AU2013364503B2

AU2013364503B2 - A method and system for determining milk characteristics for individual animals in a herd

Info

Publication number: AU2013364503B2
Application number: AU2013364503A
Authority: AU
Inventors: Anthony Bryan Pleasants; Paul Robert Shorten
Original assignee: AgResearch Ltd
Current assignee: AgResearch Ltd
Priority date: 2012-12-20
Filing date: 2013-12-20
Publication date: 2017-06-22
Anticipated expiration: 2033-12-20
Also published as: WO2014098627A1; AU2013364503A1

Abstract

The invention relations to a method of determining milk characteristics for individual animals within a cohort of milking animals, where the method is characterised by the steps of: a) recording individual identification of animals present at a milking shed for a milking event; b) recording the timing of the milking event for individual animals; c) measuring at least one milk characteristic of milk collected from the cohort of milking animals; d) correlating the identification of an individual animal with the timing of the milking event to develop a profile of one or more milk characteristics for the individual animal. A system for implementing the method is also claimed. The inventive method and system has particular application for determination of characteristics of milk from a selected animal from a group of animals, particularly, but not limited to, milk yield, based on measurement of an ensemble of milk production from a group of animals, such as a dairy herd or cohort of dairy cows.

Description

PCT/NZ2013/000244 WO 2014/098627

A METHOD AND SYSTEM FOR DETERMINING MILK CHARACTERISTICS FOR INDIVIDUAL ANIMALS IN A HERD

TECHNICAL FIELD

The invention relates to a method and system for determining milking characteristics of 5 individual dairy animals within a herd of dairy animals. The invention has particular application to estimating the milk yield and/or other milking characteristics of specific cows of a herd using measurements of the total milk collected from the herd.

BACKGROUND ART

Milk yield, the quantity of milk obtained from an animal within a specified time frame, is 10 important information for a dairy farmer. It can form the basis for decisions pertaining to management of a dairy herd or selected animals within the herd. Such decisions can extend to when animals may be culled (usually the lower producing animals) or dried off (allowing diversion of high quality feed, which contributes to good milk yield, to animals still producing good quantities of milk). 15 These decisions help to optimise the milking performance of the herd and increase the subsequent economic return.

Some aspects of an animal’s condition can be assessed by analysing the quantity and/or quality of the milk from the animal. In some instances, changes in condition of an animal are manifested in the milk before there is any readily observable physical 20 manifestation of the change.

Therefore, measurement of milk yield and other milk characteristics can be an important aspect of dairy farm management.

On many farms, it is common to use milking meters in association with the milk lines of 1 PCT/NZ2013/000244 WO 2014/098627 a milking shed. When used in conjunction with identification tags associated with the cow, the milk yield, and optionally, other characteristics of milk of a specific cow can then be measured.

For accurate measurement of milk yield it is preferable that each stall of the milking 5 shed be provided with an inline milk meter, and this is often the case with new and modern milking sheds.

Thus, the milking performance of an animal can effectively be measured in real time or within a matter of hours. This allows early detection of any changes in milking performance before those changes are manifested in physical cues such as variances in 10 the behaviour of the animal.

However, many older milking sheds, which represent the majority of dairy farms in New Zealand, have limited numbers of milk meters. To retrofit existing milking sheds with the necessary milk meters can incur significant expense. For example, some milk meters, such as those based on technology described in New Zealand Patent Nos. 253311 and 15 255600 can cost in excess of NZ$1,000.

Therefore to retrofit a milking shed with 60 stalls with a meter for each stall will require around NZ$60,000 of milking measurement equipment. Extra costs will be incurred if additional meters measuring somatic cell counts (SCC), the presence of which can be indicative of mastitis in an animal, are integrated into the milking shed. These meters 20 will often include Near Infrared (NIR) based sensors to estimate SCC, protein and fats and can cost in excess of NZ$3,000 (at current market rates).

For dairy farmers, and particularly those with smaller farms, this can represent significant financial investment and cost.

On smaller dairying operations a compromise must be found between the use of milk 25 meters and the economic investment in installing the milk meters. This may result in the 2 PCT/NZ2013/000244 WO 2014/098627 use of a reduced number of milk meters.

However, this may make it difficult for the dairy farmer to make management decisions relating to the dairy herd based on data that is not necessarily reflective of the performance of individual animals at any given time. This can make it difficult to 5 optimise dairy production.

For example, some animals in the herd may produce better quality milk than others. Milking sheds equipped with individual milk meters may detect these animals, allowing the dairy farmer to separate them out into a separate milking cohort or for breeding stock. 10 This may be difficult to achieve if milk meters were limited in number. Depending on the number of meters present, it may take several milking events to determine the performance of individual animals, if at all, of a herd based on the collected milk. The dairy farmer may have to rely on physical cues based on observations of an animal or group of animals for decisions relating to management of those animals. However, the 15 physical cues may be preceded by undetected changes in milking performance of the animal.

In many smaller dairying operations, the farmer or a herd tester may take regular samples of milk from each animal for testing. This process, referred to as herd testing, can provide useful information about specific characteristics of the milk of each animal, 20 such as percentage of fat and so forth. This can be of assistance in dairy herd management without having to resort to identifying any physical manifestations of impaired milking performance.

However, this sampling process can be time consuming and can also be expensive as it involves using external contractors, such as Livestock Improvement Corporation 25 (www.lic.co.nzL to provide the necessary equipment, collect the required milk samples 3 PCT/N Z2013/000244 WO 2014/098627 and conduct the testing.

On average in New Zealand, LIC conducts four tests on each herd per lactation period (eight months on average). Thus, in no way can this testing regime be considered a “real time” process as is the case with dairy sheds having stalls using inline milk meters. 5 Thus, a change in the milking performance of an animal may not be detected for several weeks. Furthermore, the use of herd testing in New Zealand has declined over the past 20 years.

It is an object of the present invention to address the foregoing problems or at least to provide the public with a useful choice. 10 Further aspects and advantages of the present invention will become apparent from the ensuing description which is given by way of example only.

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 15 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. 20 Throughout this specification, the word "comprise", or variations thereof such as "comprises" or "comprising", will be understood to imply the inclusion of a stated element, integer or step, or group of elements integers or steps, but not the exclusion of any other element, integer or step, or group of elements, integers or steps. 4 WO 2014/098627 PCT/NZ2013/000244

DISCLOSURE OF THE INVENTION

According to one aspect of the present invention there is provided a method of determining milk characteristics for individual animals within a cohort of milking animals, the method characterised by the steps of: 5 a) recording individual identification of animals present at a milking shed for a milking event; b) recording the timing of the milking event for individual animals; c) measuring at least one milk characteristic of milk collected from the cohort of milking animals; 10 d) correlating the identification of an individual animal with the timing of the milking event to develop a profile of one or more milk characteristics for the individual animal.

According to another aspect of the present invention, there is provided a method of determining milk characteristics for individual animals within a cohort of milking animals 15 substantially as described above, the method including the additional step of: e) repeating steps a) to d) to develop a profile of milk characteristics for the individual animal over a period of time.

According to another aspect of the present invention there is provided a system of determining milk characteristics for individual animals within a cohort of milking animals, 20 wherein the system is for use in a milking shed, wherein the milking shed includes a milk vat, a plurality of milking stalls, wherein milk from each stall flows into a common milk line to the milk vat, wherein the system includes a processor, and 5 PCT/NZ2013/000244 WO 2014/098627 at least one milk characteristic sensor positioned to receive milk collected from the cohort of milking animals, and configured to output at least one signal indicative of at least one milk characteristic of the milk, the system characterised in that 5 the processor is configured to: a) record individual identification of animals present at the milking shed for a milking event; b) record the timing of the milking event for individual animals; c) receive the signal from the milk characteristic sensor and determine a milk 10 characteristic of milk collected from the cohort of milking animals; d) correlate the identification of an individual animal with the timing of the milking event to develop a profile of one or more milk characteristics for the individual animal.

According to another aspect of the present invention, there is provided a system of 15 determining milk characteristics for individual animals within a cohort of milking animals substantially as described above, wherein the processor is configured to e) repeat steps a) to d) to develop a profile of milk characteristics for the individual animal over a period of time.

According to another aspect of the present invention there is provided an article of 20 manufacture having computer storage medium storing computer readable program code executable by a computer to implement a method for determining milk characteristics for individual animals within a cohort of animals, the code including: 6 PCT/NZ2013/000244 WO 2014/098627 computer readable program code recording individual identification of animals present at a milking shed for a milking event; computer readable program code recording the timing of the milking event for individual animals; 5 computer readable program code receiving an indication of at least one milk characteristic of milk collected from the cohort of milking animals; and computer readable program code correlating the identification of an individual animal with the timing of the milking event to develop a profile of one or more milk characteristics for the individual animal. 10 The inventive method, system and article of manufacture has particular application for determination of characteristics of milk from a selected animal from a group of animals, particularly, but not limited to, milk yield, based on measurement of an ensemble of milk production from a group of animals, such as a dairy herd or cohort of dairy cows.

Estimated characteristics may also include percentage of fat, percentage of casein, 15 somatic cell count and so forth, of individual cows. This may be particularly advantageous for smaller farms, which for economic reasons, may be limited in the number of meters available for measurement of milk characteristics.

The milk characteristics can be an indicator of the phenotype of the animal, and any changes in the phenotype or milk quality of the animal over time. From this information, 20 decisions can be made with respect to management of the animal or the group of animals. This may result in the culling of low producing animals from the group or selecting high producing animals (which is a valued phenotype) for breeding stock. The determined characteristics may also be indicative of the health of the animal, and thus is useful for detecting when the animal requires appropriate veterinary care. 7 PCT/NZ2013/000244 WO 2014/098627

Milk characteristic should be understood to be a property or properties of the milk of the specific animal. The selected characteristic ideally should be one which in a normal and healthy animal is relatively stable but which has some degree of variability to allow detection of the characteristic against the milk of the other animals in the cohort. 5 The characteristic may be indicative of a particular phenotype of the animal, for example the quantity of milk produced. For example, the milk characteristic may be total milk collected from the animal, i.e. milk yield. Animals that have a high milk yield relative to other animals have value as potential breeding stock.

The milk characteristics of a specific animal may also be indicative of its health. For 10 example, an increase in somatic cell count can indicate the presence of mastitis, a bacterial infection of the udder, which may require veterinary intervention. While not intended to be limiting, other characteristics of the milk can include milk metabolites, percentage fat, percentage protein, electrical conductivity and so forth.

The characteristic may be indicative of a particular property of the milk produced by the 15 animal, for example, its quality. Measures of percentage fat and percentage protein of the milk are useful in this regard, and again animals that have high quality milk may have value as potential breeding stock. Milk quality can be dependent on external factors as well, such as the time of season and type of feed given to the animal. This has implications for the management of the cohort or the individual animal. Persons 20 skilled in the art will readily recognise what milk characteristics of an individual animal can be assessed with the present invention for the purpose of appropriate management of that animal and/or the cohort.

The milk characteristic should be one which can be measured in real time. It will be appreciated that separating out collected milk for later analysis for specific parameters 25 will make it difficult to correlate that milk with a specific animal. Some characteristics of milk, for example the presence of ketones, are not able to be measured in real time with 8 PCT/NZ2013/000244 WO 2014/098627 existing technology. However, once this can be achieved, then these characteristics may also be investigated with the present invention.

The system should be understood to include a processor which is configured to implement the inventive method. This may be a central processing station in the milking 5 shed, or a laptop computer which integrates and is communicative with the hardware in the milking.

The animal should be understood to be an animal valued for its milk output. In preferred embodiments of the present invention, the animal is a dairy cow and shall be referred to as such throughout the remainder of this specification. 10 However, this is not meant to be limiting and the present invention may be used to estimate milk characteristics for other animal species valued for its milk output. For example, the present invention may be used for sheep, the milk of which is valued for use in production of dairy products such as cheese.

It will be understood that the present invention is for use on a dairy farm, on which one 15 or more cohorts of cows are milked in a milking shed.

It will be understood that the present method and system is for use in a milking shed. Such sheds have a plurality of milking stalls, with each stall having a milk line draining a milk claw that typically includes four teat cups (one for each teat of the cow). The milk lines of each stall, which may be referred to as a long milk line, lead to a common milk 20 line feeding a milk vat, in which the milk from the herd is collected.

The long milk line, milk claw, and the teat cups of each milking stall shall now be referred to as the milking plant. However, this is not meant to be limiting. The milking plant may also include equipment such as automatic teat cup removers, automated teat sprayers and so forth. 9 PCT/NZ2013/000244 WO 2014/098627

Milking sheds can take a number of configurations, but the most common are rotary sheds and “herringbone" sheds. A rotary shed includes a rotating platform around the circumference of which is a trench (or in many instances, the platform is elevated from the ground). The milking stalls are 5 mounted on the platform, with the rear of the cow facing outwards. The teat cups can then be placed onto and removed from the udder during milking as the platform slowly rotates. Cows enter the platform one at a time as a stall becomes available. Each stall has its own milking plant.

Herringbone sheds are those in which the milking stalls are paired either side of a 10 trench. The dairy farmer moves along the trench to attach and detach the teat cups to the cows as required. Usually, the cows are introduced to the shed in groups corresponding to the number of stalls in the shed. While not all herringbone sheds may have stalls for individual animals, they do have individual milking plant.

The present invention is readily applicable to milking sheds of both configurations. 15 A cohort should be understood to mean a group of cows. This may be a herd of cows, or a subgroup of the herd.

The milking of each cow of the cohort should be understood to be a milking event for that cow. Depending on the management of the dairy farm, the milking of the cohort of cows may occur once or twice a day. Therefore, in some embodiments of the present 20 invention, the milking event may occur once a day and in others, twice a day.

The milking event has a start time and a finish time. In some embodiments of the present invention, the start of the milking event is when the cow enters the milking shed, and the end is when the cow exits the milking shed. However, as persons skilled in the art will appreciate, there can be a time lag between the cow’s entry into the shed and 25 when milk begins to be collected from the udder. There may also be a lag between the 10 PCT/NZ2013/000244 WO 2014/098627 end of the milking event and the subsequent exit of the cow from the shed.

In preferred embodiments of the present invention, the milking event starts upon application of the teat cups to the teat and ends when the teat cups are removed. This is a more accurate determination of the timing of the milking event. 5 In these embodiments, the milking plant may include sensors to determine when they are on and off the teats of the animal. These sensors may be associated with the teat cups and thus send a signal when the cups are applied and removed from the animal. Alternatively, the sensors may be in the claw or long milk line, and send a signal upon detection of the presence and absence of milk. In some milking sheds, such sensors 10 are commonplace.

Alternatively, the sensors may be associated with other equipment within the milking plant. For example, the sensors may be associated with automatic teat cup removers, and the timing of the milking event determined by entry of the animal in the milking shed or milking stall and removal of the teat cups. Persons skilled in the art will readily 15 appreciate other ways in which the duration of the milking event may be determined.

It will be understood that the processor is in communication with the sensors, either directly or wirelessly.

The present invention requires each cow have an individual identifier. This will be a number or similar identifying means that is specifically associated with that animal. In its 20 simplest form, the individual identifier may be a marked tag attached to the ear of the cow.

For example, the individual identifier may be an ear tag on which is notated a number unique to the animal provided with the ear tag.

Preferably, the individual identifier is an electronic identifier (EID) which is unique to the 11 PCT/NZ2013/000244 WO 2014/098627 animal. The EID may take a variety of forms depending of the requirements of the user. For example, battery powered transponders may be used as an EID. These may be integrated into an ear tag, bolus or subcutaneous implant.

In particularly preferred embodiments of the present invention, the individual identifier is 5 a Radio Frequency Identifier (RFID) ear tag. These are transponders that do not require batteries. RFID ear tags are mandatory in the dairying industry of New Zealand, which has to comply with the National Animal Identification and Tracing (NAIT) scheme.

Suitable RFID ear tags are readily available and include those manufactured by Allflex® New Zealand Limited (http://www.allflex.co.nzL 10 Reference shall now be made throughout the remainder of the specification to the individual identifier being a RFID ear tag. However, as should be appreciated this is not meant to be limiting and other forms of the individual identifier are envisaged.

The present invention requires the system to have a recording device at the milking shed. The recording device is configured to record the individual identifier, i.e. the 15 individual animals when they are in the milking shed, or using a milking plant in the milking shed or within a stall of the milking shed.

In the particularly preferred embodiment of the invention, the RFID of the ear tag is configured to communicate wirelessly or otherwise with a recording device in the form of a RFID reader. This simplifies the data collection process. 20 However, the use of an RFID reader as the recording device is not meant to be limiting as it should be appreciated that the nature of the recording device may depend on the type of EID that is to be used.

The recording device may also time the start and cessation of the milking event.

In some embodiments of the present invention, the recording device is associated with 12 PCT/N Z2013/000244 WO 2014/098627 the entry and exit points of the milking shed. This allows determination of when individual cows enter and leave the milking shed. For example, the recording device may be an Allflex® panel reader integrated into the races leading into and out of the dairy shed. Therefore it will be appreciated in this embodiment of the invention, the 5 RFID reader not only records the RFID ear tag, but also the presence of the animal, to mark the start of the milking event.

The end of the milking event may be determined by the RFID reader reading the ear tag as the animal leaves the milking shed.

This may be particularly suitable for herring-bone milking sheds, which may not be 10 provided with individual milking stalls.

In some embodiments in which the invention is used in a rotary milking shed, the RFID reader may only be associated with the entry race. If no RFID reader is associated with the exit race, the exit of the animal may be determined by the position of the rotary platform at any given moment. The position of the platform is known and therefore it is 15 relatively easy to determine when the animal reaches the exit point and thus determine the length of the milking event if the entry time is known.

However, as discussed above, there may still be a lag between the cows entering (or exiting) the dairy shed and the milking event actually commencing (or finishing, as the case may be). For example, in a rotary platform, a cow may be fully milked well before 20 its stall has reached the exit point of the rotary platform. This may affect how quickly the invention can arrive at an accurate profile for an individual animal.

In preferred embodiments, the recording device is a RFID reader linked to a data logger. The presence or absence of the RFID ear tag of the animal is noted by the RFID reader and the data provided to the data logger. The data logger may also be configured to 25 receive signals indicative of the start or end of the milking event. 13 PCT/N Z2013/000244 WO 2014/098627

It will be understood that the recording device and/or data logger is linked or otherwise communicative with the processor of the present system.

In preferred embodiments of the present invention, the recording device or data logger is also associated or otherwise communicative with the sensors of the milking plant of 5 each stall. This provides a much more accurate determination of the onset and cessation of the milking event.

The method requires the recordal of the start and end of the milking event for each specific cow, along with determination of total milk flow for the cohort. Therefore, a sensor to measure the rate of milk flow into the vat of the milking shed is needed. 10 The sensor that measures the rate of milk flow is communicative or otherwise linked with the processor of the present system. It will be appreciated that the processor is configured to correlate milk data, collected by the sensor, together with the timing of the milking event (from the recording device) and the identity of the animal being milked (from the RFID ear tag). 15 In preferred embodiments of the present invention, the sensor is a milk flow meter and shall be referred to as such throughout the remainder of this specification.

However, this is not meant to be limiting and the method may be used with other or additional sensors of interest which can measure milk or its characteristics in real time. For example, the sensor may measure the somatic cell count (SCC), or indicators 20 thereof such as conductivity.

As persons skilled in the art will appreciate, SCC may be used as a measure of mastitis. Incorporating a SCC sensor and a milk flow sensor in the common milk line or at or proximate the milk vat may allow for more accurate detection of mastitis in individual animals without the need for SCC sensors in the milking plant of each stall. This can 25 represent significant cost savings for the dairy farmer. 14 PCT/NZ2013/000244 WO 2014/098627

The example described above is not meant to be limiting. Persons skilled in the art will appreciate other characteristics of the milk aside from SCC may be estimated with the present invention, depending on the types of sensors employed in the milking shed.

Essentially, the invention can be any characteristic of milk able to be measured in real 5 time by an appropriate sensor and which may be indicative of a particular phenotype of the specific animal (for example, its milk yield although this is not meant to be limiting). In preferred embodiments of the present invention, the milk flow sensor is positioned at the milk vat, or in the common milk line preceding the milk vat.

The sensor measures the milk flow and/or other properties of the milk that is collected 10 from the cohort. From this, a profile for the total milk can be established.

Following the completion of the milking event, the data relating to the identification of individual animals and the timing and duration of the milking event is transmitted, or otherwise collected from, the recorder and/or data logger along with the total milk information from the sensor for processing. Processing may be conducted using any 15 suitable processing device such as the central processing station of the milking shed or on a laptop or desktop computer. The data is then run through an algorithm.

The algorithm separates data associated with the individual animal from the total milk information to determine a milking profile for each animal. This profile may include milk yield, SCC, percentage of fat, percentage of proteins and so forth. 20 The algorithm may include additional data derived from alternative sources of information relating to the animal or the cohort. For example, the algorithm may take into account the age of the animal or its weight.

It is envisaged that by collecting data from successive milking events, a more accurate profile of a specific animal can be built in comparison with a single event in isolation. 15 PCT/N Z2013/000244 WO 2014/098627

In use, the cow (cow ‘A’) enters the milking shed and is identified, using the RFID ear tag reader, by its unique identification number and has teat cups attached to its teats.

Upon the start of the milking event, the recording device notes that cow ‘A’ has commenced milking. At the same time, the sensor associated with the common milk 5 line or milk vat measures the milk from the cohort.

The data collected from the recording device or data logger may be sent in real time to a computer processor or the like. Alternatively, the data is stored and retrieved following the end of the milking.

Each cow produces milk having a particular milk profile. The recording of the onset and 10 cessation of the milking event for specific animals means that the characteristics of the total milk flowing through the common milk line and/or entering the vat at any one time can be associated with the specific animals being milked at that time.

Each milking event is different; a specific cow may not be milked at the same time as the previous milking event. However, as the milk flowing through the common milk line 15 and/or entering the vat at any one time can still be associated with the specific animals being milked at that time, over time, the milk profile of a particular animal can be identified as being distinct from other animals in the cohort.

Thus, by application of an appropriate algorithm to the collected data, the milk characteristics of a specific animal, such as milk yield, can be determined with a 20 reasonable degree of accuracy. A profile of the animal can be established, and over time, any variations to that profile can be detected.

This provides a useful management tool for the dairy farmer. Informed decisions relating to the culling, breeding or veterinary care of specific animals in the herd can be made. Additionally, decisions relating to improving milk quality can also be made. 16 PCT/NZ2013/000244 WO 2014/098627

Certain characteristics of milk are useful for determining quality of the milk derived from the animal. Any changes in these characteristics may be related to external factors rather than as a direct result of the animal’s physiology. For example, a change in feed may affect the quality of milk. 5 The present invention allows a dairy farmer to correlate a decrease or increase in a milk characteristic that is reflective of the milk quality (or a change in the phenotype of the animal) with a change in a particular external factor (such as a change in feed).

Although the present invention may utilise only one milk flow meter, it is not beyond the scope of the present invention that additional milk flow meters be employed along the 10 common milk line.

Doing so may improve the accuracy of the estimate and potentially within a shorter time period as there would be a stronger correlation between the onset of a milking event and a specific animal within the cohort.

For example, a herringbone milking shed may be configured with a milk line for each 15 side of the shed. Such an arrangement may allow the use of two appropriately positioned milk meters, one for each line.

Having more than one milk meter along the common milk line may be useful for meter calibration purposes but is not essential to the present invention.

Another way of improving the accuracy of the estimate is by milking twice a day, which 20 provides effectively two opportunities to gather data for implementation of the method. In some countries, two milking events per day is common practice in management of the dairy farm.

The time between successive milking events, which is on average 24 hours for animals milked once a day or between 8 to 14 hours for animals milked twice daily, can be used 17 PCT/N Z2013/000244 WO 2014/098627 as a covariate to improve the way in which the profile of a specific animal is developed. Other useful covariates may be the live weight and/or age of the specific animal or information associated with previous seasons.

However, it should be appreciated that, depending on the milk characteristic being 5 determined, more (or less) time may be required to establish a reasonably accurate profile for the animal.

Embodiments of the present invention may be implemented as instructions (for example, procedures, functions, and so on) that perform the functions described. It should be appreciated that the present invention is not described with reference to any particular 10 programming languages, and that a variety of programming languages could be used to implement the present invention. Firmware and/or software codes may be stored in a memory, or embodied in any other processor readable medium, and executed by a processor or processors. The memory may be implemented within the processor or external to the processor. 15 A general purpose processor may be a microprocessor, but in the alternative, the processor may be any processor, controller, microcontroller, or state machine. A processor may also be implemented as a combination of computing devices, for example, a combination of a digital signal processor (DSP) and a microprocessor, a plurality of microprocessors, one or more microprocessors in conjunction with a DSP 20 core, or any other such configuration. The processors may function in conjunction with servers and network connections as known in the art.

The steps of a method, process, or algorithm described in connection with the present invention may be embodied directly in hardware, in a software module executed by a processor, or in a combination of the two. The various steps or acts in a method or 25 process may be performed in the order shown, or may be performed in another order. Additionally, one or more process or method steps may be omitted or one or more 18 PCT/NZ2013/000244 WO 2014/098627 process or method steps may be added to the methods and processes. An additional step, block, or action may be added in the beginning, end, or intervening existing elements of the methods and processes.

Embodiments of the present invention offer one or more of the following advantages: 5 · a cost effective method of determining milk characteristics, such as milk yield, of a specific dairy cow by measuring the total milk of the milking cohort and correlating this with timing of the milking event of the specific animal; • provides measurement of milk characteristics using relatively few milk meters and achieving an accuracy comparable to dairy shed systems having inline milk 10 meters; • is readily adapted into existing dairy sheds in a cost effective manner due to the potential for a reduced number of expensive milk sensing equipment; • a tool to optimise dairy herd performance on farms lacking inline milk meters; • at the very least, offers the public a useful choice.

15 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 is a schematic of the system of the present invention; 20 Figure 2 is a graph of the measured total milk flow into a milk vat representing a milking event for a cohort of 923 dairy cows being milked in a rotary shed;

Figure 3 is a graph of estimated and measured morning milk yield for a specific 19 PCT/N Z2013/000244 WO 2014/098627 cow of the cohort of Figure 2 over 100 days;

Figures 4 and 5 are graphs of predicted milk yields calculated using the present invention and conventional methods compared against actual measured milk yields; 5 Figure 6 is a flow chart of the step-by-step process for estimating milk yield of an individual animal from a bulk milk flow;

Figures 7 and 8 are a graph of predicted milk electrical conductivity calculated using the present invention and conventional methods compared against actual measured milk electrical conductivity; 10 Figure 9 is a graph of estimated and measured milk electrical conductivity for a specific cow of the cohort of Figure 2 over 100 days;

Figure 10 is a flow chart of the step-by-step process for estimating a milk composition of interest for the milk from an individual animal from a bulk milk flow; 15 Figure 11 is a graph of the percentage fat in the milk flow into a milk vat, representing a milking event for a cohort of 347 dairy cows being milked in a rotary shed;

Figure 12 is a graph of estimated and measured percentage fat in the milk of a specific cow over 100 days; and 20 Figures 13 and 14 are a graphs of predicted percentage milk fat calculated using the present invention and conventional methods compared against actual measured percentage milk fat. 20 PCT/NZ2013/000244 WO 2014/098627

BEST MODES FOR CARRYING OUT THE INVENTION

The system of the present invention is illustrated as a schematic in Figure 1, and depicts a herring bone milking shed (1).

The milking shed includes a plurality of milking stalls (2), each stall including a RFID 5 reader (3) to wirelessly (4) read the RFID (not shown) integrated into the ear tag (not shown), which identifies each animal (not shown).

The RFID reader (3) is linked (3a) to a data logger (5) to record the identification of each animal (not shown).

The milking stalls (2) are linked (2a) to a common milk line (6). The stalls have a milking 10 plant that includes cup sensors (7) which are linked (7a) to the data logger (5) to note the commencement and cessation of the milking event.

The milking shed also includes a milk vat (8) in which the milk from each stall (2) collects. A milk meter or sensor(s) (9) is integrated into the common milk line (6) preceding the milk vat to measure the quantity of milk or a characteristic of interest of 15 the milk flowing through the common milk line (i.e. from the cohort of animals being milked). A person skilled in the art will readily identify an appropriate meter or sensor for the task, depending on the characteristic of interest.

The data logger (5) and meter or sensor(s) (9) are in communication (5a, 9a) with a processor (10), which may be a computer used to monitor and control elements of the 20 milking system.

The system may be readily modified by a skilled addressee for implementation in a rotary shed.

Aspects of the method shall now be described with reference to the following graphs (Figures 2 to 5, 7, and 8), which is based upon data collected from a herd of 923 dairy 21 PCT/NZ2013/000244 WO 2014/098627 cows being milked in a rotary shed of 60 stalls at Tokanui, New Zealand over a four month period, and simulations performed on same to test the method.

Figure 2 is a graph showing a simulation of the measured total milk flow (kilograms per minute) in the common milk line leading to the milk vat, representing a milking event of 5 180 minutes for a cohort of dairy cows being milked in a rotary shed.

It will be observed that there is considerable variation in the milk flow, which is indicative of the start and cessation of the milking event for specific cows. By noting the time at which specific cows start and stop milking over a series of milking events it can be possible to develop a profile of each animal in the cohort. 10 From the collected data, application of an algorithm then allows the dairy farmer to estimate the milk yield for that animal. The algorithm shall now be discussed in detail.

The rate of milk production (y,·) (litres/minute) of the individual animal (the ith cow) as a function of time (t) (minutes) can be described by Equation 1: 0, t <0 a„ 0<t<tu a. lt-tu) (^2/ _ ^1/) 0, t > t2 15 where a, is the rate of milk production, t2i is the milking duration and tu is time for maximum milk production. The milk yield is Yf =\ai{t2j + /„) litres. The expected value of /„ is known to be £[/„] = 0.38/2,.

If there are N cows milked daily with each cow beginning milking at time Tt then the total milk flow (litres/minute), as a function of the milk flow for individual animals, the common 20 milk line is determined by Equation 2: 22 PCT/NZ2013/000244 WO 2014/098627

Kt) = 1=1 A necessary input into the algorithm is the total milk flow. This may be calculated using Equation 3. Assuming the total milk flow is recorded periodically (for example, At=5 seconds; this is not meant to be limiting and other time intervals may be used) at times 5 rj = jAt,)=1 ..M, then Equation 3 is:

N bj = b(rj) = Σ^/Ο} -Τ,Μ) /=1

This can be expressed as a matrix system, in which the rows of the matrix A are the times the bulk flow is measured while the columns of matrix A are the individual animals. The matrix system is defined by Equation 4: 10

Aa = b whereb = [6, b2 ... bMJ are the total flow measurements, a = [a, ... aMJ are the unknown individual yields per cow, and the (j. ) element of the matrix A is y,(rj , which is determined from the recording start and finish time of milking for each cow (and it is assumed that /,, = 0.38/2i ). 15 It should be noted that 7) (the time of commencement of milking for an individual animal) and 12i (the length of the milking event for the individual animal) are two further inputs to the algorithm.

Therefore, a matrix system can be constructed for any one day (e.g. the mth day) of milking using Equation 5, which is as follows: 20 A,a = b,„ m m 23 PCT/NZ2013/000244 WO 2014/098627

This assumes that the order that the cows are milked is different each day, which has been verified. A block matrix system over multiple days can then be constructed using Equation 6, as follows: 5 A'”+”a = b m m+n m A„, h , m lit where A“+n = m A ^m+l and b”+" = ^m+l A _ m+n _ -bm+n.

Equation 6 can be readily solved by weighted least squares, where the weights relate to the accuracy of flow measurement (simulations were conducted with n=3). The solution to this is Equation 7, which is: 10 i=|c*f ν-ΆΓ )·'((α;*· 7 v-'br) where V is the measurement variance-covariance matrix for the total flow. This can then be used to calculate an estimate of the milk yield, in litres, using Equation 8: — 2 ai (?2i tU)

The outputs of the algorithm in this instance are values a„ and for milk yield for 15 individual cows, 7,.

The variance-covariance matrix for the estimated parameter a may be calculated using Equation 9: Μ1 =((αΓ"Γ where the variance in a, is M“. 24 PCT/N Z2013/000244 WO 2014/098627

Changes in the milk yield of individual cows over a milking season can then be tracked using a Bayesian filter. In the simplest case, a Kalman filter can be employed using a, M', an initial prior estimate of milk yield (assumed to 10 ± 3 litres in simulations) and an estimate of the between day variance in milk yield (known to have a coefficient of 5 variation of around 15%). A state transition model can also be incorporated in the Kalman filter to describe the milk lactation curve (expected change in milk yield between days over a year).

Simulated total flow is calculated by Equation 10 as follows: Κ0+οξ(φ(0 10 where σ represents the amplitude of the perturbations; and ξ(ί) denotes normal white noise, which is a Gaussian distributed process with zero mean and independent increments. For simulations σ=0.07, i.e. the assumption for errors in the milk flow meter measurement is 7%.

Referring back to Figure 2, it should be noted that there is a drop in milk flow from 60 15 minutes to 80 minutes and 140 minutes to 150 minutes. This is indicative of the change over in the milking shed as one group of cows exit the shed and another group enters. Although not essential, by careful management of each group, i.e. keeping each group separated for a period of days, the accuracy of the method can be improved over a shorter period of time. It should be noted that an error of 7% is assumed for the milk 20 meter.

Figure 3 is a graph comparing the estimate and measured morning milk yield (in kilograms) of a particular animal (cow 1) of the Tokanui herd over a 100 day period beginning 1 January 2011.

It will be seen that over time the estimated yield (indicated by line) tracks the measured 25 PCT/NZ2013/000244 WO 2014/098627 yield (indicated by x). This assumes an initial morning milk yield of 10±3kg on day 0 of the trial. Error bars denote standard errors for the estimate.

Figure 4 shows the relationship between total milk yield (measured in kilograms per cow) over the four month period predicted using the present method (y axis) against the 5 measured milk yield (x axis, also in kilograms per cow) from the Tokanui herd. The correlation between the predicted milk yield and the measured yield data was r=0.91.

Figure 5 shows the relationship between milk yield estimated using conventional methods (in kilograms, horizontal axis) against the measured milk yield (in kilograms, vertical axis) from the Tokanui herd. 10 The correlation between the milk yield predicted using standard Livestock Improvement Corporation (LIC) herd testing (www.lic.co.nz/lic Herd Testinq.cfm). based on an average of four tests for each animal in the Tokanui herd during a lactation period of 120 days, and the measured yield data across the same lactation period is r=0.88. This assumes that the Tokanui milk meters and herd test milk meters are of comparable 15 accuracy.

The method (600) that tracks the change in milk yield of individual animals over time may also be exemplified in a basic form by the flow chart illustrated in Figure 6.

Prior to starting the process, in a preliminary step (601) m is set with the value 1 (day 1) and n, the number multiple days, is selected (in the described simulations, n=3). The 20 first step (602) is to obtain an initial prior estimate of milk yield for each cow based on weight, age and previous performance.

The second step (603) is to retrieve the individual identifier associated with the animal, the time of start of milking for the animal (Tj), the duration of the milking event fe), and the bulk milk flow (bj(t)) for each cow milked on days m, m+1,..., m+n after the start of 26 PCT/NZ2013/000244 WO 2014/098627 milking.

In the third step (604), the milk yield for each animal, and the variance thereof, may then be estimated using Equations 1 to 9 as discussed above. The measurement variance-covariance matrix V is specified by the accuracy of the bulk milk flow meter. 5 To track the change in the milk yield of individual animals over these m days (based on an estimate of the between day variance in milk yield and the expected milk lactation curve), a Kalman/Bayesian filter may be used in the next step (605). The process is then repeated by setting m as m+n+1 in step 606 and performing steps 603 through to 605. 10 Thus, the present invention offers more accurate estimates for milk yield relative to conventional methods for estimating milk yields. However, the use of the invention is not limited to estimation of milk yields.

Figure 7 is a graph showing the relationship between milk electrical conductivity (EC) predicted using the method of the present invention (horizontal axis) against the 15 measured average milk electrical conductivity (vertical axis), again demonstrated using a herd from Tokanui of 923 cows. This is based on EC measured in the common milk line. In this instance, the correlation between measured electrical conductivity and predicted milk conductivity was r=0.77.

This is comparable to the relationship (r=0.83) between electrical conductivity per cow 20 estimated using herd testing (horizontal axis) against measured electrical conductivity (vertical axis), as depicted in Figure 8, over the same four month period for the same Tokanui herd.

This is based on morning herd tests carried out on day 0, 60 and 120 and also assumes that the meters used on the dairy farm and by the herd testers are of comparable 25 accuracy. 27 PCT/NZ2013/000244 WO 2014/098627

Figure 9 shows estimated milk electrical conductivity (indicated by line) versus measured milk electrical conductivity (indicated by x) for one cow of the Tokanui herd over a 120 day period. It will be seen that the standard error bars are at their greatest for the early milking events but gradually decrease as the estimates of electrical 5 conductivity becomes more accurate.

Simulations have also been performed to test the method for assessing other properties of milk, on data collected from a herd of 347 dairy cows being milked in a rotary shed of 34 stalls at LIC’s Innovation Farm in New Zealand over a four month period beginning 4 July 2013. 10 Milk composition can be treated similarly where it is assumed that the individual milk yields are measured (Τηαηίιηί2ί) for each cow but the milk composition (f) of individual cows is not measured. In this case the measured composition of the total flow by the bulk composition sensor is expressed as the linear system (Equation 11): (t) = y tf b(t) 15 In this instance, the inputs are 7), tu, a,·, fe/ and c(t). These can be measured by milk meters or estimated using one of Equations 1 to 9 as previously discussed.

The procedure outlined with respect to Equations 4 to 9, as discussed above with respect to milk yield, can then be used to solve for the milk composition (/) of individual animals in Equation 11. 20 The simulated milk composition of the total flow is calculated using Equation 12 as: c(t) + a£(t)c(t) 28 PCT/NZ2013/000244 WO 2014/098627 where: σ represents the amplitude of the perturbations; and ξ(ί) denotes normal white noise, which is a Gaussian distributed process with zero mean and independent increments. For simulations, σ =0.025 (i.e. the milk composition measurement error is 2.5%). 5 The step-by-step process (1000) for estimation of milk composition, as exemplified by the basic flow chart of Figure 10, is similar to that used for estimation of milk yield (as described with respect to Figure 6).

Prior to starting the process in preliminary step 1001, m is set with the value 1 (day 1) and n, the number of multiple days is selected (as previously, n=3). The first step 10 (1002) is to obtain an initial prior estimate of a particular characteristic of interest for the milk of each cow based on weight, age and previous performance.

The second step (1003) is to retrieve the individual identifier associated with the animal, the time of start of milking for the animal (T,), the duration of the milking event (t2i), and the rate of milk production (¾) for each cow milked on days n, n+1,..., n+m after the 15 start of milking. This may be measured directly, using the appropriate individual milk metering equipment or estimated using the bulk milk algorithm (Equations 2 to 9).

In the next step 1004, the milk composition for each animal, and the variance thereof, may then be estimated using Equation 11 as defined above. The accuracy of the metering equipment measuring the composition of interest will determine the 20 measurement variance-covariance matrix V. A Kalman/Bayesian filter is then used in the fourth step 1005 to monitor any changes in the milk composition of individual cows over the number of days. This is based on an estimate of the between day variance in milk composition and the expected changes in milk composition during lactation. The process is then repeated by setting m as m+n+1 25 in step 1006 and performing steps 1003 through to 1005. 29 PCT/NZ2013/000244 WO 2014/098627

It will therefore be appreciated that the present invention offers more accurate estimates for milk composition relative to conventional methods for estimating milk composition.

Figure 11 is a graph showing the simulated measured percentage of fat in the milk flow into the milk vat for a specific milking event for the LIC herd (based on a 2.5% sensor 5 and recorded milking times, % fat and yield per cow). As with Figure 2, there is considerable variation in the milk flow, including a disruption about 118 minutes into the milking event. These variations are indicative of the start and cessation of the milking event for specific cows. Noting the time at which specific cows start and stop milking over a series of milking events and measuring of percentage fat at any given time can 10 lead to the development of a profile of the milk of each animal in the cohort.

Turning to Figure 12, this is a graph for percentage fat in the milk of one animal (cow 1) of the LIC herd (which numbered 347 animals), measured (indicated by x) and estimated (indicated by line) with the present invention, over a milking period of 120 days beginning 4 July 2013. For the first 20 days, cow 1 was not milked at all. Upon the 15 first milking event the error bars, denoting standard errors, increase but then become reduced over time. This is based on an initial guess of 4.8±1.1 % fat on day 0.

There Is a greater correlation (r=0.70) between the percentage milk fat per cow predicted (horizontal axis) with the present invention (based on % fat in the common milk line) and measured percentage milk fat per cow (vertical axis) for the 347 animal 20 herd over the 120 day period (Figure 13) than the percentage fat per cow estimated (horizontal axis) using the two-monthly LIC herd testing for the same herd, which is r=0,58 (Figure 14). This assumes that the metering equipment used on-site and herd test meters are of comparable accuracy.

Aspects of the present invention have been described by way of example only and it 25 should be appreciated that modifications and additions may be made thereto without departing from the scope thereof as defined in the appended claims. 30

Claims

WHAT WE CLAIM IS:

1. A method of using a system of determining milk characteristics for individual animals within a cohort of milking animals, wherein the system is for use in a milking shed, wherein the milking shed includes a milk vat, a plurality of milking stalls, wherein milk from each stall flows into a common milk line to the milk vat, the system including: a processor, and at least one milk characteristic sensor positioned in the common milk line and configured to output a signal indicative of at least one milk characteristic of the milk, the method characterised by the steps of: a) recording individual identification of animals present at a milking shed for a milking event; b) recording the timing of the milking event of individual animals; c) measuring at least one milk characteristic of milk collected from the cohort of milking animals; d) correlating the identification of an individual cow with the timing of the milking event to develop a profile of one or more milk characteristics for the individual animal.
2. The method of claim 1, wherein the method includes the additional step of: e) repeating steps a) to d) to develop a profile of milk characteristics for the individual animal over a period of time.
3. The method of either claims 1 or 2 wherein the animals are assigned an individual identifier.
4. The method of claim 3 wherein the individual identifier is a RFID ear tag.
5. The method of any one of claims 1 to 4 wherein the milking event is determined as being the length of time between entry and exit of the individual animal from the milking shed.
6. The method of any one of claims 1 to 4 wherein the milking event is determined as being the length of time between entry and exit of the individual animal from a milking stall of the milking shed.
7. The method of any one of claims 1 to 4 wherein the milking event is determined as being the length of time between application and removal of teat cups from the individual animal.
8. The method of any one of claims 1 to 7 wherein the milk characteristic is milk yield.
9. The method of any one of claims 1 to 7 wherein the milk characteristic is milk electrical conductivity.
10. The method of any one of claims 1 to 7 wherein the milk characteristic is indicative of the composition or quality of the milk of the individual animal.
11. The method of claim 10, wherein the milk characteristic is selected from one or more of the following: percentage of fat, percentage of lactose, percentage of casein, or percentage of protein.
12. The method of any one of claims 1 to 7 wherein the milk characteristic is indicative of the health of the individual animal.
13. The method of claim 12 wherein the milk characteristic is selected from one or more of the following: metabolites or somatic cell count.
14. The method of any one of claims 1 to 13 wherein the animals are dairy cows.
15. The method of any one of claims 1 to 14 wherein the cohort of milking animals is a dairy herd.
16. A system of determining milk characteristics for individual animals within a cohort of milking animals, wherein the system is for use in a milking shed, wherein the milking shed includes a milk vat, a plurality of milking stalls, wherein milk from each stall flows into a common milk line to the milk vat, the system including: a processor, and at least one milk characteristic sensor positioned in the common milk line to receive milk collected from the cohort of milking animals, and configured to output at least one signal indicative of at least one milk characteristic of the milk, the system characterised in that the processor is configured to: a) record individual identification of individual animals present at the milking shed for a milking event; b) record the timing of the milking events of individual animals; c) receive the signal from the milk characteristic sensor and determine a milk characteristic of milk collected from the cohort of milking animals; d) correlate the identification of an individual cow with the timing of the milking event to develop a profile of one or more milk characteristics for the individual animal.
17. The system as claimed in claim 16, wherein the processor is configured to: e) repeat steps a) to d) to develop a profile of milk characteristics for the individual animal over a period of time.
18. The system as claimed in either claim 16 or claim 17 wherein the processor includes a recording device.
19. The system as claimed in claim 18 wherein the recording device is a RFID ear tag reader linked to a data logger.
20. The system as claimed in either claim 18 or claim 19 wherein the recording device is associated with entry and exit points of the milking shed.
21. The system as claimed in either claim 18 or claim 19 wherein the recording device is associated with the milking plant.
22. The system as claimed in any one of claims 16 to 21 wherein the system includes an individual identifier for the animals.
23. The system as claimed in claim 22 wherein the individual identifier is a RFID ear tag.
24. The system as claimed in any one of claims 16 to 24 wherein the milk characteristic sensor is a milk flow meter.
25. The system as claimed in any one of claims 16 to 24 wherein the milk characteristic sensor is configured to sense a property of the milk that is indicative of composition or quality of the milk of the individual animal.
26. The system as claimed in any one of claims 16 to 24 wherein the milk characteristic sensor is configured to sense a property of the milk that is indicative of the animal’s health.
27. The system as claimed in claim 26 wherein the milk characteristic sensor is a somatic cell count sensor.
28. An article of manufacture having computer storage medium storing computer readable program code executable by a computer of a system for determining milk characteristics for individual animals within a cohort of milking animals to implement a method for determining milk characteristics for individual animals within the cohort of milking animals, wherein the system is for use in a milking shed, wherein the milking shed includes a milk vat, a plurality of milking stalls, wherein milk from each stall flows into a common milk line to the milk vat, the system including: a processor, and at least one milk characteristic sensor positioned in the common milk line to receive milk collected from the cohort of milking animals, and configured to output at least one signal indicative of at least one milk characteristic of the milk, and wherein the code includes: computer readable program code recording individual identification of animals present at a milking shed for a milking event; computer readable program code recording the timing of the milking event for individual animals; computer readable program code receiving an indication of at least one milk characteristic of milk collected from the cohort of milking animals; and computer readable program code correlating the identification of an individual animal with the timing of the milking event to develop a profile of one or more milk characteristics for the individual animal.