Final ReportSummary - BOVINOSE (Pheromone-based sensor system for detecting estrus in dairy cows)
In order to control the calving interval, and therefore to optimise milk production and maximise offspring in dairy cattle, artificial insemination (AI) is widely used. Currently, the conception rate of AI varies from 30 % to 70 %, and a successful insemination depends mainly on the right timing. Highest conception rates are achieved when the insemination is done 8 to 12 hours after the start of estrus (also known as 'standing heat'). The problem, therefore, narrows down to detection of estrus in a cow.
The BOVINOSE project aims to develop a novel type of 'electronic nose' to detect estrus in a dairy cow, and thus to determine the optimal timing of artificial insemination. The physical principle is based on detection of sex pheromones that are secreted by the cow, exclusively during estrus. These pheromones are the natural olfactory signal for the bull that the cow is in heat. The system will be an easy-to-use device that enables a dairy farmer to accurately determine a cow's ovulation timing. To this end, the device will be equipped with sensors, able to detect the specific bovine sex pheromones
This technology will help the dairy farmers in the EU, the vast majority being micro-enterprises run as family businesses. It is an opportunity for a number of SMEs that are active in supplying the dairy farmers with technology solutions. Simply put, in essence, a bull's nose will be mimicked.
Partners in the BOVINOSE consortium have been selected with a clear added value and complementarity in mind.
SMEs have teamed up from both the supply chain (manufacturers of electronic equipment and control systems, as well as AI solution providers), and from the end user perspective (dairy farmer).
The expertise of pheromone research, both on fundamental issues and on chemical characterisation, as well as application research on sensor development and signal processing, is incorporated in the project consortium. Veterinary know-how is brought in, particularly on the cow's estrus cycle and recognition of behavioural signs typical for cows in estrus.
Research institutes, represented by scientists and engineers of different disciplines support SMEs to develop and manufacture a novel and easy-to-use device that enables a dairy farmer to accurately determine a cow's ovulation timing.
The info-chemicals that make up the sex pheromones in cows have been extensively investigated using natural and induced samples from faeces, urine, and saliva from various dairy farms. Faeces were determined as most suitable source of sex pheromone based on the highest responses of bulls, largest emission rates of pheromone components, user-friendliness, and application conditions compare to the urine samples. Assessments were made to assess fine scale dynamics of volatile pheromone components as well as for measurements by sensors. The acetic and propionic acids were investigated and qualified as sufficient indicators of reproductive status of cows and could be used to define optimal insemination time.
Required BOVINOSE system properties could be identified. A set of sensors for detection of these sex pheromones was researched, designed and developed, resulting in a functional prototype. This was based on a set of defined user and technical requirements, consisting of a probe, an array of sensors and a dedicated, integrated software system for control and end-user interaction.
Extensive field trials closely accompanied by veterinarian experts have been run at a dairy farm, both on the info-chemicals as on the fitness for use for the anticipated end-user by performing a technical evaluation. The prototype that was tested under farm conditions proved to be able to measure volatiles from faces samples. There was good correlation between heat and pheromone peaks in the laboratory setting. The analyses showed that silent heat resulted in pheromone peaks. The prototype device performed under field conditions equally well as under laboratory conditions, however.
Project context and objectives:
The overarching goal of BOVINOSE is to develop a novel and easy-to-use device that enables a dairy farmer to accurately determine a cow's ovulation timing. To this end, the device will be equipped with sensors, able to detect the specific bovine sex pheromones. Simply put, in essence, a bull's nose will be mimicked.
The main scientific and technological objectives of BOVINOSE are the following:
- to identify the info-chemicals that make up the sex pheromones in cows that are uniquely released during estrus;
- to develop a set of sensors for detection of bovine sex pheromones;
- to develop an easy-to-use, yet reliable system for estrus determination in cows, based on detection of the sex pheromones;
- to develop and manufacture a functional prototype, consisting of a probe, an array of sensors and dedicated, integrated software for control and user-interaction.
The main objectives for the first twelve months of the BOVINOSE project have been dedicated to:
- defining and specifying the BOVINOSE system requirements as seen from an end-user, and the technical requirements for modules and components;
- researching sex pheromones for cows;
- design and develop prototypes of BOVINOSE sensors, both hardware and software parts.
The main objectives for the second (and final) part of the BOVINOSE project have been dedicated to:
- the development and building of a functional prototype (milestone Mi-3);
- software development programmed in Matlab, tested, and debugged;
- design and development of auxiliary tools and a graphical user-interface (GUI) for operating the BOVINOSE system;
- validation of the prototype system by extensive laboratory testing;
- full testing using a two-staged procedure of field trials at a dairy farm closely accompanied by veterinarian experts;
- full evaluation of the first version of the prototype of the BOVINOSE device resulting in new functionalities and improvements;
- final project evaluations and recommendations for further research, comprising a technical evaluation report, a cost price analysis, and further recommendations for further research;
- definition and finalisation of the required levels of dissemination and preparation for future exploitation.
All of these objectives have been fulfilled at the end of the project.
Project results:
System definition and specifications
The definition of systems requirements from a user viewpoint, in terms of performance, user-friendliness and application conditions; the specification of technical requirement for the system modules and components were all covered. This increased the reliability of the end system, what will increases the user satisfaction level.
System performance specifications
Major results on system performance specifications have been the definition of:
- basis for electronic pheromone identification;
- requirements of end user (both general considerations, functional and physical requirements);
- general structure of electronic pheromone detection and identification system.
Sensor specification
The technical requirements of the sensor have been specified, in terms of standard operating conditions, environmental conditions, sensitivity characteristics, and electro-mechanical characteristics. Typical parameters include power consumption, selectivity, accuracy, repeatability, drift, hysteresis, etc.
The metal oxide gas sensors are to be adapted for the BOVINOSE detector module. The sensors have to be adapted to function at normal indoor (and comparable) conditions at surrounding temperatures from about 10 °C to about 30 °C, relative humidity from about 30 % to about 80 %, oxygen rich atmosphere (not less than about 18 %). The sensors have to be fixed within the standard transistor container TO5. The sensitivity of the sensors should be not less than about 1 - 2 ppm of pure volatile compounds defined as pheromone components. The power consumption should not exceed about 500 mW per sensor and have to be used with a power source adapted for standard electricity supply.
The selectivity of the sensors has to be sufficient for the classification of the odours. The drift of the sensor response have to be less than about 10 % per continuous working cycle.
Considering the parameters of the sensors and possibilities to modify the characteristics of MO sensors it is reasonable to define the array of the BOVINOSE electronic odour detection system consisting of 8 MO gas sensors.
Software definition document
The BOVINOSE electronic pheromone detection system is to be equipped with the software packages that:
(i) allows control the hardware;
(ii) implements pre-processing of the sensor signals and creates a database of the output features;
(iii) performs analysis of the features; and
(iv) provides a user interface.
The technical requirements for these software packages data processing, data storage and control system have been specified.
Sampling module specifications
The odour sampling module is a peripheral tool of the BOVINOSE pheromone detection system that is optimised for sampling of volatile components specific to estrus in dairy cows. The sampling module will be modified and optimised for usage in farms in the final stage of the project. The second version of the sampling unit will be integrated within the detecting device.
The sampling unit is to be used with a standard container for collecting of faeces samples and adapted for the standard SPME fibre. The unit consists of the following components:
- Thermostat for stabilisation of temperature in the sample container during the SPME extraction of the volatile compounds from the headspace air. The working temperature of the thermostat has to be optimised during the laboratory tests.
- The headspace chamber with a holder for the special container of faeces samples and an adapter for the standard manual SPME holder, the volume of the headspace air is planned to be about 50 % of the sample volume but can be modified aiming to optimise the sampling system during the laboratory tests. The headspace chamber is adjusted to fit within the thermostat.
- (Optional) special sampling-cutting tool use as the faeces amount meter-container. The amount meter-container is to be included into the set of the sampling module if the variation in the featuring of the specific volatile compounds during the laboratory tests will be too high for reliable identification of the estrus in diary cows. Depending on the testing routine, the amount meter-container can be used instead of the standard container for sampling of faeces in the sampling unit.
- The materials of the sampling module have to be tested for the emissions of the interfering gases. The materials have to be optimised with respect to these emissions and the target volatile components.
- The methodology of usage of the sampling unit will be user-friendly and adjusted for the best effectiveness of the BOVINOSE device.
- The first version sampling unit will be used during the eNose test period in Vilnius and for the first stage reporting.
The first-stage prototype of the sampling module is a separate unit acceptable for use in the laboratory / room conditions. The module will be integrated within the complete BOVINOSE system in the second stage of development after the prove of the system acceptability for identification of the estrus in diary cows will be obtained and the project partners will agree to continue the project.
Specifications for the integrated system
BOVINOSE shall consist of the three main parts, namely the hardware, the software and the methodology for using the complete system. In the hardware, the key-part is the gas sensors and array of the sensors that determine reliability of the information acquired by the system. In the software, the key-part is the feature extraction routine and the database of the features that defines the most specific information about single test and makes it possible to compare this information with the list of the references. In the methodology, the key-part is the sampling of the right VOCs that makes it possible to obtain the advanced selectivity and sensitivity.
Pheromone research
Collection of biological samples
In total 140 samples of faeces, 105 samples of urine and 85 samples of saliva were collected in Lithuania from harmone induced ovulation in cows for assessment of bulls' reaction, measurements of release rates of sex pheromone components, to assess fine scale dynamics of volatile pheromone components as well as for measurements by sensors.
120 faeces samples were obtained from Holland from cows were under natural cycle. Samples were used for determination of correlation between amounts of the pheromone components and physiological cycle of cow as well as for measurements by sensors. The same amount of blood samples were collected and concentration of progesterone was determined.
Behavioural reaction of bulls to saliva, urine and faeces samples of cows in estrus
In total, reaction of 10 mature bulls was tested. The stimuli were applied on unestrus cow, as no reaction was recorded in absence of visual stimuli. Sixty percents of bulls showed mounting reaction to unestrus cow treated with sample of estrus faeces. Reaction to urine was less pronounced (35 %), and that to saliva was not noted at all (0 %).
Composition and usual amounts of bovine pheromone components
Acetic and propionic acids from three sex pheromone components known in cows were detected in headspace of faeces and urine by SPME equipped with carboxen-polydimethylsiloxane fibre. Both acids were present in pre-estrus, estrus and post-estrus samples of faeces and urine and reached their maximum amounts at some time before ovulation. Volatile sex pheromone components exhibit similar distribution pattern in samples of faeces and urine in respect to reproductive status of cows. During peak of release acetic and propionic acids occur at significantly higher amounts in faeces samples compare to those in urine. Faeces were determined as most suitable source of sex pheromone based on the highest responses of bulls, largest emission rates of pheromone components, user-friendliness, and application conditions compare to the urine samples.
Amounts of progesterone and volatile components of bovine pheromone during naturally occurring estrus cycles of cows
The monitoring of amounts of progesterone and volatile sex pheromone components, namely acetic and propionic acids revealed strong negative correlations between amounts of hormone and two volatile sex pheromone components in samples collected at the transition period from diestrus to estrus and at the first day of estrus which is a critical period for searching of optimal insemination time.
The acetic and propionic acids are sufficient indicators of reproductive status of cows and could be used to define optimal insemination time.
The fine scale dynamics of volatile pheromone components in estrus-induced cows
In the faeces samples of estrus-induced cows increase in amounts of acetic and propionic acids was registered around 24 hours before ovulation and after 10 - 12 hours dropped down to diestrus level. The dynamics of changes in amounts of acetic and propionic acids in the faeces samples obtained in pre-estrus and estrus periods occur in the similar manner in cows with naturally occurring and hormone induced estrus with clearly expressed peak of two sex pheromone components.
Sensor development
Prototype pheromone sensors
The prototype of sensors - a set of 24 sensors - was manufactured by the thin film technology and assembled within the standard TO5 transistor container. It consists of a sensitive layer with electrical contacts, mini-heater and thermometer. The surface of the sensitive layer was modified with catalyst non-homogeneous ultra thin film in order to improve sensitivity to pheromones. Optimal working temperature for pheromones detection was determined by test measurements. The prototype of sensors was tested and main characteristics were obtained.
A special sensor chamber is produced for the sensor module of the BOVINOSE electronic pheromone detection and recognition system.
Sensors response measurements
A description was made on VOC sampling methodology, calibration VOC and amount control, VOC injection module, measurement and control system, response measurement method, sensor parameters.
The sensors response was measured with the pure VOC, acetic acid and propionic acid that are defined being pheromones. It was found that qualitative evaluation of the sensor parameters can be performed by exposure to the headspace air of pure compounds.
The set of metal oxide gas sensors, namely SnO2-x with Pt, Au and Ru additives, were tested in the atmosphere with acetic acid, propionic acid and the VOCs from the faeces samples. It was found that qualitative evaluation of the sensor parameters can be performed by exposure to the headspace air of pure compounds. The calibration and correct evaluation of the sensor characteristics cannot be completed using non-calibrated sources of the target VOC. For this, the special sources with calibrated permeation rates were produced and the sensor calibration methodology based on these permeation tubes was developed.
Major conclusions were that the sensors in the array of BOVINOSE device have to be calibrated using reliable source of the pure target VOC because the real samples contain organic materials that are constantly changing due to processes related to the living organisms such as bacteria.
Improved prototype sensor
The sensors array was assembled and the tests with real samples from diary farm were performed. It was proved that the features obtained from the pheronose array allows to distinguish between the cows and condition of the cow during the testing time.
The sensor array was assembled from a set of various gas sensors produced in the Sensors Laboratory.
The sensors were calibrated with the pure pheromone compounds, namely acetic acid and propionic acid. The threshold amount of the pheromone components detectable by the sensors of the array is significantly below 0.1 ppm.
The real faeces samples obtained from diary farm were analysed with the pheronose. It was proved that the features obtained from the pheronose array allows to distinguish between the cows and condition of the cow during the testing time. Unfortunately, there are no possibilities to define the origin of the correlation because the biological description of the samples are absent at present stage of the investigation.
Major conclusions have been that the sensor array is acceptable for the featuring of the cow conditions. The pheronose device is acceptable for testing of the real samples in the laboratory conditions. Based on the results of present tests, some improvements can also be suggested for the next stage of the project.
Software development
A logistic regression / EM analysis of the Lithuania cow eNose data have been performed mainly by SMART and Protech. The model can be seen as a two-layer neural network. The training data consist of the data of the cows that are used for optimising the parameters (in the current model, these are the PCA and the regression parameters).
The main conclusions from this analysis have been the following:
- An EM algorithm for unsupervised learning has been developed using the prior knowledge that for the injected cows, there must be an estrus period within the given time period.
- The method is generally applicable for any probabilistic predictive model. Here, we used PCA analysis combined with logistic regression. (This could be interpreted as a particular two-layer neural network). However, other choices are possible.
- The EM algorithm, combined with a PCA -logistic regression model give reasonable and consistent results.
- In general, the results are about the same as in Arunas' PCA analyses with manually set classification boundaries.
- PCA dimension of n = 3 seems okey. Since there is still bias, it may be that there is some room for improvement with more complex neural networks.
- Results will be compared / combined with GC when this information becomes available.
Also, a Bayesian linear regression analysis of the PA/AA eNose calibration and Lithuanian cow data has been performed. Main conclusions have been:
- Bayesian linear regression to predict AA and PA concentration in air have been applied. Output is not only prediction, but also an estimate of the uncertainty.
- Leave-one-out and leave-block-out cross validation on the data set have been applied, to validate the predictability of the problem within this dataset.
Investigations have been made on the correlation of the joint sensor array signal to the identified pheromone compounds (AA/AC) in cow faeces, by using methods developed in WP1 and software developed in WP3.
RTD performer SMART developed (programmed in Matlab, tested, and debugged) software implementing the identified and designed methods and algorithms specified earlier:
- EM-based learning for partially labelled cow-data (Lithuanian cows);
- PCA-based learning for pre-processing of data;
- log-z score transform for GC data preparation;
- linear regression + Bayesian inference and PCA pre-processing for sensor array pheromone learning;
- logistic regression and Bayesian inference and hypersphere - baseline pre-processing for sensor array estrus learning.
Auxiliary tools have been developed (programmed):
- data import modules for sensor data, as well as GC data provided by KTH;
- routines for a numerical interpretation of text descriptions of estrus data (provided by GD);
- scripts for evaluating (leave-one-cow out training and testing shells) of the different methods and algorithms prototypes;
- routines for data visualisation, both for developing, reporting, and output visualisation for end-user purposes.
A graphical user-interface (GUI) for operating the BOVINOSE has been designed and developed. This user-interface includes:
- Windows based graphical user interface for controlling the BOVINOSE sensor system;
- mouse based adaptive user controls for data handling and updating;
- Excel export facility;
- data and prediction output visualisation.
Prototyping
The results on the sensor set has been further developed into the design and development of the BOVINOSE prototype.
This has lead to an electronic modular system with definite processing routine optimised for the detection and featuring of the odours specific to estrus in dairy cows. It consists of:
1. modules of detectors, control and measurement, the gas injection, subsystem for odour sampling and delivery;
2. interface electronics, signal pre-processing and tools;
3. computer and data analysis software.
The headspace air above the real faeces samples is the source of odour for the classification of cow-state by smell. A gas adsorbing material (known as the SPME fibre) is used for pre-concentration the key VOCs from a container and transportation to the sensor chamber. Emitted VOCs from the SPME fibre are carried by a flow and produce specific gaseous surrounding in the vicinity of gas sensors in narrow space of the test chamber.
A special holder is developed for manual extraction of VOCs from the container with faeces samples. The holder is adapted to the sampling module of BOVINOSE box. If manually inserted in the injection module the construction of holder is acceptable for automatic delivery of VOCs into the sensor chamber.
An odour sampling module producing the headspace air with volatile compounds from the real faeces samples has been developed. The module stabilises temperature in the container and blows out the VOCs from it.
The module is adapted for manual collection of VOCs by a special probe with a SPME fibre. The carrier gas flow and temperature are both controlled by the main control unit of the device.
BOVINOSE prototype MKII
The evaluation of the first version of the prototype of the BOVINOSE device has resulted in new functionalities and improvements:
1) the device is smaller and it contains a completely integrated sampling module. Therefore, the needle is no longer required to be inserted separately;
2) the device is controlled by a new version program;
3) since the sampling module is integrated, the complete duration of the test from the start to the end includes also sampling of smell, and after that, the measurement starts automatically. No additional actions are required.
The second version prototype of the BOVINOSE device is a table-top instrument with dimensions 245 x 290 x 140 mm3. The table-top version prototype instrument was produced according to the requirements accepted for the demonstration and field trials stage in this project.
The single box instrument includes several modules with individual original arrangement and differing by functions. These modules are as follows:
1. module of sensors;
2. module of control and measurement;
3. module of VOCs injection;
4. subsystem for VOCs sampling and delivery;
5. interface electronics and signal pre-processing subsystem.
Gas sensor module with eight MO sensors, sampling-injection subsystem and analogical signal adapter
The module of sensors is based on an array of eight metal oxide gas sensors. Individual sensor is assembled within the standard transistor can package 'TO-5' case style (simply metal can) with diameter 9 mm and height 4.3 mm. The sensors are mounted within the special exposure section in the BOVINOSE device.
The analogical signal adapter modules are developed and adapted for the high resistance sensors in this module (interface circuitry). The individual circuitry adjusts high resistance of sensors with resistance of measurement channel. In addition, the signal adapters provide optimum voltage level on the ADC input.
The chamber is made of a metal slab with axial channel with diameter 8 mm. Two sensors are inserted in special holes on each side of the chamber with rectangular cross-section. On each sensor an individual adapter is fixed. The adapters are required especially for the high resistance sensors and can be individually tuned for the certain sensor. The resistance of the sensors can be from about 10 kOhm to about 200 MOhms. The sensors fit in the chamber tightly and it is remaining practically negligible dead space volume inside. The chamber is connected to the air flow tubing and with the sampling-injection module.
The VOCs sampling-injecting module
The module is designed to automatically handle the SPME fibre during the measurements. It takes control of the fibre after it is inserted in special load-lock cavity. The module consists of the guiding-protecting frame, driving mechanics, electrical motor and electrical wiring. The sub-system is adapted to handle a fibre holder that is specially designed for this prototype.
The novel sampling-injection module is totally integrated with the sensor module (sensor chamber) in the second version prototype of the BOVINOSE device. In this version, the SPME syringe remains inside the module all the time during the entire cycle of sampling, injection, measurement and cleaning procedures.
The VOCs are collected by the SPME fibre from the faeces samples during a three-step procedure. The syringe is moved, stopped and exposed to the VOCs after the start of the special programme. After collecting the VOCs the syringe is extracted from the exposure section. In the step 5, the measurement phase begins. No operator actions are required in this step.
The injection of VOCs is accomplished in two steps during which the syringe is moved forward and the core with adsorbing material is introduced into the evaporation section through which the carrying gas is flowing. The flow of carrying gas transports the VOCs from the adsorbing core to the sensor module.
After the measurement, the syringe is extracted from the evaporation section, the system is cleaned by the flow of air and the syringe is returned to the initial position. The device is ready for new measurement cycle and, consequently, the step 9 actually corresponds with the step 1 of the next measurement cycle. An operator has not manipulate with the syringe between the sequential cycles of the measurements.
In addition, the carrying gas flow system is equipped with special gas filter eliminating contaminants from the surrounding air and producing the carrying flow of clean air within the tubing in the device. Since the components of the gas flow and control subsystem are adapted to the sensors, these components, the sampling-injection module and the sensor array mainly define the dimensions of the BOVINOSE device.
Clean air supply sub-system
The clean air supply system is constructed for fresh air supply into the sensor chamber aiming to:
(i) produce a synthetic atmosphere for functioning of sensors;
(ii) to carry the target VOC from the SPME fibre to the sensors; and
(iii) to blow out the contaminating gases from the chamber.
The clean air supply sub-system is based on tubing with diameter 6 mm and the standard Swagelock connections. The sub-system includes a mini-compressor, an air filter, an adapting vent, input and output.
Data acquisition electronics with signal converters
The electronics is assembled on the main board of the device. The electronics is adapted for sampling up to eight sensors using individual analogical signal channel. The signals are converted into 16 bit digital output that can be transferred immediately to a PC.
The main electronic board module consists of interface electronics for sensors connection to the board, and elements of such subsystems are:
1. temperature stabilisation
2. AFS
3. SPME interface
4. sensors voltage source
5. solenoid interface
6. sound indicator.
The eight-channel sensor commutator is developed on base of precision eight-channel analogue multiplexer of MAX-308 series and operation amplifier of AD-8672. The positive-polarity sensor voltage source is developed on the basis of adjustable voltage regulators of LM-317 series. The data acquisition electronics is required for sampling the sensor signals and transferring the signals to the PC interface. The ADC/DAC module is based on 16-bit DAC-microchip with parallel digital input of AD-669 series. Operation amplifier's input of AD8671 series (element DA1) is an analogical input of module. Output of operation amplifier AD8671 series (element DA2) is an analogical output of module. The air flow force depends on micro compressors electrical motor voltage level. This level is adjusted by operation amplifier AD-8672 and output transistor IRF-630.
Control electronics
The control electronics performs control functions for the SPME module, air flow sub-system and gas sensor heating. The control electronics also provides an access for tuning of the variables in the device. The electronic components of the control sub-system are placed on the main electronics board and integrated in the main subsystems of the main electronic board. Output indicators of the control electronics subsystem (LEDs) are placed on the front panel of the device.
PC interface
A PC has to be connected to the BOVINOSE device through the RS-232 standard interface for the serial binary single-ended data and control signals connecting circuitry. A converter RS-232/USB is used for connecting the device to a PC with the special software package for the system control and data analysis.
The PC interface is designed for connecting the device to a PC. The interface consists of the electronic circuitry, connector and special software converter - a PC system driver. The PC interface is designed on the basis of the RS232 standard and physically uses an USB connection on a PC.
Power supply module
The power supply module produces stabilised voltage for supplying to all the electrical circuit and components in the prototype. The power should be optimised depending on the final design of the device.
The power supply module consists of:
1. power transformer - 1 pcs.
2. voltage rectifier - 4 pcs.
3. voltage stabilisers PCB - 1 pcs.
The power transformer is developed individually.
Software package
The software package performs an external control of the device and determines the variables in the electronic box. The software also provides a simple user's interface for device control, inspection of the performance and the data obtained from the device. The software is integrable within the data pre-processing and the data analysis package. For this special user's interface was developed by partner SMART (SMART Research BV, the Netherlands).
Main conclusions
The construction and the design of the second version prototype was optimised regarding the laboratory trials and the field tests. The construction can still be optimised making the commercial device more compact and handy. The improvements of the design and functioning can be made in the next stage of development after general approval for practical application is acquired. The main improvements in the second version prototype device are as follows:
1. Based on the information acquired from the field trials, the second version prototype of the BOVINOSE device was developed as a portable table-top instrument with totally automatic sampling-injection subsystem, data acquisition and processing. This second version prototype device is also adapted for the pheromone detection in the faeces samples. The detection limits for the pheromones are less than 1 ppm in air.
2. Part of drawbacks of the first version prototype device was solved by developing the second version of the prototype device. Revealed during the field trials, the weak points of the first version prototype device, namely user's interface, semi-automatic sampling system, manual handling of the SPME syringe and portability, were practically eliminated in the second version prototype of the BOVINOSE device developed and manufactured for the very last period of the BOVINOSE project.
3. Duration of separate test cycle (about 5 min) is recognised being too long for practical monitoring of large drove of cows in the farm. A few approaches were proposed for reduction of the measurement time in the commercial version of the device. It is proposed that additional technical research can provide the most effective way to reduce of the measurement time by any:
(i) automatic tray for handling of sample containers;
(ii) simultaneous testing of the samples in multichannel device;
(iii) development of special air socking and VOCs sampling module.
4. It is also recognised that some components of electronics and the RS-232 interface can be replaced by the more modern components. Such replacement should succeed the prototype device with reduced dimensions, improved compatibility with PC, increased speed of data acquisition and transfer and extended possibilities for additional functions in future.
5. For the future development it is proposed being attractive for practical application a new generation of multi-purpose and multi-parameter robotic system combining diverse methods of characterisation and acceptable for complex and continuous monitoring of cow's health and productivity.
Software activities
SMART has led the design, development, and implementation of:
- integration of data import, learning and prediction software and test scripts (developed under WP3) for prototype performance optimisation;
- integration of data import, prediction software and GUI (developed under WP4) with Protech's data acquisition software.
Testing and validation
Achievements on laboratory test report
The purpose of this task has been to prepare a test protocol for the usefulness of the electronic nose in comparison with traditional techniques. Main objectives of the task have been to:
- determine the moment of estrus;
- determine the level of pheromone emission using gas chromatography during the oestrus cycle;
- determine the level of pheromone emission using the electronic nose.
Background of the test
In order to select the most suitable sensor for sex pheromone detection the source of highest emission rates had to be determined. To that end a preliminary experiment was conducted, during which it was established that pheromones can be reliably detected in faeces and urine. However urine was not used in the laboratory and field tests of the e-nose. Two reasons led to such decision:
- in preliminary experiments a clear difference between the results obtained using urine and those using faeces was detected;
- collection of urine samples under field conditions is so complicated, that it was deemed too impractical. Faeces samples, on the other hand, can be collected easily.
Since faeces are easy to collect at regular intervals it was decided to use it as a main source to determine the excretion of sex pheromones during the estrus cycle.
Furthermore, besides the main objective (detection of estrus) an important added value of the measurement of sex pheromones is the ability to determine whether cows are still in estrus after insemination, enabling accurate pregnancy state assessment. Having this in mind, the animals have been monitored after insemination as well.
How it was done
In order to perform laboratory testing of the device, samples of cow faeces were collected and estrus status of the cows was recorded in two farms located in Lithuania and Netherlands. Seven and four samples were collected respectively in special plastic vessels accepted as the standard for the project experiments in the project meeting in Deventer in November. Faeces were collected before, during and after ovulation period established by standard technique based on behavioural and rectal reactions evaluated by experienced veterinarian as well as on application of ultrasound technique. After the preparation of individual collections of samples for the each cow, the samples were frozen at -28 °C and kept at the constant temperature. One of the sets was delivered to the KTH (Sweden) for the analysis by gas chromatography method. Another set of samples was delivered to the Sensors laboratory (partner Protech, Vilnius).
The gas sensing and sampling components were tested in experiments with these faeces samples. Two series of the samples were used in the volatile compounds tests during which the volatile compounds from the samples were injected into the sensor chamber and the response of sensors was recorded.
Results
To sum up the work done under the task 6.2 comparison of cow status evaluation results obtained by standard techniques and that of electronic nose was deemed satisfactory, as the results coincided in more than 70 % of the samples.
Regarding the detection rate - it must be noted that in large dairy farms precise estrus determination is rather complicated and often inaccurate, resulting in low insemination success rates due to untypical heat course in cows, misbalanced by specific handling and unnatural environment (created in order to increase milk production capacity). Such conditions decrease the percentage of accurate ovulation detections and may negatively affect prognosis of timeframe suitable for insemination by both means (standard and electronic nose). Under such conditions 70 % detection rate is considered to be a very good output.
It can be concluded that the characteristics of the BOVINOSE device and the accuracy of measurement is sufficient to obtain a set of features acceptable for discrimination between the separated faeces samples by the odour detection and pre-processing of output signals.
Software related achievements
Testing and validation of various methods, algorithms and prototype for both pheromone and estrus in prediction based on software developed under WP3 and WP5.
1. Laboratory tests: The system has been tested extensively on different batches of laboratory data (i.e. cow faeces taken from field measured in laboratory, comparing with GC data, farmer data)
2. Field tests: The tests by GC have been supported.
Field evaluations
The e-nose prototype was evaluated during a field trial on a dairy farm in the Netherlands. The device was tested for its capacity to predict estrus under field conditions and for its user friendliness. The device was also evaluated against criteria that were defined in the requirements from the user viewpoint.
The results were promising. The device was in compliance with the acceptable criteria. Many of the ideal criteria were not fulfilled; however, as a result of the field test several improvements were made. The software was further developed as described in this report. Furthermore the device was made smaller and the needle was integrated in the device. These improvements greatly enhanced the user friendliness.
Several functionalities were tested under field conditions. Frozen samples were compared with fresh samples and it was shown that although both samples could be used the results could be different. Also prolonged freezing and thawing results in unreliable measurements.
Predictions for field trials cannot be straightforwardly based on laboratory results. More field data are required and new training is needed to analyse the difference in the PCA plots.
This has lead to the following major conclusions:
1. The prototype that was tested under farm conditions proved to be able to measure volatiles from faces samples. The measurements were only done on a limited number of animals and they were used to make recommendations for the development of a second prototype.
2. The results showed that it is advisable to treat samples always in the same way, either frozen or fresh. Once samples have been thawed after freezing they should not be frozen again.
3. Much time was spent on improving the software. The new software is not yet fully tested but it is a considerable improvement. This software can be the basis for further development.
4. There was good correlation between heat and pheromone peaks in the laboratory setting. The analyses showed that silent heat resulted in pheromone peaks. After insemination some cows showed pheromone peaks which make it difficult to use the pheromones as an indicator for pregnancy. In the field setting, more data are needed to confirm this result.
5. The prototype device performed under field conditions equally well as under laboratory conditions, however predictions for field trials cannot be straightforwardly based on laboratory results.
Evaluation and recommendations
Technical performance assessment
Technical aspects of design, functioning, manufacturing and application were investigated for the first and second versions prototype devices of pheromone detection in the faeces samples of diary cows. The findings of the laboratory tests and the field trials were analysed and readiness of the device for practical application was evaluated. It was recognised that some technical shortages existed in the design and functioning of the device and solutions for major issues were provided.
The laboratory tests and field trials were carried out for the first version of BOVINOSE device. The laboratory tests were accomplished under well controlled working conditions. The field trials were performed for the same prototype device at a dairy farm under realistic conditions. The results of both tests were used for evaluation of technical aspects in functioning of the device.
Major conclusions:
1. Two versions of the BOVINOSE device prototype are developed in Vilnius (Lithuania) and probed in the laboratory tests in Vilnius (Lithuania) and in the field tests in the G. Aveskamp's farm (the Netherlands) and the Animal Health Service Center (Deventer, the Netherlands) during the project period.
2. The first version prototype of the BOVINOSE device is a table-top instrument with semi-automatic sampling-injection subsystem acceptable for detection of pheromone components, namely propionic acid and acetic acid, at amounts below 1 ppm in air. The prototype device is adapted for detection of pheromones in the faeces samples of cows.
3. Based on the information acquired from the field trials, the second version prototype of the BOVINOSE device was developed as a portable table-top instrument with totally automatic sampling-injection subsystem, data acquisition and processing. This second version prototype device is also adapted for the pheromone detection in the faeces samples. The detection limits for the pheromones are less than 1 ppm in air.
4. The long term stability tests were performed at various periods over about 10 months interval during the BOVINOSE project. In each experiment, the same faeces samples were probed with the device under typical laboratory room conditions. It can be concluded that the response outputs of the BOVINOSE device are repeatedly obtained for similar samples (pure VOCs and real faeces samples) over practically entire exploitation period within the frame of the BOVINOSE project. It is demonstrated by these long term tests that the long term stability, reproducibility and repeatability of functioning of the BOVINOSE device are acceptable for practical application.
5. The influences of external factors, namely surrounding temperature (T), relative humidity of surrounding air (RH), presence of NH3 in the atmosphere (PNH3), were analysed by the laboratory tests with a special container. Specific working conditions were intentionally produced for the BOVINOSE device within this space limiting container. It was proved that the tested factors produce negligible influence on the output signals of the BOVINOSE device. It was also demonstrated that the unexpected deviations between the output signals occur for some faeces samples that cannot be explained only by the technical aspects of device functioning.
6. Part of drawbacks of the first version prototype device was solved by developing the second version of the prototype device. Revealed during the field trials, the weak points of the first version prototype device, namely user's interface, semi-automatic sampling system, manual handling of the SPME syringe and portability, were practically eliminated in the second version prototype of the BOVINOSE device developed and manufactured for the very last period of the BOVINOSE project.
7. Duration of separate test cycle (about 5 min) is recognised being too long for practical monitoring of large drove of cows in the farm. A few approaches were proposed for reduction of the measurement time in the commercial version of the device. It is proposed that additional technical research can provide the most effective way to reduce of the measurement time by any:
(i) automatic tray for handling of sample containers;
(ii) simultaneous testing of the samples in multichannel device;
(iii) development of special air socking and VOCs sampling module.
8. It is also recognised that some components of electronics and the RS-232 interface can be replaced by the more modern components. Such replacement should succeed the prototype device with reduced dimensions, improved compatibility with PC, increased speed of data acquisition and transfer and extended possibilities for additional functions in future.
9. For the future development it is proposed being attractive for practical application a new generation of multi-purpose and multi-parameter robotic system combining diverse methods of characterisation and acceptable for complex and continuous monitoring of cow's health and productivity.
10. Model predictions of pheromone compounds AA and PA based on BOVINOSE measurements on faeces samples from diary cows correlate strongly with results of direct GC measurements of those compounds in these samples.
11. Model predictions of estrus based on BOVINOSE measurements on faeces samples from diary cows correlate in general strongly with observations of estrus by human experts.
12. A disadvantage of the current models is that (1) the pheromone prediction basically only makes sense if a whole series is measured (to reveal the pattern of the series - since individual predictions make no sense due to scaling), and (2) the estrus prediction needs to be calibrated per cow by initial data from samples taken when the cow is not in heat, (nor pregnant).
Numerical prediction results of laboratory tests as well as of field trials have been extensively analyzed and evaluated, and discussed and reported. Following strong and weak points have been identified:
Strong points:
- the strong, robust correlation of eNose pheromone predictions with the actual pheromone data in the laboratory tests;
- the quite good correlation of eNose estrus predictions with the actual estrus data according to human experts.
Weak points for further research and improvements:
- the need of a reliable baseline determination. The required measurements needed for this baseline determination; the sensitivity of predictions to an erroneous baseline determination;
- deviations of field trial data compared to laboratory data (however, more field trials are needed for definite conclusions).
Economic assessment
Cost price calculations have been made and analysed for the latest prototype. The outcome of these activities has been described in deliverable D7.2.
Value analysis techniques were used to identify potential for cost reductions of the first prototype while maintaining functionality, and to guide further product development and improvements. Costs are meaningless without the context of revenues generated (profits), so, to support future exploitation, a customer value proposition has been included, in this case based on payback time and Internal Rate of Return calculations. These have been made for several dairy herd sizes.
The analysis reveals that both for the customer (mainly the farmer, but also veterinary doctors and AI stations) and for the device manufacturer the device is highly attractive in terms of annual returns.
With a modest improvement of the conception rate from 50 % to 54 % per cycle, the payback time for the farmer is about a year for a herd size of 30 dairy cows. For a relatively small herd size of 15 dairy cows the payback time is with about 26 months still very acceptable.
For the device manufacturer, the break-even point for earning back the investments in product R&D and market development is reached at selling 2152 units, less than about 1 % of the anticipated market size of 250 000 units. This implies a great market outlook. Assuming a gradual ('S-shape') market diffusion with 10 % market penetration after 5 years, and 12 % anticipated profits, the internal rate of return has been calculated at 66 % (5 years) and 78 % (10 years), respectively. These figures make it attractive for the BOVINOSE partners from a cost/benefits viewpoint to pursue further R&D and market development.
IMV and NIFA have conducted a market survey amongst French, Dutch and German farmers. All results have been listed in D7.2. For the cost price analysis the relevant data sections have been used as inputs. These surveys were also useful input for the deliverable D8.4 on the proposed BOVINOSE marketing strategy.
Recommendations for further research
In cows, the e-nose detects an estrus by measuring volatile chemical pheromone components which quantitatively correlates with ovulation time and in that way allows determination of optimal insemination time. The fact that domestic animals like horses, sheep and goats use olfactory signals to obtain information about reproduction status of partner, indicates that electronic nose could be applied for detection of ovulation time and consequently for determination of optimal insemination time. In addition, e-nose could be used in forestry and agriculture when rapid, sensitive and reliable method of volatile chemical marker(s) detection is needed.
Major conclusion
Electronic nose could be used in farming, forestry and agriculture when rapid, sensitive and reliable method of certain odour detection is essential for resolving tasks.
Dissemination and preparation for future exploitation
PUDF
The plan for use and dissemination of foreground (PUDF) has been completed in two stages. This document integrates both stages of the development on the PUDF.
A policy of wide dissemination of project results will be pursued in particular focused on potential end users of the project results, i.e. dairy farmers and veterinary / reproduction specialists and agricultural consultants. This second group plays an important role as they have usually strong working relationships with the farmers, and their opinion bears a lot of weight to the farmer. Therefore, scientifically sound proof of the working principle will in particularly disseminated to veterinary doctors and agricultural consultants who are active in the dairy value chain.
Peer-reviewed articles
Preparations have been made for four peer-reviewed texts for publication in scientific journals or as stand-alone publications in line with editorial and other requirements as requested by the publisher. These publications will contribute to disseminate research findings.
During the final six project months, the RTD performers have selected a number of potential fields of publication based on the results of the BOVINOSE project. RTDs KTH and SMART have taken the lead in preparing these. SMEs have been formally asked for approval.
Four scientific publications have been selected for publication. One has actually been pressed and three have been prepared for future publication.
1. W. Wiegerinck, A. Setkus, V. Buda, A.-K. Borg- Karlsson, R. Mozuraitis, A. De Gee, 'BOVINOSE: Pheromone-based sensor system for detecting estrus in dairy cows', 2011, Physics Procedia, in press.
2. Mozuraitis, R. et al., 'Sex pheromone dynamics during ovulation period in estrus-induced cows', 2011, in preparation.
3. Mozuraitis, R. et al., 'Sex pheromone dynamics during estrus cycle in dairy cows', 2011, in preparation.
4. Wiegerinck, W., et al., 'Rapid identification of dairy cows sex pheromone based on gas sensor featuring', 2011, in preparation.
One article has already been published in Physics Procedia 00 (2011) 1-2.
Three have been prepared for publication of which one article abstract has been used in the conference dealing with chemical signals in vertebrates 2011.
Marketing strategy report
Part of this task has been the research on specific consumer demands. Therefore, a dedicated dairy farmers questionnaire has been defined and developed. This has been used in interviewing dairy farmers and vetenarian doctors. The outcomes have been assembled, analysed and worked out into a description of the BOVINOSE market strategy as a basis for a successful market introduction. Surveys have been conducted amongst Dutch, German and French dairy farmers (mostly SMEs).
Market drivers have been identified that impact the future exploitation of BOVINOSE, in order to define appropriate measures integrated in a coherent and successful marketing strategy.
The final technology business plan has been identified and defined, describing the market, the product, the business, the marketing strategy, manufacturing, forecast of sales, cash flow, and breakeven, management and control of the business, as well as a further required financing.
Potential impact:
All beneficiaries will receive generated royalties pro rata the input they provided to the project (costs basis). At month 12 it is assumed that the project budget will be followed as a. This will be adjusted after the end of the project when actual costs are known.
The planning and the methodology of the exploitation for all five SME beneficiaries is presented below and will be refined along with the project.
PM
Measuring and diagnostic instrument manufacturer
- Define manufacturing plan for assembling BOVINOSE end-products, including machinery requirements, required finances, agreements with suppliers, quality control, personnel, shipment of goods etc.
- BOVINOSE system assembly.
Distribution / shipment of assembled end products: 21.27 %
COMPLEX
Electronic device design engineering and component manufacturer
- Define manufacturing plan for electronic components including machinery requirements, required finances, agreements with suppliers, quality control, personnel, etc.
Electronic component manufacturer (supplier to PM): 12.95 %
NIFA
AI product provider
- Assist in defining the BOVINOSE marketing strategy plan, including price setting, costs breakdown, price performance analysis, product leaflets / datasheets, press releases, promotion campaign, sales training
Area sales organisation Benelux, Germany, and USA: 20.21 %
IMV
AI product developer and manufacturer
- Coordinate BOVINOSE business plan
- Exploitation manager
- Coordination of defining the BOVINOSE marketing strategy plan, including price setting, costs breakdown, price performance analysis, product leaflets / datasheets, press releases, promotion campaign, sales training.
- Main sales organisation (EU and non-EU)
Exploitation manager: 34.23 %
AVESKAMP
Dairy farmer (end user)
- Field tests site for various stages of prototype sensors
- Test bed for component and system optimisation both during the project as during the first years of commercialisation after the project: 11.35 %
Exploitable foreground (overview)
Following is an overview of exploitable results generated by BOVINOSE.
As the final prototype is not fully finished at month 21, the partners could not make any decision on potential patents. At the moment, it looks like no patents will be applied for.
The project website is http://www.bovinose.eu
The BOVINOSE project aims to develop a novel type of 'electronic nose' to detect estrus in a dairy cow, and thus to determine the optimal timing of artificial insemination. The physical principle is based on detection of sex pheromones that are secreted by the cow, exclusively during estrus. These pheromones are the natural olfactory signal for the bull that the cow is in heat. The system will be an easy-to-use device that enables a dairy farmer to accurately determine a cow's ovulation timing. To this end, the device will be equipped with sensors, able to detect the specific bovine sex pheromones
This technology will help the dairy farmers in the EU, the vast majority being micro-enterprises run as family businesses. It is an opportunity for a number of SMEs that are active in supplying the dairy farmers with technology solutions. Simply put, in essence, a bull's nose will be mimicked.
Partners in the BOVINOSE consortium have been selected with a clear added value and complementarity in mind.
SMEs have teamed up from both the supply chain (manufacturers of electronic equipment and control systems, as well as AI solution providers), and from the end user perspective (dairy farmer).
The expertise of pheromone research, both on fundamental issues and on chemical characterisation, as well as application research on sensor development and signal processing, is incorporated in the project consortium. Veterinary know-how is brought in, particularly on the cow's estrus cycle and recognition of behavioural signs typical for cows in estrus.
Research institutes, represented by scientists and engineers of different disciplines support SMEs to develop and manufacture a novel and easy-to-use device that enables a dairy farmer to accurately determine a cow's ovulation timing.
The info-chemicals that make up the sex pheromones in cows have been extensively investigated using natural and induced samples from faeces, urine, and saliva from various dairy farms. Faeces were determined as most suitable source of sex pheromone based on the highest responses of bulls, largest emission rates of pheromone components, user-friendliness, and application conditions compare to the urine samples. Assessments were made to assess fine scale dynamics of volatile pheromone components as well as for measurements by sensors. The acetic and propionic acids were investigated and qualified as sufficient indicators of reproductive status of cows and could be used to define optimal insemination time.
Required BOVINOSE system properties could be identified. A set of sensors for detection of these sex pheromones was researched, designed and developed, resulting in a functional prototype. This was based on a set of defined user and technical requirements, consisting of a probe, an array of sensors and a dedicated, integrated software system for control and end-user interaction.
Extensive field trials closely accompanied by veterinarian experts have been run at a dairy farm, both on the info-chemicals as on the fitness for use for the anticipated end-user by performing a technical evaluation. The prototype that was tested under farm conditions proved to be able to measure volatiles from faces samples. There was good correlation between heat and pheromone peaks in the laboratory setting. The analyses showed that silent heat resulted in pheromone peaks. The prototype device performed under field conditions equally well as under laboratory conditions, however.
Project context and objectives:
The overarching goal of BOVINOSE is to develop a novel and easy-to-use device that enables a dairy farmer to accurately determine a cow's ovulation timing. To this end, the device will be equipped with sensors, able to detect the specific bovine sex pheromones. Simply put, in essence, a bull's nose will be mimicked.
The main scientific and technological objectives of BOVINOSE are the following:
- to identify the info-chemicals that make up the sex pheromones in cows that are uniquely released during estrus;
- to develop a set of sensors for detection of bovine sex pheromones;
- to develop an easy-to-use, yet reliable system for estrus determination in cows, based on detection of the sex pheromones;
- to develop and manufacture a functional prototype, consisting of a probe, an array of sensors and dedicated, integrated software for control and user-interaction.
The main objectives for the first twelve months of the BOVINOSE project have been dedicated to:
- defining and specifying the BOVINOSE system requirements as seen from an end-user, and the technical requirements for modules and components;
- researching sex pheromones for cows;
- design and develop prototypes of BOVINOSE sensors, both hardware and software parts.
The main objectives for the second (and final) part of the BOVINOSE project have been dedicated to:
- the development and building of a functional prototype (milestone Mi-3);
- software development programmed in Matlab, tested, and debugged;
- design and development of auxiliary tools and a graphical user-interface (GUI) for operating the BOVINOSE system;
- validation of the prototype system by extensive laboratory testing;
- full testing using a two-staged procedure of field trials at a dairy farm closely accompanied by veterinarian experts;
- full evaluation of the first version of the prototype of the BOVINOSE device resulting in new functionalities and improvements;
- final project evaluations and recommendations for further research, comprising a technical evaluation report, a cost price analysis, and further recommendations for further research;
- definition and finalisation of the required levels of dissemination and preparation for future exploitation.
All of these objectives have been fulfilled at the end of the project.
Project results:
System definition and specifications
The definition of systems requirements from a user viewpoint, in terms of performance, user-friendliness and application conditions; the specification of technical requirement for the system modules and components were all covered. This increased the reliability of the end system, what will increases the user satisfaction level.
System performance specifications
Major results on system performance specifications have been the definition of:
- basis for electronic pheromone identification;
- requirements of end user (both general considerations, functional and physical requirements);
- general structure of electronic pheromone detection and identification system.
Sensor specification
The technical requirements of the sensor have been specified, in terms of standard operating conditions, environmental conditions, sensitivity characteristics, and electro-mechanical characteristics. Typical parameters include power consumption, selectivity, accuracy, repeatability, drift, hysteresis, etc.
The metal oxide gas sensors are to be adapted for the BOVINOSE detector module. The sensors have to be adapted to function at normal indoor (and comparable) conditions at surrounding temperatures from about 10 °C to about 30 °C, relative humidity from about 30 % to about 80 %, oxygen rich atmosphere (not less than about 18 %). The sensors have to be fixed within the standard transistor container TO5. The sensitivity of the sensors should be not less than about 1 - 2 ppm of pure volatile compounds defined as pheromone components. The power consumption should not exceed about 500 mW per sensor and have to be used with a power source adapted for standard electricity supply.
The selectivity of the sensors has to be sufficient for the classification of the odours. The drift of the sensor response have to be less than about 10 % per continuous working cycle.
Considering the parameters of the sensors and possibilities to modify the characteristics of MO sensors it is reasonable to define the array of the BOVINOSE electronic odour detection system consisting of 8 MO gas sensors.
Software definition document
The BOVINOSE electronic pheromone detection system is to be equipped with the software packages that:
(i) allows control the hardware;
(ii) implements pre-processing of the sensor signals and creates a database of the output features;
(iii) performs analysis of the features; and
(iv) provides a user interface.
The technical requirements for these software packages data processing, data storage and control system have been specified.
Sampling module specifications
The odour sampling module is a peripheral tool of the BOVINOSE pheromone detection system that is optimised for sampling of volatile components specific to estrus in dairy cows. The sampling module will be modified and optimised for usage in farms in the final stage of the project. The second version of the sampling unit will be integrated within the detecting device.
The sampling unit is to be used with a standard container for collecting of faeces samples and adapted for the standard SPME fibre. The unit consists of the following components:
- Thermostat for stabilisation of temperature in the sample container during the SPME extraction of the volatile compounds from the headspace air. The working temperature of the thermostat has to be optimised during the laboratory tests.
- The headspace chamber with a holder for the special container of faeces samples and an adapter for the standard manual SPME holder, the volume of the headspace air is planned to be about 50 % of the sample volume but can be modified aiming to optimise the sampling system during the laboratory tests. The headspace chamber is adjusted to fit within the thermostat.
- (Optional) special sampling-cutting tool use as the faeces amount meter-container. The amount meter-container is to be included into the set of the sampling module if the variation in the featuring of the specific volatile compounds during the laboratory tests will be too high for reliable identification of the estrus in diary cows. Depending on the testing routine, the amount meter-container can be used instead of the standard container for sampling of faeces in the sampling unit.
- The materials of the sampling module have to be tested for the emissions of the interfering gases. The materials have to be optimised with respect to these emissions and the target volatile components.
- The methodology of usage of the sampling unit will be user-friendly and adjusted for the best effectiveness of the BOVINOSE device.
- The first version sampling unit will be used during the eNose test period in Vilnius and for the first stage reporting.
The first-stage prototype of the sampling module is a separate unit acceptable for use in the laboratory / room conditions. The module will be integrated within the complete BOVINOSE system in the second stage of development after the prove of the system acceptability for identification of the estrus in diary cows will be obtained and the project partners will agree to continue the project.
Specifications for the integrated system
BOVINOSE shall consist of the three main parts, namely the hardware, the software and the methodology for using the complete system. In the hardware, the key-part is the gas sensors and array of the sensors that determine reliability of the information acquired by the system. In the software, the key-part is the feature extraction routine and the database of the features that defines the most specific information about single test and makes it possible to compare this information with the list of the references. In the methodology, the key-part is the sampling of the right VOCs that makes it possible to obtain the advanced selectivity and sensitivity.
Pheromone research
Collection of biological samples
In total 140 samples of faeces, 105 samples of urine and 85 samples of saliva were collected in Lithuania from harmone induced ovulation in cows for assessment of bulls' reaction, measurements of release rates of sex pheromone components, to assess fine scale dynamics of volatile pheromone components as well as for measurements by sensors.
120 faeces samples were obtained from Holland from cows were under natural cycle. Samples were used for determination of correlation between amounts of the pheromone components and physiological cycle of cow as well as for measurements by sensors. The same amount of blood samples were collected and concentration of progesterone was determined.
Behavioural reaction of bulls to saliva, urine and faeces samples of cows in estrus
In total, reaction of 10 mature bulls was tested. The stimuli were applied on unestrus cow, as no reaction was recorded in absence of visual stimuli. Sixty percents of bulls showed mounting reaction to unestrus cow treated with sample of estrus faeces. Reaction to urine was less pronounced (35 %), and that to saliva was not noted at all (0 %).
Composition and usual amounts of bovine pheromone components
Acetic and propionic acids from three sex pheromone components known in cows were detected in headspace of faeces and urine by SPME equipped with carboxen-polydimethylsiloxane fibre. Both acids were present in pre-estrus, estrus and post-estrus samples of faeces and urine and reached their maximum amounts at some time before ovulation. Volatile sex pheromone components exhibit similar distribution pattern in samples of faeces and urine in respect to reproductive status of cows. During peak of release acetic and propionic acids occur at significantly higher amounts in faeces samples compare to those in urine. Faeces were determined as most suitable source of sex pheromone based on the highest responses of bulls, largest emission rates of pheromone components, user-friendliness, and application conditions compare to the urine samples.
Amounts of progesterone and volatile components of bovine pheromone during naturally occurring estrus cycles of cows
The monitoring of amounts of progesterone and volatile sex pheromone components, namely acetic and propionic acids revealed strong negative correlations between amounts of hormone and two volatile sex pheromone components in samples collected at the transition period from diestrus to estrus and at the first day of estrus which is a critical period for searching of optimal insemination time.
The acetic and propionic acids are sufficient indicators of reproductive status of cows and could be used to define optimal insemination time.
The fine scale dynamics of volatile pheromone components in estrus-induced cows
In the faeces samples of estrus-induced cows increase in amounts of acetic and propionic acids was registered around 24 hours before ovulation and after 10 - 12 hours dropped down to diestrus level. The dynamics of changes in amounts of acetic and propionic acids in the faeces samples obtained in pre-estrus and estrus periods occur in the similar manner in cows with naturally occurring and hormone induced estrus with clearly expressed peak of two sex pheromone components.
Sensor development
Prototype pheromone sensors
The prototype of sensors - a set of 24 sensors - was manufactured by the thin film technology and assembled within the standard TO5 transistor container. It consists of a sensitive layer with electrical contacts, mini-heater and thermometer. The surface of the sensitive layer was modified with catalyst non-homogeneous ultra thin film in order to improve sensitivity to pheromones. Optimal working temperature for pheromones detection was determined by test measurements. The prototype of sensors was tested and main characteristics were obtained.
A special sensor chamber is produced for the sensor module of the BOVINOSE electronic pheromone detection and recognition system.
Sensors response measurements
A description was made on VOC sampling methodology, calibration VOC and amount control, VOC injection module, measurement and control system, response measurement method, sensor parameters.
The sensors response was measured with the pure VOC, acetic acid and propionic acid that are defined being pheromones. It was found that qualitative evaluation of the sensor parameters can be performed by exposure to the headspace air of pure compounds.
The set of metal oxide gas sensors, namely SnO2-x with Pt, Au and Ru additives, were tested in the atmosphere with acetic acid, propionic acid and the VOCs from the faeces samples. It was found that qualitative evaluation of the sensor parameters can be performed by exposure to the headspace air of pure compounds. The calibration and correct evaluation of the sensor characteristics cannot be completed using non-calibrated sources of the target VOC. For this, the special sources with calibrated permeation rates were produced and the sensor calibration methodology based on these permeation tubes was developed.
Major conclusions were that the sensors in the array of BOVINOSE device have to be calibrated using reliable source of the pure target VOC because the real samples contain organic materials that are constantly changing due to processes related to the living organisms such as bacteria.
Improved prototype sensor
The sensors array was assembled and the tests with real samples from diary farm were performed. It was proved that the features obtained from the pheronose array allows to distinguish between the cows and condition of the cow during the testing time.
The sensor array was assembled from a set of various gas sensors produced in the Sensors Laboratory.
The sensors were calibrated with the pure pheromone compounds, namely acetic acid and propionic acid. The threshold amount of the pheromone components detectable by the sensors of the array is significantly below 0.1 ppm.
The real faeces samples obtained from diary farm were analysed with the pheronose. It was proved that the features obtained from the pheronose array allows to distinguish between the cows and condition of the cow during the testing time. Unfortunately, there are no possibilities to define the origin of the correlation because the biological description of the samples are absent at present stage of the investigation.
Major conclusions have been that the sensor array is acceptable for the featuring of the cow conditions. The pheronose device is acceptable for testing of the real samples in the laboratory conditions. Based on the results of present tests, some improvements can also be suggested for the next stage of the project.
Software development
A logistic regression / EM analysis of the Lithuania cow eNose data have been performed mainly by SMART and Protech. The model can be seen as a two-layer neural network. The training data consist of the data of the cows that are used for optimising the parameters (in the current model, these are the PCA and the regression parameters).
The main conclusions from this analysis have been the following:
- An EM algorithm for unsupervised learning has been developed using the prior knowledge that for the injected cows, there must be an estrus period within the given time period.
- The method is generally applicable for any probabilistic predictive model. Here, we used PCA analysis combined with logistic regression. (This could be interpreted as a particular two-layer neural network). However, other choices are possible.
- The EM algorithm, combined with a PCA -logistic regression model give reasonable and consistent results.
- In general, the results are about the same as in Arunas' PCA analyses with manually set classification boundaries.
- PCA dimension of n = 3 seems okey. Since there is still bias, it may be that there is some room for improvement with more complex neural networks.
- Results will be compared / combined with GC when this information becomes available.
Also, a Bayesian linear regression analysis of the PA/AA eNose calibration and Lithuanian cow data has been performed. Main conclusions have been:
- Bayesian linear regression to predict AA and PA concentration in air have been applied. Output is not only prediction, but also an estimate of the uncertainty.
- Leave-one-out and leave-block-out cross validation on the data set have been applied, to validate the predictability of the problem within this dataset.
Investigations have been made on the correlation of the joint sensor array signal to the identified pheromone compounds (AA/AC) in cow faeces, by using methods developed in WP1 and software developed in WP3.
RTD performer SMART developed (programmed in Matlab, tested, and debugged) software implementing the identified and designed methods and algorithms specified earlier:
- EM-based learning for partially labelled cow-data (Lithuanian cows);
- PCA-based learning for pre-processing of data;
- log-z score transform for GC data preparation;
- linear regression + Bayesian inference and PCA pre-processing for sensor array pheromone learning;
- logistic regression and Bayesian inference and hypersphere - baseline pre-processing for sensor array estrus learning.
Auxiliary tools have been developed (programmed):
- data import modules for sensor data, as well as GC data provided by KTH;
- routines for a numerical interpretation of text descriptions of estrus data (provided by GD);
- scripts for evaluating (leave-one-cow out training and testing shells) of the different methods and algorithms prototypes;
- routines for data visualisation, both for developing, reporting, and output visualisation for end-user purposes.
A graphical user-interface (GUI) for operating the BOVINOSE has been designed and developed. This user-interface includes:
- Windows based graphical user interface for controlling the BOVINOSE sensor system;
- mouse based adaptive user controls for data handling and updating;
- Excel export facility;
- data and prediction output visualisation.
Prototyping
The results on the sensor set has been further developed into the design and development of the BOVINOSE prototype.
This has lead to an electronic modular system with definite processing routine optimised for the detection and featuring of the odours specific to estrus in dairy cows. It consists of:
1. modules of detectors, control and measurement, the gas injection, subsystem for odour sampling and delivery;
2. interface electronics, signal pre-processing and tools;
3. computer and data analysis software.
The headspace air above the real faeces samples is the source of odour for the classification of cow-state by smell. A gas adsorbing material (known as the SPME fibre) is used for pre-concentration the key VOCs from a container and transportation to the sensor chamber. Emitted VOCs from the SPME fibre are carried by a flow and produce specific gaseous surrounding in the vicinity of gas sensors in narrow space of the test chamber.
A special holder is developed for manual extraction of VOCs from the container with faeces samples. The holder is adapted to the sampling module of BOVINOSE box. If manually inserted in the injection module the construction of holder is acceptable for automatic delivery of VOCs into the sensor chamber.
An odour sampling module producing the headspace air with volatile compounds from the real faeces samples has been developed. The module stabilises temperature in the container and blows out the VOCs from it.
The module is adapted for manual collection of VOCs by a special probe with a SPME fibre. The carrier gas flow and temperature are both controlled by the main control unit of the device.
BOVINOSE prototype MKII
The evaluation of the first version of the prototype of the BOVINOSE device has resulted in new functionalities and improvements:
1) the device is smaller and it contains a completely integrated sampling module. Therefore, the needle is no longer required to be inserted separately;
2) the device is controlled by a new version program;
3) since the sampling module is integrated, the complete duration of the test from the start to the end includes also sampling of smell, and after that, the measurement starts automatically. No additional actions are required.
The second version prototype of the BOVINOSE device is a table-top instrument with dimensions 245 x 290 x 140 mm3. The table-top version prototype instrument was produced according to the requirements accepted for the demonstration and field trials stage in this project.
The single box instrument includes several modules with individual original arrangement and differing by functions. These modules are as follows:
1. module of sensors;
2. module of control and measurement;
3. module of VOCs injection;
4. subsystem for VOCs sampling and delivery;
5. interface electronics and signal pre-processing subsystem.
Gas sensor module with eight MO sensors, sampling-injection subsystem and analogical signal adapter
The module of sensors is based on an array of eight metal oxide gas sensors. Individual sensor is assembled within the standard transistor can package 'TO-5' case style (simply metal can) with diameter 9 mm and height 4.3 mm. The sensors are mounted within the special exposure section in the BOVINOSE device.
The analogical signal adapter modules are developed and adapted for the high resistance sensors in this module (interface circuitry). The individual circuitry adjusts high resistance of sensors with resistance of measurement channel. In addition, the signal adapters provide optimum voltage level on the ADC input.
The chamber is made of a metal slab with axial channel with diameter 8 mm. Two sensors are inserted in special holes on each side of the chamber with rectangular cross-section. On each sensor an individual adapter is fixed. The adapters are required especially for the high resistance sensors and can be individually tuned for the certain sensor. The resistance of the sensors can be from about 10 kOhm to about 200 MOhms. The sensors fit in the chamber tightly and it is remaining practically negligible dead space volume inside. The chamber is connected to the air flow tubing and with the sampling-injection module.
The VOCs sampling-injecting module
The module is designed to automatically handle the SPME fibre during the measurements. It takes control of the fibre after it is inserted in special load-lock cavity. The module consists of the guiding-protecting frame, driving mechanics, electrical motor and electrical wiring. The sub-system is adapted to handle a fibre holder that is specially designed for this prototype.
The novel sampling-injection module is totally integrated with the sensor module (sensor chamber) in the second version prototype of the BOVINOSE device. In this version, the SPME syringe remains inside the module all the time during the entire cycle of sampling, injection, measurement and cleaning procedures.
The VOCs are collected by the SPME fibre from the faeces samples during a three-step procedure. The syringe is moved, stopped and exposed to the VOCs after the start of the special programme. After collecting the VOCs the syringe is extracted from the exposure section. In the step 5, the measurement phase begins. No operator actions are required in this step.
The injection of VOCs is accomplished in two steps during which the syringe is moved forward and the core with adsorbing material is introduced into the evaporation section through which the carrying gas is flowing. The flow of carrying gas transports the VOCs from the adsorbing core to the sensor module.
After the measurement, the syringe is extracted from the evaporation section, the system is cleaned by the flow of air and the syringe is returned to the initial position. The device is ready for new measurement cycle and, consequently, the step 9 actually corresponds with the step 1 of the next measurement cycle. An operator has not manipulate with the syringe between the sequential cycles of the measurements.
In addition, the carrying gas flow system is equipped with special gas filter eliminating contaminants from the surrounding air and producing the carrying flow of clean air within the tubing in the device. Since the components of the gas flow and control subsystem are adapted to the sensors, these components, the sampling-injection module and the sensor array mainly define the dimensions of the BOVINOSE device.
Clean air supply sub-system
The clean air supply system is constructed for fresh air supply into the sensor chamber aiming to:
(i) produce a synthetic atmosphere for functioning of sensors;
(ii) to carry the target VOC from the SPME fibre to the sensors; and
(iii) to blow out the contaminating gases from the chamber.
The clean air supply sub-system is based on tubing with diameter 6 mm and the standard Swagelock connections. The sub-system includes a mini-compressor, an air filter, an adapting vent, input and output.
Data acquisition electronics with signal converters
The electronics is assembled on the main board of the device. The electronics is adapted for sampling up to eight sensors using individual analogical signal channel. The signals are converted into 16 bit digital output that can be transferred immediately to a PC.
The main electronic board module consists of interface electronics for sensors connection to the board, and elements of such subsystems are:
1. temperature stabilisation
2. AFS
3. SPME interface
4. sensors voltage source
5. solenoid interface
6. sound indicator.
The eight-channel sensor commutator is developed on base of precision eight-channel analogue multiplexer of MAX-308 series and operation amplifier of AD-8672. The positive-polarity sensor voltage source is developed on the basis of adjustable voltage regulators of LM-317 series. The data acquisition electronics is required for sampling the sensor signals and transferring the signals to the PC interface. The ADC/DAC module is based on 16-bit DAC-microchip with parallel digital input of AD-669 series. Operation amplifier's input of AD8671 series (element DA1) is an analogical input of module. Output of operation amplifier AD8671 series (element DA2) is an analogical output of module. The air flow force depends on micro compressors electrical motor voltage level. This level is adjusted by operation amplifier AD-8672 and output transistor IRF-630.
Control electronics
The control electronics performs control functions for the SPME module, air flow sub-system and gas sensor heating. The control electronics also provides an access for tuning of the variables in the device. The electronic components of the control sub-system are placed on the main electronics board and integrated in the main subsystems of the main electronic board. Output indicators of the control electronics subsystem (LEDs) are placed on the front panel of the device.
PC interface
A PC has to be connected to the BOVINOSE device through the RS-232 standard interface for the serial binary single-ended data and control signals connecting circuitry. A converter RS-232/USB is used for connecting the device to a PC with the special software package for the system control and data analysis.
The PC interface is designed for connecting the device to a PC. The interface consists of the electronic circuitry, connector and special software converter - a PC system driver. The PC interface is designed on the basis of the RS232 standard and physically uses an USB connection on a PC.
Power supply module
The power supply module produces stabilised voltage for supplying to all the electrical circuit and components in the prototype. The power should be optimised depending on the final design of the device.
The power supply module consists of:
1. power transformer - 1 pcs.
2. voltage rectifier - 4 pcs.
3. voltage stabilisers PCB - 1 pcs.
The power transformer is developed individually.
Software package
The software package performs an external control of the device and determines the variables in the electronic box. The software also provides a simple user's interface for device control, inspection of the performance and the data obtained from the device. The software is integrable within the data pre-processing and the data analysis package. For this special user's interface was developed by partner SMART (SMART Research BV, the Netherlands).
Main conclusions
The construction and the design of the second version prototype was optimised regarding the laboratory trials and the field tests. The construction can still be optimised making the commercial device more compact and handy. The improvements of the design and functioning can be made in the next stage of development after general approval for practical application is acquired. The main improvements in the second version prototype device are as follows:
1. Based on the information acquired from the field trials, the second version prototype of the BOVINOSE device was developed as a portable table-top instrument with totally automatic sampling-injection subsystem, data acquisition and processing. This second version prototype device is also adapted for the pheromone detection in the faeces samples. The detection limits for the pheromones are less than 1 ppm in air.
2. Part of drawbacks of the first version prototype device was solved by developing the second version of the prototype device. Revealed during the field trials, the weak points of the first version prototype device, namely user's interface, semi-automatic sampling system, manual handling of the SPME syringe and portability, were practically eliminated in the second version prototype of the BOVINOSE device developed and manufactured for the very last period of the BOVINOSE project.
3. Duration of separate test cycle (about 5 min) is recognised being too long for practical monitoring of large drove of cows in the farm. A few approaches were proposed for reduction of the measurement time in the commercial version of the device. It is proposed that additional technical research can provide the most effective way to reduce of the measurement time by any:
(i) automatic tray for handling of sample containers;
(ii) simultaneous testing of the samples in multichannel device;
(iii) development of special air socking and VOCs sampling module.
4. It is also recognised that some components of electronics and the RS-232 interface can be replaced by the more modern components. Such replacement should succeed the prototype device with reduced dimensions, improved compatibility with PC, increased speed of data acquisition and transfer and extended possibilities for additional functions in future.
5. For the future development it is proposed being attractive for practical application a new generation of multi-purpose and multi-parameter robotic system combining diverse methods of characterisation and acceptable for complex and continuous monitoring of cow's health and productivity.
Software activities
SMART has led the design, development, and implementation of:
- integration of data import, learning and prediction software and test scripts (developed under WP3) for prototype performance optimisation;
- integration of data import, prediction software and GUI (developed under WP4) with Protech's data acquisition software.
Testing and validation
Achievements on laboratory test report
The purpose of this task has been to prepare a test protocol for the usefulness of the electronic nose in comparison with traditional techniques. Main objectives of the task have been to:
- determine the moment of estrus;
- determine the level of pheromone emission using gas chromatography during the oestrus cycle;
- determine the level of pheromone emission using the electronic nose.
Background of the test
In order to select the most suitable sensor for sex pheromone detection the source of highest emission rates had to be determined. To that end a preliminary experiment was conducted, during which it was established that pheromones can be reliably detected in faeces and urine. However urine was not used in the laboratory and field tests of the e-nose. Two reasons led to such decision:
- in preliminary experiments a clear difference between the results obtained using urine and those using faeces was detected;
- collection of urine samples under field conditions is so complicated, that it was deemed too impractical. Faeces samples, on the other hand, can be collected easily.
Since faeces are easy to collect at regular intervals it was decided to use it as a main source to determine the excretion of sex pheromones during the estrus cycle.
Furthermore, besides the main objective (detection of estrus) an important added value of the measurement of sex pheromones is the ability to determine whether cows are still in estrus after insemination, enabling accurate pregnancy state assessment. Having this in mind, the animals have been monitored after insemination as well.
How it was done
In order to perform laboratory testing of the device, samples of cow faeces were collected and estrus status of the cows was recorded in two farms located in Lithuania and Netherlands. Seven and four samples were collected respectively in special plastic vessels accepted as the standard for the project experiments in the project meeting in Deventer in November. Faeces were collected before, during and after ovulation period established by standard technique based on behavioural and rectal reactions evaluated by experienced veterinarian as well as on application of ultrasound technique. After the preparation of individual collections of samples for the each cow, the samples were frozen at -28 °C and kept at the constant temperature. One of the sets was delivered to the KTH (Sweden) for the analysis by gas chromatography method. Another set of samples was delivered to the Sensors laboratory (partner Protech, Vilnius).
The gas sensing and sampling components were tested in experiments with these faeces samples. Two series of the samples were used in the volatile compounds tests during which the volatile compounds from the samples were injected into the sensor chamber and the response of sensors was recorded.
Results
To sum up the work done under the task 6.2 comparison of cow status evaluation results obtained by standard techniques and that of electronic nose was deemed satisfactory, as the results coincided in more than 70 % of the samples.
Regarding the detection rate - it must be noted that in large dairy farms precise estrus determination is rather complicated and often inaccurate, resulting in low insemination success rates due to untypical heat course in cows, misbalanced by specific handling and unnatural environment (created in order to increase milk production capacity). Such conditions decrease the percentage of accurate ovulation detections and may negatively affect prognosis of timeframe suitable for insemination by both means (standard and electronic nose). Under such conditions 70 % detection rate is considered to be a very good output.
It can be concluded that the characteristics of the BOVINOSE device and the accuracy of measurement is sufficient to obtain a set of features acceptable for discrimination between the separated faeces samples by the odour detection and pre-processing of output signals.
Software related achievements
Testing and validation of various methods, algorithms and prototype for both pheromone and estrus in prediction based on software developed under WP3 and WP5.
1. Laboratory tests: The system has been tested extensively on different batches of laboratory data (i.e. cow faeces taken from field measured in laboratory, comparing with GC data, farmer data)
2. Field tests: The tests by GC have been supported.
Field evaluations
The e-nose prototype was evaluated during a field trial on a dairy farm in the Netherlands. The device was tested for its capacity to predict estrus under field conditions and for its user friendliness. The device was also evaluated against criteria that were defined in the requirements from the user viewpoint.
The results were promising. The device was in compliance with the acceptable criteria. Many of the ideal criteria were not fulfilled; however, as a result of the field test several improvements were made. The software was further developed as described in this report. Furthermore the device was made smaller and the needle was integrated in the device. These improvements greatly enhanced the user friendliness.
Several functionalities were tested under field conditions. Frozen samples were compared with fresh samples and it was shown that although both samples could be used the results could be different. Also prolonged freezing and thawing results in unreliable measurements.
Predictions for field trials cannot be straightforwardly based on laboratory results. More field data are required and new training is needed to analyse the difference in the PCA plots.
This has lead to the following major conclusions:
1. The prototype that was tested under farm conditions proved to be able to measure volatiles from faces samples. The measurements were only done on a limited number of animals and they were used to make recommendations for the development of a second prototype.
2. The results showed that it is advisable to treat samples always in the same way, either frozen or fresh. Once samples have been thawed after freezing they should not be frozen again.
3. Much time was spent on improving the software. The new software is not yet fully tested but it is a considerable improvement. This software can be the basis for further development.
4. There was good correlation between heat and pheromone peaks in the laboratory setting. The analyses showed that silent heat resulted in pheromone peaks. After insemination some cows showed pheromone peaks which make it difficult to use the pheromones as an indicator for pregnancy. In the field setting, more data are needed to confirm this result.
5. The prototype device performed under field conditions equally well as under laboratory conditions, however predictions for field trials cannot be straightforwardly based on laboratory results.
Evaluation and recommendations
Technical performance assessment
Technical aspects of design, functioning, manufacturing and application were investigated for the first and second versions prototype devices of pheromone detection in the faeces samples of diary cows. The findings of the laboratory tests and the field trials were analysed and readiness of the device for practical application was evaluated. It was recognised that some technical shortages existed in the design and functioning of the device and solutions for major issues were provided.
The laboratory tests and field trials were carried out for the first version of BOVINOSE device. The laboratory tests were accomplished under well controlled working conditions. The field trials were performed for the same prototype device at a dairy farm under realistic conditions. The results of both tests were used for evaluation of technical aspects in functioning of the device.
Major conclusions:
1. Two versions of the BOVINOSE device prototype are developed in Vilnius (Lithuania) and probed in the laboratory tests in Vilnius (Lithuania) and in the field tests in the G. Aveskamp's farm (the Netherlands) and the Animal Health Service Center (Deventer, the Netherlands) during the project period.
2. The first version prototype of the BOVINOSE device is a table-top instrument with semi-automatic sampling-injection subsystem acceptable for detection of pheromone components, namely propionic acid and acetic acid, at amounts below 1 ppm in air. The prototype device is adapted for detection of pheromones in the faeces samples of cows.
3. Based on the information acquired from the field trials, the second version prototype of the BOVINOSE device was developed as a portable table-top instrument with totally automatic sampling-injection subsystem, data acquisition and processing. This second version prototype device is also adapted for the pheromone detection in the faeces samples. The detection limits for the pheromones are less than 1 ppm in air.
4. The long term stability tests were performed at various periods over about 10 months interval during the BOVINOSE project. In each experiment, the same faeces samples were probed with the device under typical laboratory room conditions. It can be concluded that the response outputs of the BOVINOSE device are repeatedly obtained for similar samples (pure VOCs and real faeces samples) over practically entire exploitation period within the frame of the BOVINOSE project. It is demonstrated by these long term tests that the long term stability, reproducibility and repeatability of functioning of the BOVINOSE device are acceptable for practical application.
5. The influences of external factors, namely surrounding temperature (T), relative humidity of surrounding air (RH), presence of NH3 in the atmosphere (PNH3), were analysed by the laboratory tests with a special container. Specific working conditions were intentionally produced for the BOVINOSE device within this space limiting container. It was proved that the tested factors produce negligible influence on the output signals of the BOVINOSE device. It was also demonstrated that the unexpected deviations between the output signals occur for some faeces samples that cannot be explained only by the technical aspects of device functioning.
6. Part of drawbacks of the first version prototype device was solved by developing the second version of the prototype device. Revealed during the field trials, the weak points of the first version prototype device, namely user's interface, semi-automatic sampling system, manual handling of the SPME syringe and portability, were practically eliminated in the second version prototype of the BOVINOSE device developed and manufactured for the very last period of the BOVINOSE project.
7. Duration of separate test cycle (about 5 min) is recognised being too long for practical monitoring of large drove of cows in the farm. A few approaches were proposed for reduction of the measurement time in the commercial version of the device. It is proposed that additional technical research can provide the most effective way to reduce of the measurement time by any:
(i) automatic tray for handling of sample containers;
(ii) simultaneous testing of the samples in multichannel device;
(iii) development of special air socking and VOCs sampling module.
8. It is also recognised that some components of electronics and the RS-232 interface can be replaced by the more modern components. Such replacement should succeed the prototype device with reduced dimensions, improved compatibility with PC, increased speed of data acquisition and transfer and extended possibilities for additional functions in future.
9. For the future development it is proposed being attractive for practical application a new generation of multi-purpose and multi-parameter robotic system combining diverse methods of characterisation and acceptable for complex and continuous monitoring of cow's health and productivity.
10. Model predictions of pheromone compounds AA and PA based on BOVINOSE measurements on faeces samples from diary cows correlate strongly with results of direct GC measurements of those compounds in these samples.
11. Model predictions of estrus based on BOVINOSE measurements on faeces samples from diary cows correlate in general strongly with observations of estrus by human experts.
12. A disadvantage of the current models is that (1) the pheromone prediction basically only makes sense if a whole series is measured (to reveal the pattern of the series - since individual predictions make no sense due to scaling), and (2) the estrus prediction needs to be calibrated per cow by initial data from samples taken when the cow is not in heat, (nor pregnant).
Numerical prediction results of laboratory tests as well as of field trials have been extensively analyzed and evaluated, and discussed and reported. Following strong and weak points have been identified:
Strong points:
- the strong, robust correlation of eNose pheromone predictions with the actual pheromone data in the laboratory tests;
- the quite good correlation of eNose estrus predictions with the actual estrus data according to human experts.
Weak points for further research and improvements:
- the need of a reliable baseline determination. The required measurements needed for this baseline determination; the sensitivity of predictions to an erroneous baseline determination;
- deviations of field trial data compared to laboratory data (however, more field trials are needed for definite conclusions).
Economic assessment
Cost price calculations have been made and analysed for the latest prototype. The outcome of these activities has been described in deliverable D7.2.
Value analysis techniques were used to identify potential for cost reductions of the first prototype while maintaining functionality, and to guide further product development and improvements. Costs are meaningless without the context of revenues generated (profits), so, to support future exploitation, a customer value proposition has been included, in this case based on payback time and Internal Rate of Return calculations. These have been made for several dairy herd sizes.
The analysis reveals that both for the customer (mainly the farmer, but also veterinary doctors and AI stations) and for the device manufacturer the device is highly attractive in terms of annual returns.
With a modest improvement of the conception rate from 50 % to 54 % per cycle, the payback time for the farmer is about a year for a herd size of 30 dairy cows. For a relatively small herd size of 15 dairy cows the payback time is with about 26 months still very acceptable.
For the device manufacturer, the break-even point for earning back the investments in product R&D and market development is reached at selling 2152 units, less than about 1 % of the anticipated market size of 250 000 units. This implies a great market outlook. Assuming a gradual ('S-shape') market diffusion with 10 % market penetration after 5 years, and 12 % anticipated profits, the internal rate of return has been calculated at 66 % (5 years) and 78 % (10 years), respectively. These figures make it attractive for the BOVINOSE partners from a cost/benefits viewpoint to pursue further R&D and market development.
IMV and NIFA have conducted a market survey amongst French, Dutch and German farmers. All results have been listed in D7.2. For the cost price analysis the relevant data sections have been used as inputs. These surveys were also useful input for the deliverable D8.4 on the proposed BOVINOSE marketing strategy.
Recommendations for further research
In cows, the e-nose detects an estrus by measuring volatile chemical pheromone components which quantitatively correlates with ovulation time and in that way allows determination of optimal insemination time. The fact that domestic animals like horses, sheep and goats use olfactory signals to obtain information about reproduction status of partner, indicates that electronic nose could be applied for detection of ovulation time and consequently for determination of optimal insemination time. In addition, e-nose could be used in forestry and agriculture when rapid, sensitive and reliable method of volatile chemical marker(s) detection is needed.
Major conclusion
Electronic nose could be used in farming, forestry and agriculture when rapid, sensitive and reliable method of certain odour detection is essential for resolving tasks.
Dissemination and preparation for future exploitation
PUDF
The plan for use and dissemination of foreground (PUDF) has been completed in two stages. This document integrates both stages of the development on the PUDF.
A policy of wide dissemination of project results will be pursued in particular focused on potential end users of the project results, i.e. dairy farmers and veterinary / reproduction specialists and agricultural consultants. This second group plays an important role as they have usually strong working relationships with the farmers, and their opinion bears a lot of weight to the farmer. Therefore, scientifically sound proof of the working principle will in particularly disseminated to veterinary doctors and agricultural consultants who are active in the dairy value chain.
Peer-reviewed articles
Preparations have been made for four peer-reviewed texts for publication in scientific journals or as stand-alone publications in line with editorial and other requirements as requested by the publisher. These publications will contribute to disseminate research findings.
During the final six project months, the RTD performers have selected a number of potential fields of publication based on the results of the BOVINOSE project. RTDs KTH and SMART have taken the lead in preparing these. SMEs have been formally asked for approval.
Four scientific publications have been selected for publication. One has actually been pressed and three have been prepared for future publication.
1. W. Wiegerinck, A. Setkus, V. Buda, A.-K. Borg- Karlsson, R. Mozuraitis, A. De Gee, 'BOVINOSE: Pheromone-based sensor system for detecting estrus in dairy cows', 2011, Physics Procedia, in press.
2. Mozuraitis, R. et al., 'Sex pheromone dynamics during ovulation period in estrus-induced cows', 2011, in preparation.
3. Mozuraitis, R. et al., 'Sex pheromone dynamics during estrus cycle in dairy cows', 2011, in preparation.
4. Wiegerinck, W., et al., 'Rapid identification of dairy cows sex pheromone based on gas sensor featuring', 2011, in preparation.
One article has already been published in Physics Procedia 00 (2011) 1-2.
Three have been prepared for publication of which one article abstract has been used in the conference dealing with chemical signals in vertebrates 2011.
Marketing strategy report
Part of this task has been the research on specific consumer demands. Therefore, a dedicated dairy farmers questionnaire has been defined and developed. This has been used in interviewing dairy farmers and vetenarian doctors. The outcomes have been assembled, analysed and worked out into a description of the BOVINOSE market strategy as a basis for a successful market introduction. Surveys have been conducted amongst Dutch, German and French dairy farmers (mostly SMEs).
Market drivers have been identified that impact the future exploitation of BOVINOSE, in order to define appropriate measures integrated in a coherent and successful marketing strategy.
The final technology business plan has been identified and defined, describing the market, the product, the business, the marketing strategy, manufacturing, forecast of sales, cash flow, and breakeven, management and control of the business, as well as a further required financing.
Potential impact:
All beneficiaries will receive generated royalties pro rata the input they provided to the project (costs basis). At month 12 it is assumed that the project budget will be followed as a. This will be adjusted after the end of the project when actual costs are known.
The planning and the methodology of the exploitation for all five SME beneficiaries is presented below and will be refined along with the project.
PM
Measuring and diagnostic instrument manufacturer
- Define manufacturing plan for assembling BOVINOSE end-products, including machinery requirements, required finances, agreements with suppliers, quality control, personnel, shipment of goods etc.
- BOVINOSE system assembly.
Distribution / shipment of assembled end products: 21.27 %
COMPLEX
Electronic device design engineering and component manufacturer
- Define manufacturing plan for electronic components including machinery requirements, required finances, agreements with suppliers, quality control, personnel, etc.
Electronic component manufacturer (supplier to PM): 12.95 %
NIFA
AI product provider
- Assist in defining the BOVINOSE marketing strategy plan, including price setting, costs breakdown, price performance analysis, product leaflets / datasheets, press releases, promotion campaign, sales training
Area sales organisation Benelux, Germany, and USA: 20.21 %
IMV
AI product developer and manufacturer
- Coordinate BOVINOSE business plan
- Exploitation manager
- Coordination of defining the BOVINOSE marketing strategy plan, including price setting, costs breakdown, price performance analysis, product leaflets / datasheets, press releases, promotion campaign, sales training.
- Main sales organisation (EU and non-EU)
Exploitation manager: 34.23 %
AVESKAMP
Dairy farmer (end user)
- Field tests site for various stages of prototype sensors
- Test bed for component and system optimisation both during the project as during the first years of commercialisation after the project: 11.35 %
Exploitable foreground (overview)
Following is an overview of exploitable results generated by BOVINOSE.
As the final prototype is not fully finished at month 21, the partners could not make any decision on potential patents. At the moment, it looks like no patents will be applied for.
The project website is http://www.bovinose.eu