to transmit environmental parameter data while a product is in transit to a selected destination market. If the model of the present invention shows significant degradation during transit, the product may be redirected to an alternative market while still in transit. A similar approach may also be employed in stand alone isolated data logging configurations of the invention which provides such environmental parameter data once a product shipment reaches its destination.

BRIEF DESCRIPTION OF DRAWINGS

Further aspects of the present invention will become apparent from the following description which is given by way of example only and with reference to the accompanying drawings in which:

Figure 1 shows a block schematic flow chart of steps executed by a method of quality estimation provided in accordance with a preferred embodiment;

Figure 2 shows a table of quality estimate results provided over time in conjunction with a further embodiment of the invention.

Figure 3 shows a block schematic flowchart of steps executed by a method of defining and using a model in accordance with a further embodiment of the invention.

BEST MOPES FOR CARRYING OUT THE INVENTION

Figure 1 shows a block schematic flow chart of steps executed by a method of quality estimation provided in accordance with a preferred embodiment.

The first stage of the methodology discussed is shown in step (A) where a model is defined for the change in quality expected for a product over time. This basic initial model may be formed to provide quality estimation distributions using recordings of environmental parameters over the period of time in question.

The second stage (B) of this process involves an initial calculation made of a quality estimate, preferably formed by a microbe population frequency distribution at a set time. The model defined in step (A) is used to calculate this estimated population count distribution based on an initial measurement of the population distribution at a starting time and a recording of temperature variations during the expiry of the time period under investigation.

The following stage (C) of this process involves a recording of a corresponding quality measurement distribution from actual micro-organism population counts at the set time for which a population estimate distribution was calculated. The recorded measurements provide a comparison between actual and estimated quality values to be used to determine whether the model used is valid.

Those skilled in the art should appreciate that stage (B) and (C) can also encompass the collation or buffering of multiple paired sets of quality estimate and measurement distributions over time before the following stage (D) is executed. Alternatively stage (D) may be executed once a single paired set of estimate and measurement value distributions are obtained.

This following stage D is executed to determine whether the model defined initially is valid. This validation step is completed through making a comparison between

estimated versus measured quality value distributions and in a preferred embodiment estimated population level distributions versus measured population level distributions for micro-organisms on a food product.

Preferably this validity assessment may be made through the use of Bayesian statistical techniques employed to compare two frequency distributions with one another. User defined tolerances which specify how different these two distributions are can then be employed to make a decision as to whether the model is valid or invalid.

If the model is determined to be valid then stages (B) and (C) are repeated for subsequence future production and distribution runs of the product. Conversely if the model is determined to be invalid stage (E) of this process is executed.

At stage (E) the initial model is reconfigured using one or more of the recorded quality measurements provided at stage (C). Preferably a Bayesian probability technique is employed to provide a filter to the output of the basic model defined and in the case of a preferred embodiment, modify a pair of control parameters adjusted to in turn reconfigure the model. These specific changes made to two control parameters (A, B) are illustrated by way of example.

Once the model has been reconfigured it can continue to be used in a role of estimated the quality of a product through repeating steps (B) through (D) and possibly stage (E) if in future the model is deemed to be invalid. The feedback provided through actual measurements at stage (C) ensures that reconfiguration of the model will allow the models output to track the actual conditions experienced by the product under investigation.

Figure 2 shows a table of quality estimate results provided over time.

The table of figure 2 illustrates the action of a particle filter employed in a preferred

embodiment to reconfigure a model by changing the control parameters A and B. As can be seen from figure 2 the predicted measurements of the model are gradually pulled into line with actual measurements over 12 separate reconfiguration operations. Each of the control parameters A, B are adjusted gradually and in combination to ensure that the predicted measurement provided is pulled closer into alignment with actual measurements obtained.

Figure 3 shows a block schematic flowchart of steps executed by a method of defining and using a model in accordance with a further embodiment of the invention. In particular figure 3 elaborates further on the actions taken with respect to the method discussed and illustrated with respect to figure 1.

The iterative process illustrated with respect to figure 3 can be instigated through obtaining initial product quality data (such as for example microbe counts) which in turn are used to generate an initial frequency distribution. This initial frequency distribution is used as an initial quality description of a food product.

As can be seen from the iterative process illustrated by figure 3, the initial product data is recorded a number of times to provide a plurality of initial quality measurements over time. This iterative looping process will then build up a collection of such product quality data to allow for the generation and monitoring of a CUSUM or cumulative sum chart. The CUSUM chart is updated each time product quality data becomes available and is used to compare the initial quality description defined with the subsequently recorded product quality data.

The updated CUSUM chart is used to test if this initial frequency distribution has changed and if so this will trigger a replacement or change being made to the initial frequency distribution used.

Once an appropriate initial quality description and distribution has been arrived at it can be used to define a stochastic population model which reflects the change in

quality of the food product involved.

Next, environmental parameter data such as temperatures, pH levels and so forth can be collected and the model defined can be used to subsequently predict a frequency distribution estimate for the future time at which a product is to have its quality tested. This acceptance testing is generally completed at a destination point for exported products. Acceptance testing allows for the collection of acceptance data sourced from samples taken from the product at the time at which the model was used to prepare a quality estimate distribution.

Once the acceptance testing data is available a comparison may next be made between the frequency distribution of the acceptance data and the estimated frequency distribution prepared using the model. This comparison allows for a determination as to whether the model is invalid. A model determined to be invalid is reconfigured through having its parameters updated using Bayesian statistics in the example discussed with respect to figure 3.

At the end of this process the model is either validated, or reconfigured and is then ready to use again. With the iteration of this process the initial frequency distribution used to define the model is tested and also re-initialised if required. The actual output of the model is again tested with every iteration through the acceptance data collected, allowing the model to again be tested for a second time during each iteration.

Also as illustrated with respect to figure 3, the output of the model provided may be employed to seed an initial quality description for a further model which focuses on changes in quality of the product after the acceptance sampling discussed above. For example, in one embodiment both the acceptance sampling data and also the predicted or estimated frequency distribution can be employed to provide an initial frequency distribution of a new model.

Aspects of the present invention have been described by way of example only and it should be appreciated that modifications and additions may be made thereto without departing from the scope thereof as defined in the appended claims.