System and method for analyzing a fluid

Abstract

Systems and methods for analyzing milk are described. A plurality of first type sensors are provided, each first type sensor being associated with a respective one of a plurality of teat cup sets of a milking system and configured to analyze milk extracted by the teat cup set from an individual animal to determine at least one first type sensor value of a parameter of the milk across an event period. At least one of a second type sensor associated with at least one of the plurality of teat cup sets is configured to analyze milk extracted by the teat cup set from the individual animal to determine at least one second type sensor value of the parameter of the milk within the event period, wherein the second type sensor is less susceptible to animal-specific bias than the first type sensor. The number of second type sensors in the system is less than the number of first type sensors. Based on the at least one first type sensor value and the at least one second type sensor value of the parameter determined for the individual animal, an animal-specific bias correction is determined for the individual animal. The animal-specific bias correction is applied to the first type sensor values of the parameter obtained from the first type sensors for milk extracted from the individual animal.

Description

[0001] Cross Reference to Related Applications

[0002] This application is based on the specification filed in respect of New Zealand patent application number 762276, the entire contents of which are incorporated herein by reference. TECHNICAL FIELD

[0003] The present disclosure relates to systems and methods for analysing a fluid, and more particularly to systems and methods for analysing milk using sensing devices within a milking environment. BACKGROUND

[0004] It is known to use sensors to obtain information relating to milk collected from a milk producing animal. Such information is used in relation to decisions about matters such as processing of milk, selection, breeding, medical treatment, animal specific feed rationing and measurement of milk production efficiency.

[0005] In some milking systems, in-line sensors are provided which are able to collect data throughout a milking, sensing characteristics of the milk flowing past them. Ideally, such sensors are installed for each milking stall of a milking system to allow data to be collected from individual animals at a high frequency (in terms of the data collected each time an animal is milked).

[0006] However, in order to achieve an acceptable price-point to allow for mass installation to provide high milking stall coverage and to meet the constraints imposed by flowing milk, such in-line sensors typically have a lower accuracy than other known sensor types. This impacts on the quality or certainty of the data collected and therefore the effectiveness of decisions made based on that data.

[0007] It is an object of the present invention to address the foregoing problems or at least to provide the public with a useful choice.

[0008] All references, including any patents or patent applications cited in this specification, are hereby incorporated by reference in their entirety. The references disclosed herein are not

[0009] Unless the context clearly requires otherwise, throughout the description and the claims, the words 'comprise', 'comprising', and the like are to be construed in an inclusive sense as

[0010] Other aspects and advantages of the present invention will become apparent from the following description, given by way of example only, and with reference to the accompanying drawings. SUMMARY

[0011] According to one aspect of the present technology, there is provided a system for analysing milk, comprising:

[0012] a plurality of first type sensors, each first type sensor being associated with a respective one of a plurality of milking clusters of a milking system, and configured to analyse milk extracted by the milking cluster from an individual animal to determine at least one first type sensor value of a parameter of the milk across an event period;

[0013] at least one second type sensor of a second type of sensor, associated with at least one of the plurality of milking clusters, and configured to analyse milk extracted by the milking cluster from an individual animal to determine at least one second type sensor value of the parameter of the milk within the event period, wherein the second type sensor is less susceptible to animal-specific bias than the first type sensor, and wherein the number of second type sensors in the system is less than the number of first type sensors; and

[0014] at least one processor configured to:

[0015] determine an animal-specific bias correction for the individual animal based on the at least one first type sensor value and the at least one second type sensor value of the parameter determined for the individual animal; and

[0016] apply the animal-specific bias correction to first type sensor values of the parameter obtained from the first type sensors for milk extracted from the individual animal.

[0017] According to one aspect of the present disclosure, there is provided a method for analysing milk in a system having a plurality of milking clusters, each milking cluster being configured to extract milk from an individual animal, the method comprising:

[0018] analysing milk extracted by the milking cluster from an individual animal using a first type sensor associated with the milking cluster to determine at least one first type sensor value of a parameter of the milk across an event period;

[0019] analysing milk extracted by the milking cluster from an individual animal using a second type sensor associated with the milking cluster to determine at least one second type sensor value of the parameter of the milk within the event period, wherein the second type sensor is less susceptible to animal-specific bias than the first type sensor, and wherein the number of second type sensors in the system is less than the number of first type sensors; and

[0020] determining an animal-specific bias correction for the individual animal based on at least one first type sensor value of a parameter determined for the individual animal and at least one second type sensor value of the parameter; and

[0021] applying the animal-specific bias correction to first type sensor values of the parameter obtained from the first type sensor for milk extracted from the individual animal.

[0022] There are many sensors for automatically sensing various parameters of milk, including various components of milk (such as, but not limited to, one or more of fat, protein and lactose) and attributes such as volume (i.e. yield) of milk extracted. Animal-specific bias in sensor measurements occurs when one or more attributes of the milk or the animal itself that affect the measurement of the target parameter vary between animals but remain relatively consistent for an individual animal. For example, if the measurement of milk fat is affected by the colour of the milk, then the concentration of the molecular compounds that affect the colour of the milk will likely affect the measurement. If the animals in a group have different but persistent concentrations of these compounds, then they will each bias the measurement according to their respective concentrations, resulting in animal-specific bias. It is believed that animal-specific bias can be more pronounced in groups of animals with more genetic diversity, for example multi-breed groups, where genetic diversity contributes to greater variation in the aforementioned milk attributes. Sensors that exhibit animal-specific bias produce measurements for individual animals that tend to be biased relative to other animals in a group of animals, thus affecting comparisons between animals and decisions based on those comparisons.

[0023] Likewise, reference to a correction for animal-specific bias should be understood to mean a correction (e.g. value or function) associated with an individual animal to be applied to the parameter(s) measurement of milk extracted from that animal for which animal-specific bias is determined for a particular type of sensor. It should be noted that devices for identifying individual animals and recording data collected for milk extracted from that animal are well known in the art.

[0024] In an example, the first type sensor can be an optical sensor. Many such sensors are known for milking animals. For example, the first type sensor can be a Protrack® optical sensor by LIC Automation Limited (www.licautomation.co.nz), an AfiLab® fat, protein and lactose concentration sensor by Afimilk Limited (www.afimilk.com), or a Lely MQC® fat and protein concentration sensor by Lely (www.lely.com). TM volume, fat and protein sensors (www.licautomation.co.nz), an AfiLab® fat, protein and lactose concentration sensor by Afimilk Limited (www.afimilk.com), or a Lely MQC® fat and protein concentration sensor by Lely (www.lely.com). TM volume, fat and protein sensors (www.licautomation.co.nz), an AfiLab® fat, protein and lactose concentration sensor by Afimilk Limited (www.afimilk.com), or a Lely MQC® fat and protein concentration sensor by Lely (www.lely.com). TM volume, fat and protein sensors (www.licautomation.co.nz), an AfiLab® fat, protein and lactose concentration sensor by Afimilk Limited (www.afimilk.com), or a Lely MQC® fat and protein concentration sensor by Lely (www.lely.com).

[0025] The optical sensor can be sensitive to an optical property of the milk, in particular the colour and / or the degree of optical scattering of the milk. It can be believed that the optical sensor can exhibit animal-specific bias related to the milk properties contributing to the colour and optical scattering of the milk. It will be appreciated that while reference can be made to the first type of sensor being an optical based sensor, it can be clearly contemplated that examples of the present technology can be applied to systems using other sensor types that are sensitive to animal-specific bias.

[0026] In an example, it is envisaged that the first type of sensor can be an in-line sensor. Reference to an in-line sensor herein should be understood to mean a sensor that analyses fluid flowing through one or more sensing devices to determine at least one parameter of the fluid at a particular point in time or across a time period, i.e. without collecting discrete samples from the flow. Dairy plants often include separate milk transport conduits from extraction points (e.g. using milking claw sets comprising teat cups) that are connected to a common transport line for delivery to a storage vessel. It is known in the art to provide in-line sensors within the separate milk transport conduits, allowing analysis of the milk extracted from individual animals as it flows through those separate milk transport conduits. It is often desirable to associate the sensor with as high a percentage of the milking claw sets as possible to enable high frequency data collection for individual animals (i.e. data is collected each time they are milked).

[0027] As a further example, it can be believed that animal-specific bias can be observed in in-line volume measurements, where milk volume is derived from the cross-sectional area of the milk flow. It is believed that variations in the shape of the animal teat or hair coverage between animals can result in differences in the amount of air entering the teat cup of a milking claw set. This can affect the cross-sectional area, speed or air content of the milk through such in-line sensors, potentially leading to biased measurements.

[0028] The need to install such sensors in relatively high numbers provides a practical driving force to reduce their price point, which contributes to their having lower precision than other known sensors. Additionally, the fact that inline milk sensors analyze milk as it flows past them prevents the use of sample handling that can improve the measurement, as is known in devices that analyze discrete samples of milk. For example, known ultrasonic milk analyzers precisely control milk temperature to achieve higher precision measurements. Known mid-infrared analyzers also control milk temperature and require much narrower measurement cells than can be placed in-line with a sensor, typical tubing for milk flow. Other handling, including bubble elimination, addition of reagents, and homogenization, can be used with sensors that analyze discrete samples rather than in-line sensors and can improve measurement performance. Furthermore, in-line sensors need to be made using materials and geometries to meet the hygiene requirements of milking systems, which also contributes to their relatively low precision. For completeness, it is contemplated that aspects of the present technology can be applied to systems that utilize a first type of sensor that is not configured for in-line installation.

[0029] Conversely, due to cost considerations (e.g., due to high capital cost or ongoing use of consumables), it is generally not commercially feasible to install a sensor for each milking cup set that is insensitive to animal-specific bias and generally higher precision (i.e., a second type of sensor). For example, a second type of sensor can utilize measurement techniques that use ultrasound, acoustics, electromagnetic radiation (e.g., near-infrared or mid-infrared), and electronic impedance. For example, a second type of sensor can implement the off-line LactiCheck® milk analyzer by Page & Pedersen International, Inc. (www.pagedersen.com) or the off-line MIRIS® milk analyzer by Miris Holding AS (www.mirissolutions.com). TM For example, a second type of sensor can implement the off-line LactiCheck® milk analyzer by Page & Pedersen International, Inc. (www.pagedersen.com) or the off-line MIRIS® milk analyzer by Miris Holding AS (www.mirissolutions.com). TM For example, a second type of sensor can implement the off-line LactiCheck® milk analyzer by Page & Pedersen International, Inc. (www.pagedersen.com) or the off-line MIRIS® milk analyzer by Miris Holding AS (www.mirissolutions.com). TM For example, a second type of sensor can implement the off-line LactiCheck® milk analyzer by Page & Pedersen International, Inc. (www.pagedersen.com) or the off-line MIRIS® milk analyzer by Miris Holding AS (www.mirissolutions.com). TM For example, a second type of sensor can implement the off-line LactiCheck® milk analyzer by Page & Pedersen International, Inc. (www.pagedersen.com) or the off-line MIRIS® milk analyzer by Miris Holding AS (www.mirissolutions.com).

[0030] In an example, the second type sensor can be a sensor system comprising multiple types of sensing devices, e.g. sensing devices of the first type sensor in addition to another sensor type, which collectively provide a measurement less susceptible to animal-specific bias than the first type sensor in isolation. Further details of such a sensor system can be found in, for example, PCT patent application number PCT / NZ2018 / 050153 entitled “System and Method for Analysis of a Fluid”, which is incorporated herein by reference. It will be appreciated that in such an example, the animal-specific bias correction can be determined based on sensor values from (a) the included first type sensor in isolation (i.e. “at least one first type sensor value for the parameter”) and (b) the sensor system which can be partially determined using the first type sensor value (i.e. “at least one second type sensor value for the parameter”). Furthermore, in such an example, it will be appreciated that the animal-specific bias correction can be applied to values obtained from the first type sensor, rather than values included in the second type sensor.

[0031] Aspects of the technology address animal-specific bias by correcting measurements of the first type sensor using measurements of the second type sensor, thereby compensating for respective limitations of the first type sensor and the second type sensor.

[0032] In an example, determining the animal-specific bias value comprises determining, each time the animal is milked using a teat cup set having an associated first type sensor and second type sensor, a difference between the first type sensor value for the parameter and the second type sensor value for the parameter.

[0033] In an example, determining the animal-specific bias value comprises determining, over a time period comprising a plurality of instances in which the animal is milked, an average of the difference between the first type sensor value for the parameter and the second type sensor value for the parameter.

[0034] In one example, the time period can be a complete lactation period (i.e. for a dairy cow, the time period between one calving and the next calving for the cow). In one example, the time period can be a portion of a lactation period, e.g. one or more stages of a lactation cycle.

[0035] In an example, the animal-specific bias correction can be applied retrospectively, i.e. at the end of the time period, the animal-specific bias correction can be applied to past low-precision results. This can be particularly applicable when the results are used for the purpose of animal evaluation at the end of a season, for example.

[0036] In an example, the moving animal-specific bias correction can be used from the start of the time period, for example for daily animal management purposes. It is expected that multiple results can be required before the animal-specific bias correction becomes reliable. It will be appreciated that the results can continue to be updated retrospectively as new data is obtained, i.e. once a new data point for an animal is obtained and the animal-specific bias correction is updated, the corrected result is overwritten.

[0037] In an example, determining the animal-specific bias correction takes into account trends across the time period. For example, the animal-specific bias value can be determined by fitting a curve over time to a relationship between the first type of sensor value for a parameter and the second type of sensor value for the parameter. For example, it is envisaged that the animal-specific bias can shift throughout lactation. To illustrate this, the equation for calculating the animal-specific bias correction from a predetermined time (e.g. days post-calving) can be fitted for each animal.

[0038] In an example, the determination of the animal-specific bias correction can exclude data from instances of the animal being milked where the first type of sensor value for a parameter and / or the second type of sensor value for the instance is determined to be an outlier over the time period. In an example, once the animal-specific bias correction is determined, the first type of sensor value for a parameter can be adjusted by the animal-specific bias correction to provide an error correction, i.e. to produce a corrected first type of sensor value for the parameter. For completeness, it will be appreciated that reference to an error correction is intended to mean at least a reduction in the animal-specific error component compared to a measurement determined from the first type of sensor alone.

[0039] In an example, determining whether a parameter value is an outlier can comprise determining whether the parameter value is an untrustworthy result, i.e. not within a reasonable biological range. In an example, determining whether a parameter value is an outlier can comprise determining whether the parameter value is a‘contemporary group’ outlier, i.e. a statistical outlier compared to other animals within a contemporary group. A contemporary group is defined as a group of animals having one or more comparable statistics (e.g. breed, age and / or stage of lactation). In an example, determining whether a parameter value is an outlier can comprise determining whether the parameter value is an intra-animal outlier, i.e. a statistical outlier compared to other results for the same animal at a similar stage of lactation. In an example, determining whether a parameter value is an outlier can comprise determining whether the parameter value is an intra-animal difference outlier, i.e. an outlier arising from a milking where the difference between the first type of sensor result and the second type of sensor result is a statistical outlier compared to differences from different milkings of the same animal at a similar stage of lactation.

[0040] In an example, parameter values in respective data sets comprising first type sensor values and second type sensor values can be calibrated for overall bias. In an example, a median of second type sensor results over all milking parlours for a specific day can be determined to provide a reference for calibration. In an example, a median of first type sensor values of the parameter for each milking parlour in a day can be determined. In an example, an adjustment value for a day can be determined as the difference between the median of second type sensor results over all milking parlours and the median of first type sensor values of the parameter for each milking parlour.

[0041] For firmware and / or software (also referred to as computer program) implementations, the techniques of this disclosure can be implemented by way of instructions (e.g., software) that are processable by a processing device. Such instructions can be stored in memory and executed by the processing device. The instructions can be software executable, firmware, or both, and can be implemented in one or more computer programs. The software can be stored on any computer readable medium, for example, but not limited to RAM, ROM, EEPROM, flash memory, or a combination thereof. The software can be stored as a computer program product that can be loaded into the processing device. The processing device can be a microprocessor, but in the alternative, the processing device can be a controller, a microcontroller, a state machine, or any other processing device known in the art. The processing device can be a combination of one or more of the above devices or any other processing device known in the art. The processing device can be a component of a cloud computing device. The processing device can be a component of a server and network connectivity known in the art. As an example, the first type sensor and the second type sensor and the central processor can communicate with each other through a controller area network (CAN) bus system. Other performance sensors (e.g. flow or yield sensors), animal identification devices and milking plant sensors can also communicate with the central processor in a milking context. In one example embodiment, the animal identifier, data from the sensors and any other data can be stored in a data cloud.

[0042] The steps of a method, process, or algorithm described in connection with the present disclosure can be embodied directly in hardware, in a software module executed by one or more processors, or in a combination of the two. Each step or action in a method or process can be performed in the order shown or can be performed in a different order. Additionally, one or more process or method steps can be omitted or added to the methods and processes. Additional steps, blocks, or actions can be added to the methods and processes at any of their beginning, end, or intervening points. BRIEF DESCRIPTION OF DRAWINGS

[0043] Other aspects of the technology will become apparent from the following description, taken in conjunction with the accompanying drawings, illustrating the principles of the technology in example embodiments.

[0044] Figure 1A is a schematic diagram of an example livestock management system that can implement an aspect of the technology;

[0045] Figure 1B is a schematic diagram of an example first sensor arrangement for use in an example livestock management system;

[0046] Figure 1C is a schematic diagram of an example second sensor arrangement for use in an example livestock management system;

[0047] Figure 2 is a flowchart of a method of analysing milk in accordance with aspects of the technology;

[0048] Figure 3 is a scatter plot of milk fat error results for a first type of sensor and a second type of sensor for individual animals;

[0049] Figures 4A-4C is a scatter plot showing the correlation between sensor results and herd test results for three measurement types of fat (first type of sensor, second type of sensor, first type of sensor corrected for animal-specific bias) at individual milking levels;

[0050] Figures 5A-5C is a set of scatter plots showing the correlation between sensor results and herd test results for three measurement types of protein (first type of sensor, second type of sensor, first type of sensor corrected for animal-specific bias) at individual milking levels;

[0051] Figures 6A-6C is a set of scatter plots showing the correlation between sensor results and herd test results for three measurement types of fat (first type of sensor, second type of sensor, first type of sensor corrected for animal-specific bias) at dairy cow average levels; and

[0052] Figures 7A-7C is a set of scatter plots showing the correlation between sensor results and herd test results for three measurement types of protein (first type of sensor, second type of sensor, first type of sensor corrected for animal-specific bias) at dairy cow average levels. DETAILED DESCRIPTION

[0053] Aspects of the technology are described herein in the context of milk analysis. However, it will be appreciated that the principles of the disclosure discussed herein can be applied to the analysis of other fluids.

[0054] Figure 1AA livestock management system 100 is illustrated in which a local hardware platform 102 manages the collection and transmission of data related to the operation of a milking facility. The hardware platform 102 has a processor 104, a memory 106, and other components typically found in computing devices. In the illustrated example embodiment, the memory 106 stores information that is accessible to the processor 104, including instructions 108 that can be executed by the processor 104, and data 110 which can be retrieved, manipulated or stored by the processor 104. The memory 106 can be of any suitable type known in the art including a computer readable medium or other medium that stores data which can be read by an electronic device. The processor 104 can be any suitable device known to those having ordinary skill in the art. While the processor 104 and the memory 106 are illustrated as being within a single unit, it is understood that this is not intended to be limiting and that the functionality of each described herein can be performed by multiple processors and memories which can be located remotely from one another or which can not be located remotely from one another. The instructions 108 can comprise any set of instructions suitable for execution by the processor 104. For example, the instructions 108 can be stored as computer code on a computer readable medium. The instructions can be stored in any suitable computer language or format. The data 110 can be retrieved, stored or modified by the processor 104 in accordance with the instructions 110. The data 110 can also be formatted in any suitable computer readable format. Moreover, while the data is shown to be contained in a single location, it is understood that this is not intended to be limiting and that the data can be stored in multiple memories or locations. The data 110 can also include records 112 for aspects of the system 100 by control routines.

[0055] The hardware platform 102 can communicate with various devices associated with a milking facility, such as: a first type of sensor 150a-150n associated with a plurality of individual clusters of milking cups within the milking facility, and a second type of sensor 152a-152(n-x) associated with a subset of the individual clusters of milking cups. Reference can be made herein to milk collected at or from a milking stall. A milking stall is a place within a milking facility where an animal can be placed to be milked. In some milking facilities, clusters of milking cups are associated with milking stalls in a one-to-one relationship (e.g., in a typical rotary milking parlor), while in other milking facilities, clusters of milking cups can be shared between two or more milking stalls (e.g., in a herringbone configuration).

[0056] Figure 1BA first type of sensors 150a-150n and a second type of sensors 152a-152(n-x) are illustrated as connected with the hardware platform 102 over a controller area network (CAN) bus. It will be appreciated that although not illustrated, additional performance sensors (e.g. performance sensors such as milk flow or yield sensors) can also be connected to and communicate over the CAN bus. Each of the first type of sensors 150a-150n and the second type of sensors 152a-152(n-x) is associated with a respective teat cup cluster in the milking facility, i.e. the sensor data output by the respective sensors is related to the milk of an individual animal being milked by that teat cup cluster.

[0057] In an example, the first type of sensors 150 can be in-line sensors configured to determine at least the fat and / or protein content of the milk, e.g. Protrack Milk TM Volume, Fat and Protein sensors by LIC Automation Ltd, or AfiLab TM Fat, Protein and Lactose concentration sensors by Afimilk Ltd, or Lely MQC TM Fat and Protein concentration sensors by Lely. In an example embodiment, a first type of sensor 150 can be provided for each teat cup cluster in the milking facility. However, it will be appreciated that this is not intended to limit each embodiment of the present disclosure. For example, it is conceivable that only a subset of the teat cup clusters can have an associated first type of sensor 150.

[0058] According to aspects of the present technology, the second type of sensors 152a-152(n-x) are provided on a less than one-to-one basis with the first type of sensors 150a-150n, i.e. the second type of sensors 152a-152(n-x) are only provided for a subset of those teat cup clusters that also have a first type of sensor 150a-150n.

[0059] The second type of sensors 152 are configured to analyze the milk for at least one of the same parameter(s) as the first type of sensors 150, but are less affected by animal-specific bias. In an example, the second type of sensors 152 can implement an offline LactiCheck TM milk analyzer by Page & Pedersen International A / S performing an ultrasound-based sensing method, or an offline MIRIS TM milk analyzer by Miris Holding performing a mid-infrared based sensing method. In an example, the second type of sensors 152 are configured to analyze a milk sample obtained from the milk extracted by the associated teat cup cluster (e.g. using sampling equipment to pass the extracted milk sample to one of the above-mentioned offline sensors).

[0060] Reference is made to Figure 1C In an alternative example, the second type sensor 152 can be a sensor system 160 comprising multiple types of sensing devices, for example, first system sensors 162 equivalent to the first type sensors 150 and second system sensors 164. Results from the first system sensors 162 and the second system sensors 164 can be used collectively to produce a measurement value that is less susceptible to animal-specific bias than the first type sensor 150 in isolation (i.e. to provide a “second sensor type” value). Further details of the operation of such a sensor system can be found in, for example, PCT Patent Application No. PCT / NZ2018 / 050153. It will be appreciated that the example sensor system 160 can thus produce both first type sensor values and second type sensor values for milk being analysed.

[0061] Returning to Figure 1A The animal identification devices 154a to 154n are provided for determining animal identification (“animal ID”) of individual animals entering or within the milking facility. More specifically, the animal identification devices 154a to 154n can be used to associate an animal ID with each of the teat cup clusters associated with the first type sensors 150a to 150n (and the second type sensors 152a to 152n) such that sensor data can be attributed to individual animals. A number of methods are known for determining animal ID, for example, radio frequency identification (“RFID”) readers configured to read RFID tags carried by the animals. In alternative embodiments, or in conjunction with the animal identification devices 154a to 154n, a user can manually input (or correct) animal IDs via a user device.

[0062] The hardware platform 102 can also communicate with user devices, such as a touchscreen 120 and local workstation 122 located within the milking facility for monitoring system operations. The hardware platform 102 can also communicate with one or more server devices 126 over a network 124, the server devices 126 having associated memory 128 for storing and processing data collected by the local hardware platform 102. It will be appreciated that the server devices 126 and memory 128 can take any suitable form known in the art, for example a “cloud-based” distributed server architecture. The network 124 can include various configurations and protocols, including the Internet, an intranet, a virtual private network, a wide area network, a local area network, a private network using one or more proprietary communication protocols of a company, whether wired or wireless, or combinations thereof. It will be appreciated that the illustrated network 124 can include different networks and / or connections: for example, a local network through which the user interface can be accessed near the milking facility, and an Internet connection through which the cloud server is accessed. Information about the operation of the system 100 can be communicated over the network 124 to user devices such as a smartphone 130 or tablet 132.

[0063] Reference is made to Figure 2 A method 200 is provided for analyzing milk extracted from individual animals by the system 100. In a first step 202, milk extracted from an individual animal by a set of milking cups is analyzed using an associated first type sensor 150 to obtain first type sensor values for a parameter of the milk (e.g. fat and / or protein content). The animal is identified and the animal identification is recorded together with the first type sensor values. In a second step 204, the milk is also analyzed by a second type sensor 152 associated with the same set of milking cups to obtain second type sensor values for the parameter of the milk, the second type sensor values being recorded against the animal identification.

[0064] In a third step 206, the first type sensor values and the second type sensor values for the parameter are used to determine an animal-specific bias value for the individual animal. In one example, the animal-specific bias value can be the difference between the first type sensor values and the second type sensor values for the parameter. In one example, the animal-specific bias value is an average of the difference between the first type sensor values and the second type sensor values over a time period including multiple instances in which the animal is milked (e.g. a lactation period of the animal or a partial lactation period of the animal). In examples, the animal-specific bias value can include a regression coefficient determined by applying a linear regression to the difference between the first type sensor values and the second type sensor values for the animal over the day of milking.

[0065] In an example, one or more automatic outlier detection processes can be applied to the data sets comprising the first type sensor values and the second type sensor values, respectively, to remove such outliers prior to determining the animal-specific bias values. For example, the outlier detection processes can be implemented in the form of software scripts. In an example, the outlier detection can comprise one or more of: determining whether a parameter value is an untrustworthy result, determining whether a parameter value is a same-period group outlier, determining whether a parameter value is an intra-animal outlier, and determining whether a parameter value is an intra-animal poor outlier.

[0066] In an example, the parameter values in the respective data sets comprising the first type sensor values and the second type sensor values can be calibrated for overall bias. In an example, the second type sensor values can be calibrated for overall bias using bulk tank data as a reference, where bulk tank data refers to parameter values obtained for milk collected from a bulk tank collecting milk from all milking parlours. In one embodiment, the calibration for overall bias can comprise determining (for each date for which data is collected) a median of the second type sensor results across all milking parlours (“all-parlour second type median”) to provide a benchmark for calibration. For each milking parlour, a median of the first type sensor values for the parameter (“current-parlour first type median”) can be determined, and a parlour-by-day adjustment determined as the all-parlour second type median minus the current-parlour first type median. The parlour-by-day adjustment can be applied to all first type sensor results by adding it to the raw results to produce adjusted results for further processing. It is envisaged that this can reduce bias between milking parlours and thus produce less noise in the separate estimation of animal-specific bias. In an example, if for a particular day a predetermined number of results for a milking parlour are not recorded, the adjusted results can be excluded from further analysis.

[0067] In a fourth step 208, the first type sensor values for the milk parameter are adjusted by subtracting the animal-specific bias values from the first type sensor values obtained for the individual animals. These adjusted values can then be used for further data analysis and decision making known in the art.

[0068] It is envisaged that aspects of the present technology can be particularly applicable to examples in which the system 100 is installed in a rotary milking parlour. Some milking animals, such as dairy cows, can be highly consistent in their own order of appearance for milking. As a result, in some milking parlour configurations, there can be a relatively high likelihood that animals are consistently milked at the same or similar milking stall cluster of teat cups. This can result in a situation in which certain animals are less likely to be milked by a cluster of teat cups having an associated second type sensor that can be used to determine an animal-specific bias for that animal. In a rotary milking parlour, when the milking platform is continuously rotated, the order of appearance of the animals for milking themselves has no impact on the assignment to a particular milking stall. As a result, the assignment of animals to a milking stall having an associated second type sensor is essentially random, thereby increasing the likelihood of developing an animal-specific bias correction for each animal within a cluster.

[0069] In examples, the order of animal entry into the milking parlour can be controlled (e.g. using a draw gate) to encourage the assignment of animals to milking stalls having second type sensors.

[0070] Experimental example: Correction using Protrack TM Milk obtained fat and protein sensor results

[0071] The following describes an experimental implementation of the present technology in the form of correction of fat and protein sensor results obtained using Protrack TM Milk (“PT-Milk”) sensors were installed on 17 milking stalls, while second type sensors (in the form of a sensor system 160 as described above with reference to Figure 1C to a second system sensor 164 being an ultrasonic based sensor) were installed on 17 other milking stalls, i.e. sensors were installed on a total of 34 milking stalls.

[0072] In this example, sensor results for fat and protein were adjusted by comparing seven day averages with average results of regularly collected milk samples tested using a laboratory based reference method (referred to herein as “herd testing”). It should be noted that in practice, sensors can be regularly calibrated by comparison with herd average fat and protein determined from a large number of milk sample results provided by a milk processing company, and as such can not require this form of result adjustment. In alternative examples, overall bias correction as described above can be utilised.

[0073] In this experimental implementation, data from 31 cows with at least 8 milking events were used to calculate performance metrics using valid herd test, PT-Milk, and ultrasound-based sensor results.

[0074] The data set was filtered for unreliable measurements. In this experiment, initially unreliable results were eliminated based on their visual inconsistency with other results for the animal. PT-Milk fat and protein results that were inconsistent with the animal trend were flagged as outliers. If fat or protein was identified as an outlier, the fat and protein results from PT-Milk were excluded from subsequent analysis. Ultrasound-based sensor fat and protein results were flagged as outliers and excluded from subsequent analysis in the same manner.

[0075] Performance statistics were calculated only for milking results with valid PT-Milk, ultrasound-based sensor, and herd test results from 31 cows with more than 8 fully matched milking events ("fully matched milking events"). Using the herd test results as ground truth, standard deviations and mean errors of individual test results for PT-Milk and ultrasound-based sensor were determined. Standard deviations and mean errors of cow means were also determined as a measure of animal-specific bias within the herd.

[0076] Animal-specific bias (ASB) correction for each cow was determined using the assumption that 6% stall coverage (2 / 34 stalls) would provide a sufficient number of tests to estimate animal-specific bias. For the two stalls with system sensors (Stall 2 and Stall 3), all milking events with valid PT-Milk and ultrasound-based results were used to calculate ASB correction for each of the 31 cows. Each cow had a minimum of 13 milking events and an average of 21 milking events. Figure 3 Differences between ultrasound and PT-Milk results for an example cow are shown, where the circles represent measurements on either Stall 2 or Stall 3 that were used to calculate the ASB correction for that cow (indicated by the solid line). The fat ASB correction for each cow was calculated as the cow-averaged difference between ultrasound-based and PT-Milk fat results. Protein ASB correction for each cow was calculated in the same manner using protein results. All PT-Milk results were adjusted using the individual ASB corrections. As described above, performance statistics were calculated from the adjusted PT-Milk results.

[0077] Table 1, Figures 4A-4C , Figures 5A-5C , Figures 6A-6C and Figures 7A-7CThe performance of three milk composition estimates is shown: PT-Milk, the second type of sensor based on ultrasound, and PT-Milk adjusted using ASB correction (“adjusted PT-Milk”).

[0078] Table 1: Performance statistics for the three milk composition estimates (g / 100 mL) at individual test and cow average level.

[0079]

[0080] At cow average level, the standard deviation (SD) of the cow average error values for PT-Milk and the ultrasound based sensor were consistent with those obtained previously in a separate experiment. The SD of the cow average error values for adjusted PT-Milk was smaller than for PT-Milk: the cow average error SD for adjusted PT-Milk was 0.19 g / 100 mL for fat, compared to 0.43 g / 100 mL for PT-Milk alone, and 0.15 g / 100 mL for protein, compared to 0.25 g / 100 mL for PT-Milk alone. It can thus be seen that the influence of ASB was significantly reduced by applying the respective ASB correction.

[0081] Furthermore, the SD of the adjusted cow average error for PT-Milk was similar to the ultrasound based sensing method. This indicates that the low ASB of the second type of sensing based on ultrasound can be achieved on PT-Milk using ASB correction, which is determined from ultrasound based sensors installed at only 6% of the milking stalls (2 out of 34 in total in this case). This allows high milking stall coverage to be achieved with high precision using the lower cost PT-Milk sensor.

[0082] At individual test level, the error SD for PT-Milk was 0.55 and 0.24 g / 100 mL for fat and protein, respectively. This is consistent with the performance of this technology measured previously. The error SD for ultrasound based fat (0.39 g / 100 mL) was similar to that measured previously, but the error SD for ultrasound based protein (0.40 g / 100 mL) was slightly higher than measured previously. The inventors hypothesize that this can be due to the relatively weak outlier detection method applied in this experiment, and that if a more stringent outlier detection was applied, the ultrasound based protein SD error would be expected to be improved.

[0083] The inventors observed that the error SD for the adjusted PT-Milk protein (0.22 g / 100 mL) was superior to the ultrasound-based protein (0.40 g / 100 mL). It can be believed that the reason for this is that the bulk of the ultrasound-based error is random error. In many of the tests comparing PT-Milk and ultrasound-based results, the ASB correction was averaged, which is believed to reduce random error, leading to good estimates of ASB. In contrast, the PT-Milk error is believed to be mostly ASB, so when the ASB correction is applied to individual results, very good individual measurements are obtained.

[0084] The inventors note that the implementation in this experiment was relatively simple. More complex methods can be used, potentially improving the resulting ASB correction. For example, outlier detection of the ultrasound-based measurements is believed to be an area for refinement, and excluding outliers based on their being an unusual trend for the cow can be automated. Another layer of individual cow outlier detection can also be applied to the individual ASB estimates.

[0085] Furthermore, in this experiment, the ASB correction was constant during the data being corrected. For completeness, it is envisioned that the ASB correction can be adjusted throughout lactation, for example using a polynomial curve fitted to selected milking stalls as shown in Figure 3

[0086] The present technology provides methods and systems for correcting animal-specific bias in automated milk analysis sensors that are sensitive to such bias.

[0087] The entire disclosure of all applications, patents and publications cited above are hereby incorporated by reference.

[0088] The reference in this specification to any prior publication or patent does not constitute an admission that the prior publication or patent is part of the common general knowledge in any country in the world.

[0089] The present invention can also be generally described as including parts, elements and features referred to or indicated in the specification of this application, individually or collectively, and any or all combinations of any two or more of said parts, elements or features.

[0090] In the preceding description, reference has been made to the use of means or apparatus for performing various functions. It is to be understood that such means or apparatus are merely illustrative of the application of the principles of the present invention. Numerous other arrangements can be devised by those skilled in the art without departing from the scope of the present invention as will be set forth in the following claims.

[0091] ​It should be noted that various changes and modifications to the presently preferred embodiments described herein will be apparent to those skilled in the art. Such changes and modifications can be made without departing from the spirit and scope of the application and without diminishing its attendant advantages. It is therefore intended that such changes and modifications be included within the scope of the application.

[0092] Aspects of the application have been described by way of example only and it should be appreciated that modifications and additions can be made thereto without departing from the scope thereof.

Claims

1. A system for analyzing milk, comprising: a plurality of first type sensors, each first type sensor being associated with a respective one of a plurality of teat cup sets of a milking system and configured to analyze milk extracted by the teat cup set from an individual animal to determine at least one first type sensor value of a parameter of the milk across an event period; at least one of a second type sensor, the second type sensor being associated with at least one of the plurality of teat cup sets and configured to analyze the milk extracted by the teat cup set from the individual animal to determine at least one second type sensor value of the parameter of the milk within the event period, wherein the second type sensor is less susceptible to animal-specific bias than the first type sensor, and wherein a number of the second type sensors in the system is less than a number of the first type sensors; and at least one processor configured to: determine an animal-specific bias correction for the individual animal based on the at least one first type sensor value of the parameter and the at least one second type sensor value of the parameter determined for the individual animal; and apply the animal-specific bias correction to first type sensor values of the parameter obtained from the first type sensors for milk extracted from the individual animal.

2. The system of claim 1, wherein the first type sensor is an optical sensor.

3. The system of claim 1 or claim 2, wherein the first type sensor is an in-line sensor.

4. The system of claim 1 or claim 2, wherein the at least one second type sensor utilizes a measurement technique comprising one or more of: ultrasound, acoustics, electromagnetic radiation, and electronic impedance.

5. The system of claim 1 or claim 2, wherein the determination of the animal-specific bias correction by the at least one processor comprises: determine a difference between the first type sensor value of the parameter and the second type sensor value of the parameter each time the animal is milked with a teat cup set having an associated first type sensor and second type sensor.

6. The system of claim 5, wherein the determination of the animal-specific bias correction by the at least one processor comprises: determine an average of the difference between the first type sensor value of the parameter and the second type sensor value of the parameter over a period of time comprising a plurality of instances in which the animal was milked.

7. The system of claim 6, wherein the period of time is a complete lactation.

8. The system of claim 6, wherein the period of time is a portion of a lactation.

9. The system of any one of claims 6 to 8, wherein the animal-specific bias correction is a moving animal-specific bias correction used from a beginning of the period of time.

10. The system of any one of claims 1, 2, and 6-8, wherein the determination of the animal-specific bias correction by the at least one processor comprises: determine the animal-specific bias correction taking into account a trend across the period of time.

11. The system of claim 10, wherein the animal-specific bias correction is determined by fitting a curve to the difference between the first type sensor value of the parameter and the second type sensor value of the parameter over the period of time.

12. The system according to any one of claims 1, 2, 6-8 and 11, wherein the at least one processor is configured to retrospectively apply the animal-specific bias correction.

13. The system of any one of claims 1, 2, 6-8, and 11, wherein determining, by the at least one processor, the animal-specific bias correction comprises: Exclude data from instances of the milked animal where the first type of sensor value of the parameter and / or the second type of sensor value of the instance are determined to be outliers.

14. The system according to any one of claims 1, 2, 6-8 and 11, wherein the parameter of the milk is fat.

15. The system according to any one of claims 1, 2, 6-8 and 11, wherein the parameter of the milk is protein.

16. The system according to any one of claims 1, 2, 6-8 and 11, wherein each of the first type sensor and the at least one second type sensor is configured to determine values ​​for a plurality of parameters for the milk, wherein the plurality of parameters include at least milk and fat.

17. A method for analyzing milk in a system having multiple milking cup sets, each milking cup set configured to extract milk from an individual animal, the method comprising: Using a first-type sensor associated with the milking cup set, milk extracted from an individual animal by the milking cup set is analyzed to determine at least one first-type sensor value of a parameter of the milk across event time periods. Using a second type of sensor associated with the milking cup set, the milk extracted from the individual animal by the milking cup set is analyzed to determine at least one second type of sensor value of the parameter of the milk during the event period, wherein the second type of sensor is less susceptible to animal-specific bias than the first type of sensor, and wherein the number of the second type of sensor in the system is less than the number of the first type of sensor. as well as Based on at least one first-type sensor value and at least one second-type sensor value of the parameter determined for the individual animal, an animal-specific bias correction is determined for the individual animal; and The animal-specific bias correction is applied to the first-type sensor value of the parameter obtained from the first-type sensor for milk extracted from the individual animal.

Images