SE545695C2

SE545695C2 - Machine learning in a beverage distribution system

Info

Publication number: SE545695C2
Application number: SE2051546A
Authority: SE
Inventors: Martin Renck; Mattias Liss; Niclas Emdelius; Ulf Stenerhag
Original assignee: Wayout Int Ab
Priority date: 2020-12-22
Filing date: 2020-12-22
Publication date: 2023-12-05
Also published as: SE2051546A1

Abstract

There is provided a beverage distribution system (1) comprising: a plurality of reusable and portable beverage containers (2) where each of the plurality of beverage containers (2) carries a unique identification tag (9),the system (1) further comprising a plurality of identification tag readers (10), where at least some tag readers (10) are located at different geographical locations, the tag readers (10) being configured to read the identification tags (9),said tag readers (10) being connected to a server (3),said server (3) being configured to store a state of each of the plurality of beverage containers (2) in a digitally stored dataset (15), where the state is that the beverage container (2) is present at one of the tag readers (10),where the system (1) is configured to apply machine learning to the data set (15) to produce a decision rule for making a decision about taking an individual beverage container (2) out of service.

Description

Machine learning in a beverage distribution system Field of the invention This invention relates to systems and methods for distribution of beverage, in particular using individually traceable containers.

Background Although centrally provided water (”tap water") is available most countries, at least in densely populated areas, the water is often not suitable for human consumption. lnstead, potable water is produced and packaged (bottled) centrally and single use containers/bot- tles are then distributed using a fleet of trucks or other vehicles to stores, for example gro- cery stores, where they are available for purchase. Other beverages than water, such as beer and soda are also provided in grocery stores.

This has several drawbacks. First, because a human consumes large amounts of drinking water, it is a hassle to shop and carry the necessary amounts of drinking water (or other beverage). I\/|oreover, single use containers need to be produced. After use, the single use containers are, at best, recycled and used as fuel in waste burning plants. At worst, single use containers becomes a part ofthe increasingly troubling world-wide trash problem.

Summary of invention ln a first aspect of the invention there is provided a beverage distribution system compris- ing: a plurality of reusable and portable beverage containers where each of the plurality of beverage containers carries a unique identification tag, the system further comprising a plurality of identification tag readers, where at least some tag readers are located at different geographical locations, the tag reader being configured to read the identification tags, said tag readers being connected to a server, said server being configured to store a state of each ofthe plurality of beverage containers in a digitally stored dataset, where the state is that the beverage container is present at one of the tag readers, where the system is configured to apply machine learning to the data set to produce a de- cision rule for making a decision about taking an individual beverage container out of ser- vice.

The system, devices and methods described herein provides beverage in a reliable and safe manner. Production and distribution of beverage can be timed and fine-tuned to suit de- mand and consumption at the individual or population level. Beverage production can be fine-tuned to obtain a ”virtual pipe" of beverage. ln such a system there is a need for identifying and eliminating containers that are faulty. Either because some manufacturing error or because containers or another part ofthe sys- tem has been tampered with. The decision rule for making a decision about a container can be used for identifying and eliminating such containers, providing a safer experience to end USEFS.

The decision rule may comprise a behaviour pattern for the plurality of beverage containers.

The state is stored together with a time point. This provides information that a container was present at the tag reader at a certain time point.

The system may comprise at least one beverage dispensing station comprising one ofthe identification tag readers, said beverage dispensing station being adapted to connect to the beverage container such that beverage can be dispensed from said beverage con- tainer, where the dispensing station is configured to use the tag reader to detect a bever- age container is present at said beverage dispensing station.

The beverage dispensing station may have a dispensing prevention means, and the server is configured to, activate the dispensing prevention means when a beverage container is flagged for taking out service.

The beverage dispensing station may have a sensor, the dispensing station being config- ured to co||ect data from the sensor and provide it, using the wireless data connection, to the server, the server further being configured to include the sensor data in the dataset. This provides additional data for use in machine learning. The sensor may be for example a weight determining means or an accelerometer.

The system may further comprise a beverage filling station and at least one identification tag reader of the system is comprised in the beverage filling station.

Ina second aspect ofthe invention there is provided a method in beverage distribution system said beverage distribution system comprising a plurality of reusable and portable beverage containers where each of the plurality of beverage containers carries a unique identification tag, the system further comprising a plurality of identification tag readers, where at least some tag readers are located at different geographical locations, the tag reader being config- ured to read the identification tags, said tag readers being connected to a server, said server being configured to store a state of each of the plurality of beverage containers in a digitally stored dataset, where the state is that the beverage container is present at one ofthe tag readers, the method comprising a) applying machine learning to the data set to produce a decision rule, where the decision rule is configured to make a decision about taking an individual bever- age container out of service, b) applying the decision rule to the plurality of beverage containers.

The decision rule ofthe method may comprises a statistical model for the behaviour of the plurality of beverage containers in the system.

The obtained decision rule may be applied at least every predetermined time period, where a container that departs from the behaviour pattern is flagged in the dataset for being taken out of service.

Drawings The accompanying drawings form a part of the specification and schematically illustrate preferred embodiments of the invention and serve to illustrate the principles of the inven- tion.

Figs. 1 is a schematic drawing of a system.

Fig. 2 is a schematic drawing of a system.

Fig. 3a and 3b are schematic drawings of a beverage container.

Fig. 4 shows a beverage container.

Fig. 5 is a schematic drawing of a beverage dispensing station.

Fig. 6 shows a beverage dispensing station.

Fig. 7 is a beverage container in an upside-down position attached to a tap of a beverage dispensing station.

Fig. 8 is a beverage container supported to and connected to a beverage dispensing station. Fig. 9 is a cut away view of a beverage container and a beverage dispensing station.

Figs 10-11 are schematic top views of beverage dispensing station and beverage containers. Fig. 12 is a schematic drawing of parts of a beverage dispensing station.

Figs. 13-18 are flow charts showing methods.

Fig. 19 is a schematic drawing of control circuitry.

Detailed description The systems, methods and devices provided herein provide beverages at the point of con- sumption in a safe and reliable manner, in a ”virtual pipe" manner. Hence safe beverage can be produced and distributed to a consumer in a safe just-in time manner. Traceable containers and smart dispensing stations allow high-level monitoring of system 1 and its Va FlOUS COmpOHentS.

With reference to Fig. 1, the system 1 comprises a plurality of beverage containers 2 and a server 3. Beverage containers 2 are preferably reusable and portable. The system 1 may also comprise one or more beverage filling stations 4, one or more identification tag readers 10, and one or more beverage dispensing stations 50. The beverage dispensing stations are for example located in the homes of end-users, or at restaurants, bars, schools, hospitals, offices or other places of work, sporting facilities, or other places where it is useful to pro- vide beverage. Fig 1 also shows cell network 16 and wide area network 17, for example the internet.

The system 1 may comprise at least one beverage filling station 4 which is configured to fill reusable beverage containers 2 with beverage. Hence beverage filling station 4 may have means for producing, storing and filling beverage in containers 2 and for washing and stor- ing the containers 2. Hence beverage filling station 4 may comprise a beverage producing unit. The beverage producing unit may comprise a water purification unit 5 which is able to purify water using any suitable technology such as filters, heat, ultraviolet light or chemical means. The beverage filling station 4 will be described in more detail below. ln various em- bodiment the system 1 may also receive pre-filled containers 2 from an outside source, and system may then comprise a beverage container storageThe beverage distributed in the system 1 is preferably water or beer. ln various embodiment the beverage is potable water. ln other embodiments the beverage is beer.

Looking at Fig. 2, the system 1 may comprise a plurality of beverage filling stations 4, such as two, three, four, five or more filing stations 4, which will be described in more detail be- low. Each beverage container 2 may be logically associated in a dataset 15 with a certain filling station 4 so that each filling station 4 has a ”f|eet” of beverage containers 2 and dis- pensing stations 50 (see below), as seen in Fig 2. Fig. 2 is a schematic drawing showing how three different filling stations 4a, 4b 4c, each are associated with a f|eet 70a, 70b, 70c of containers 2 and separate f|eets 71a, 71b, 71c of dispensing stations With reference to Fig. 1 and 2, beverage filing station 4 may have a computer 19 that may be in network communication with server 3. ln various embodiments computer 19 of filling station 4 provides information to a filling station user and may enable filling station user to monitor and manage his or her f|eet of beverage containers 2 and dispensing stations Such a user may have access to only certain functions of the system Server comprises or has access to dataset 15, and dataset 15 will be described in more de- tail below. Server 3 may provide instructions to computer 19 regarding production of bev- erage and filling beverage into containers 2. Server 3 may be able to make predictions, us- ing data set 15, about beverage consumption in system 1 and use the predictions to pro- vide instructions to beverage filling station 4 regarding how much beverage should be pro- duced or how many filled beverage containers 2 should be produced. For example, when an increase of demand is predicted, production may be increased. Server 3 may also com- prise a price determining mechanism for determining a price for a filled beverage con- tainer The system 1, for example server 3, may comprise resources for placing orders for bever- age containers 2, such as replacement of empty beverage containers 2, payment, booking oftransport resources, and pricing determination mechanisms. End-users may be able to log into system 1 with a user account and place an order for a beverage container 2 using a mobile device, for example. Furthermore, server 3 may comprise logic that automatically ships beverage container 2 to a dispensing station 50 when needed.

Beverage is stored in tank 6 of beverage filling station 4. Beverage filling station 4 may also comprise a tap 14 and one or more areas for storage of beverage containers 2. Beverage filling station 4 may also have a washing station 7. Water purification unit 5 may be con- nected to an inlet 8 for non-purified water from some water supply such as a lake, a river or a utility that provides water. The filling station 4 may have an in|et or door (or both) for accepting beverage containers 2 and for discharging them. The beverage filling station 4 is described in even more detail at the end of the description.

Beverage containers 2 are filled with beverage in filling station 4. Preferably they are washed in washing station 7 before filing. The containers 2 then leave filling station 4 to be distributed to end-users.

End users typically have one more beverage dispensing stations 50. End-users may pick up beverage containers 2 at filling station 4 or more preferably have the containers 2 delivered to them. ln Fig. 1, T designates a beverage container 2 that is in transfer from filling station 4 to a dispensing station 50. W designates a beverage container 2 that has just left the washing station 7. S designates a storage of filled beverage containers 2 waiting for transport to end-users (beverage dispensing stations 50). Washing station 7 may be orga- nized as a processing line where the beverage container 2 enters filling station 4, is washed in washing station 7 and then exits washing station 7. Hence filling station 4 may have ded- icated areas for storing containers 2, washing containers 2, and filling containers 2. Washing station 7 may be completely automatic or semi-automatic.

As used herein and ”end-user” is a user that operates a beverage dispensing station 50. End user may buys filled beverage containers 2 or has filled beverage containers delivered to him/her. End-users may be able to order filled beverage containers.

A ”server user” is a user that manages the system 1 in some respect. A server user has access to the server 3 or parts thereof using a user interface. Server users may have different per- missions, such that some server users may have control over the entire system 1, whereas other server users have more limited access. Some users may only be allowed to control one or a few beverage fillings stations (4a, 4b, 4c), as system 1 may comprise more than one filling station 4, or a subset (70a, 70b, 70c) of beverage containers 2 or a subset (71a, 71b, 71c) of beverage dispensing stations 50 with reference to Fig.

The reusable beverage containers 2 in the system 1 each have a unique identification tag 9. The identity of the identification tag 9 is unique for each beverage container 2. Off course, by mistake there may be two or more containers 2 that have the same identity, but the purpose is that each beverage container 2 has a unique identity. Hence at least some ofthe plurality of beverage containers 2 have unique identities. The identity may be a unique combination of symbols such as digits or letters, for example a number.

A tag reader 10 is able to read the identity stored in the identification tag 9. In general, identification tag reader 10 can be of different kinds and is adapted to the kind of identifi- cation tag 9 used for the beverage container 2. For example, the tag reader 10 may be able to use induction to induce a current in circuits of the tag 9 to cause tag 9 to transmit a signal that can be detected by tag reader 10. It is preferred that the identification tag 9 is be an RFID tag in particular of a type that can be read with the aid of NFC (Near Field com- munication) technology. The unique identification tag 9 may be for example a RFID tag which can be read by a RFID tag reader 10. Or the identification tag 9 may be a bar code that can be read by a bar code reader. The identification tag 9 and the identification tag reader 10 may also be implemented using 5G technology, in which case the identification tag 9 com prises a battery or other energy storage. It is preferred; however, the identifica- tion tag does not comprise a battery.

A schematic drawing of a beverage container 2 is seen in Figs 3a and 3b. A non-limiting ex- ample ofa container 2 made of stainless steel is shown in Fig. 4, and Figs. 7- The plurality of reusable beverage containers 2 preferably have a volume of from 2 to 100 litres, more preferably 3-30 litres and even more preferably 8 - 25 litres. The containers 2 are preferably portable when filled with beverage. For example, the container 2 may have a handle 11 (Fig. 4) for carrying the container. The containers 2 are preferably made of Stainless steel or a polymer material such as plastic. The material may be non-transparent.

The container 2 may have any suitable shape, for example a cylindric shape.

A different suitable shape and size may be the type of container used in ”water coolers" frequently found in office spaces etc. The beverage container 2 is typically not designed to drink directly from.

The containers 2 should be easy to clean and have a low weight and be durable to allow use many times. The number of use cycles is preferably predicted to be from 10 - 1000 for an individual container 2, where each cycle includes one filling the container 2 with at least some beverage and providing it to an end user.

The identification 9 is attached to the container 2 or is otherwise comprised in container 2. For example, an NFC tag may be glued to the container 2 or a bar code may be in the form of an adhesive sticker attached to the outer surface ofthe container 2. A tag 9 may also be inserted into container 2 at production, for example during casting of a containermade of a polymer material.

The identification tag 9 is preferably attached to the container 2 in a permanent manner.

However, single use tags or limited time use tags 9 may also be used.

That the identification tag 9 is attached in a ”permanent manner” means that it is in- tended to be attached to the beverage container 2 for the duration of the expected life- time of the beverage container.

The tag 9 preferably has a low weight and a small size. The weight of tag 9 is preferably negligible compared to weight of container 2. This is also true for second tag 18 described in more detail below. The weight of the identification tag 9 or second tag 18 is preferably less than 50 grams, even more preferably less than 10 grams and most preferably less than 1 gram.

The beverage container 2 has at least one opening 12 for filling and tapping beverage. The opening 12 is closable, for example with a lid 13. The lid 13 may be disposable or reusable. lt may be possible to pour beverage from the opening 12 which is then used as a spout. Alternatively, the beverage container 2 is attached to a beverage dispensing stationand beverage dispensed from the station 50, as described in more detail below.

The system 1 comprises a plurality of reusable beverage containers 2. The number of bev- erage containers may be at least 100, preferably at least The tag reader 10 is arranged to detect the identity tag 9 from a suitable distance. The tag reader 10, is preferably, as mentioned, an RFID or NFC tag reader and hence comprises suitable hardware as is known in the art. The tag reader 10 may be arranged not to detect tags 9 that are beyond a certain distance, which may be 10 m, preferably 5 meters, 1 me- ter or 20 cm, more preferably 10 cm and most preferably 5 cm. Hence the tag reader 10 may have detection distance limit. The specificity ofthe tag reader 10 may be increased by using a directional antenna in the tag reader 10, as described in more detail below with reference to Figs. 10 and For example, the dispensing station may have a directional antenna that directs the zone of detection 57 towards the intended position of the beverage container 2 on the dispens- ing station.

System 1 may comprise at least two different tag readers 10. The at least two tag readers 10 may have different geographical locations, meaning that they are at least 20, more preferably at least 100 meters apart. For example, one tag reader 10 may be at a filling station 4 or a beverage container storage 22 whereas one tag reader may be at a dispens- ing station 50. For example, tag readers 10 may be present in dispensing stations 50 at dif- ferent geographical locations in a geographical area such as a city, for example different homes of end-users. Hence the tag readers 10 may be present at stationary geographical positions. However, tag readers 10 may also be carried in trucks or cars used fordistribution of beverage containers 2 or manually carried by personnel involved in distri- bution of beverage containers The geographical position is preferably a known geographical position. The geographical position may be known by for example, storing the coordinates or an address in the dataset 15. However, geographical positions are not necessarily known. Positions may be obtained for example, automatically by using positioning services (GPS), addresses of end users as known from client registers in system 1, or by manually entering the positions. lt may be sufficient, for example in the machine learning application of Fig. 17, to know the distance between various geographical positions, for example distances between a beverage filling station 4 and various beverage dispensing stations ln various embodiments at least one tag reader 10 is comprised in one or more a beverage dispensing stations 50. The tag readers 10 may be used by system 1 to detect that a bever- age container 2 is present at one of tag readers 10, for example present at the one of the dispensing stations 50. The tag reader 10 may be used in system 1 to detect that a bever- age container 2 is present at, supported by or connected to a certain dispensing station Tag reader 10 is in digital communication with the server 3. System 1 may use tag reader 10 to gather useful information in dataset 15 about a certain beverage container 2. 1The infor- mation is at least that the beverage container 2 is present at the tag reader 10. The system 1 may also use a tag reader 10 to record the state of a certain beverage container 2, for example that a beverage container 2 is in a ”washed” state, ”present at beverage station" state or similar.

A tag reader 10 may be arranged to detect that a beverage container 2 is present at one of the tag readers 10, for example at beverage dispensing station 50, and provide that infor- mation to the server 3. This may be stored in dataset 15, preferably together with a time point for detection.The tag reader 10 is preferably a part of a beverage dispensing station 50 or provided in a beverage filling station 4, and the tag reader 10 may use hardware and software ofthose components to communicate with the server 3. However, tag reader 10 may also be inde- pendent from a filling station 4 or a beverage dispensing stations 50 such as the case may be with a tag reader present at distribution centre comprising a beverage container stor- age 22 or present in a vehicle or carried by a person, such as a person that distributes bev- erage containers 2. Such independent tag reader 10 may each comprise a battery or other power source, a processor, a memory, and a wireless network interface, and control cir- cuitry (Fig. 19), and are capable of communicating with server Filling station 4 may have one or more tag readers 10. For example, a tag reader 10 may be located in a frame on a door on beverage filing station 4 to detect movement in or out from the beverage filling station 4. As a further example, a tag reader 10 may be located at the exit of the washing station 7. Various parts of all of beverage filling station 4 may have a dedicated workflow so that a beverage container 2 moves from one part to the next, for example, entering filling station 4, going through washing station 7, being filled at tap 14, being stored, and then exiting filling station 4 and tag reader 10 may be arranged to detect the presence of the container 2 at the various stations or areas of beverage filling station The tag reader 10 may be for example be arranged so that is able to detect when a bever- age container 2 has left the washing station 7 and hence can be assumed to be clean. For example, when the tag 9 is an NFC tag the tag reader 10 is an NFC reader arranged at a suitable distance from exit of washing station 7 so that beverage containers 2 that have been washed (and not other beverage containers 2) are detected as the pass the exit of washing station 7. Similar arrangements may apply wherever tag reader 10 is placed in the system 1. Hence a tag reader 10 is placed in system 1 to ensure that the detected state can be determined with some certainty.

Tag reader 10 or a plurality of tag readers 10 may thus be arranged to detect beverage container 2 as it passes various parts of system 1 and provide data about this to server 3,and server 3 may store such information in dataset 15. The server 3 receives the infor- mation from the tag reader 10 and may use the information to change a state of an indi- vidual beverage container 2 in dataset The server 3 may be able to combine the information (i.e., the identity of the beverage container 2) from the tag reader 10 with a predetermined piece of information associated with a certain tag reader 10. For example, an identity reading from a certain tag reader 10 always determines a certain predefined state for a container 2. For example, reading from a certain tag reader 10 may change the state ofthe container 2 from ”unwashed” to ”washed” for example, or from ”in station” to ”have left station” (”unwashed” may also be referred to ”have entered station"), as described in further detail below.

Hence a tag reader 10 may have an identity in system 1. When the tag reader is part of dispensing station 50, the identity ofthe tag reader is preferably the same as the identity of the dispensing station.

Detection of a tag 9 of a beverage container 2 by a tag reader 10 may for example result in the server 3 setting the state of the beverage container 2 in the dataset 15 to one ofthe following: a) the container 2 has entered the beverage filling station b) the container 2 has been washed, c) the container 2 has been filled with beverage in the beverage filling station 4, d) the container 2 has left the beverage filling station 4, e) the container 2 is in distribution, f) the container 2 has been left with an end user of a beverage dispensing station 50, g) the container 2 is present at a beverage dispensing station 50, h) the container 2 is not present at any beverage dispensing station States may be exclusive such that an individual container 2 can only have one state at a time. For example, a container 2 should not be able to be present at a beveragedispensing station 50 and not present at any beverage dispensing station 50. This may be provided with logic in server 3, for example database logic.

The tag reader 10 may be able to read a plurality of tags 9 simultaneously or during a short time. For example, if a truck loaded with a number of beverage containers 2, an RFID tag reader 10 may be arranged to read the tags 9 of all the beverage containers 2 in the truck as the truck passes the RFID tag reader Returning to Fig 1, system 1 comprises server 3. As described below, server 3 may be a vir- tual server. Server 3 maintains or comprises a digitally stored dataset 15, which may be in the form of a database. Any suitable form of database architecture, such as an SQL data- base, may be used. The identity of each beverage container 2 is stored in dataset 15. Da- taset 15 may maintain information that is particular to an individual beverage container 2 such as for example information about when a container 2 was put into use in system 1, number of filling cycles, date of manufacture, make, etc. This information is associated with the identity of the beverage container 2. The dataset 15 may also comprise infor- mation about the current state ofthe beverage container 3 as recorded by tag readeras described above. lt may also be possible, for a server user, to add containers 2 to database and then logi- cally associate them with a particular filling station 4. lt may also be possible to manually enter information into dataset 15, for example if a certain beverage container 2 is broken and/or has to be taken out of service.

Hence, the server 3 is configured to store data relating to the plurality of reusable bever- age containers 2 in a digitally stored dataset 15. Preferably the server 3 is also configured to store data relating to the plurality dispensing stations 50 in the dataset Server 3 comprises suitable logic components and software to carry out the methods as described herein, such as but not limited to: a beverage consumption tracking module, beverage container tracking module, error detection logic, container error flagging module, dispensing prevention means activation module, beverage production control logic, database interface, network interface and operating system.

For example the following information may be associated with the identity of certain bev- erage container 2 in the dataset 15 (some of these will be explained in more detail below): type of beverage, beverage container 2 has been filled, amount of beverage left in bever- age container 2, number of filling cycles, date of manufacture, current position, beverage in container 2 should not be consumed, identity of second tag 18 (tampering tag), state of second tag 18, the container 2 has entered the beverage filling station 4, the container 2 has been washed in the beverage filling station 4, the container 2 has been filled with bev- erage in the beverage filling station 4, the container 2 has left the beverage filling station 4, the container 2 is in distribution, the container 2 has been left with a user of a beverage dispensing station 50, the container 2 is present at a beverage dispensing station 50, iden- tity of filling station 4 to which the beverage container 2 belongs, the container 2 is not present at any beverage dispensing station 4, identity of a previous dispensing station 50 the beverage container 2 has been associated with, dates of previous filing cycles, date and time of tag readings, beverage dispensing events, beverage consumption data. These are non-limiting examples.

For example the following information may be associated with the identity of a dispensing station 50 in the dataset 15: identity of current beverage container 2 present at dispensing station 50, no beverage container 2 present, sensor data, beverage consumption data, communication log, flag for no changes since previous communication session, dispensing prevention means 65 are to be activated, dispensing prevention means 65 have been acti- vated, error messages, battery status, beverage dispensing events, sensor data, identity of filling station 4 to which the dispensing station 50 is associated, identity and contact de- tails of end user, log in details for end-user, order status etc. These are non-limiting exam- ples.

All data may be stored for previous events so that a history for each beverage containerand dispensing station 50 is stored in datasetServer 3 may comprise logic that detects errors with regard to a beverage container 2 or a beverage dispensing station. Hence an individual beverage container 2 should not be able to change state directly from ”unwashed” to "filled". Such a change (from ”unwashed” to ”filled”) should only be allowed ifthe container 2 has been in the state ”washed” after ”unwashed". lf tag readers 9 provides such conflicting information, there is an error. The error may be detected by software in the server 3, for example by database software. The database software may for example be adapted to generate an error flag. Errors can be detected in other ways for example the machine learning method of Fig. 17. An error re- garding a container 2 may lead to the generation of an error report by server 3 or may lead to that the container 2 is excluded from distribution. Exclusion may for example be carried out such that a certain container 2 is flagged to not be filled or not to be distrib- uted to an end-user. Warning may be provided as sound signals, or in a display or a flash- ing light in filling station 4, a message to end user or server user or activation of dispensing prevention means 56 (see below).

With reference to Figs. 1-2 and Figs. to 5-11, system 1 preferably comprises one or more beverage dispensing stations 50. The dispensing station 50 is used for dispensing beverage from a beverage container Many different designs for the beverage dispensing station 50 are possible. The beverage dispensing station 50 of Figs. 6-9 comprises a free-standing housing 51 that supports the beverage container 2 which is placed on top of the housing 51 with the opening 12 down- wards. A separate tap 52 is attached to the opening 12 and is used to control dispensing by the end user. Beverage may then be dispensed using gravity flow. Tap 52 comprises a valve (not shown), handle 66 for operating the valve and connector 54. The tap 52 rests in a V-shaped notch 53 of the housing 51. Again, this is just an example and tap 52 may also be integrated with housing 51. ln Figs. ln general, a tap 52 of a beverage dispensing station 50 may be operated in any suitable way such as for example with by a push button, or anyother suitable means for controlling a valve. Tap 52 may for example comprise an elec- tronically controlled valve for dispensing beverage. ln general, the housing 51 may comprise control circuitry and various sensors such as weight determining means 55, accelerometer 56, battery 63, a LED light 64 and wireless network interface 58 as described below.

There are numerous possibilities for the design of dispensing station 50. Other possible designs include a design similar to a ”water cooler" frequently found in office workspaces.

The dispensing station 50 may be suitable to stand on a bar, a table or a bench top or it may be designed to be free standing on a floor or on the ground. The beverage dispensing station 50 may also be integrated into furniture, in particular furniture suitable for kitch- ens, restaurants, and bars. The dispensing station 50 may also be integrated into a fridge.

Beverage may be dispensed from the dispensing station 50 as flowing beverage using gravity flow or by providing pressure inside container 2. Figs. 6-9 show a beverage dis- pensing station 50 suitable for dispensing beverage with gravity flow. The beverage is preferably dispensed from the beverage container 2 while being connected to the bever- age dispensing station 50, or while being supported to the beverage dispensing station When pressure is used, as may be the case with beer, the dispensing station 50 may com- prise means for pressurising the beverage container 2 as is known in the art. Typically, the beverage dispensing station 50 then comprises means for providing pressurized carbon di- oxide (other inert gas) into the beverage container 2. The means for providing carbon di- oxide may comprise a first pipe that is inserted into the container 2 and provides pressur- ised carbon dioxide to drive the beverage out from the container 2 from a second pipe. The dispensing station 50 with use for beer may be designed so that the opening 12 of container 2 is directed upwards, and the dispensing station 50 or a part thereof may then be designed to rest on top of the container 2. ln particular, a connector for piping and means for providing carbon dioxide may rest on top ofthe openingDispensing station 50 may be configured to mechanically support a beverage container 2. Hence, a part or the entire weight of beverage container 2 may rest on dispensing station 50. Dispensing station 50 or a part thereof may be arranged to mechanically receive a bev- erage container 2. The beverage container 2 may be able to be mechanically connected to the dispensing station 50, for example with the use of lugs, bayonet coupling, click-on or threads or other connection means. The dispensing station 50 may be able to engage with the beverage container 2. Preferably the dispensing station 50 is such that only one bever- age container 2 is supported by or connected to beverage dispensing station The container 2 can be mechanically connected to the dispensing station 50 such that bev- erage can be dispensed from the container 2. The dispensing of beverage preferably oc- curs through a part ofthe dispensing station 50. For example, the dispensing stationmay have a connector 54 which connects to the opening 12 of the beverage container System 1 preferably comprises a plurality of dispensing stations 50, such as two, three, four or more dispensing stations 50, such at least ten or at least 100 dispensing stations 50. Each dispensing station 50 may have a unique identity in system 1. The identity is stored in the dataset 15. A beverage dispensing station 50 is may be logically connected to one certain beverage filling station 4, for example in dataset 15, as shown in Fig. 2. The logical connection may be in the dataset The beverage container 2 may be switched such that a first beverage container 2 that is mechanically connected to or supported by the dispensing station 50 is be replaced with a second beverage container 2 which then becomes mechanically connected to, or sup- ported by, the dispensing station Looking at Fig. 12, beverage dispensing station 50 has control circuitry comprising a memory 60, a processor 61, and a bus 62. The control circuitry may be partly arranged on a printed circuit board (PCB). The control circuitry is electrically connected to the various sensors, the tag reader 10, the wireless communication interface 58, the battery 63, theLED 64 and the dispensing prevention means 65, and. The control circuitry may comprise a timer and a clock. Memory 60 stores software for pa rticipating in various methods de- scribed herein.

The dispensing station 50 furthermore may further comprise a battery 63 for powering the various electric and electronic components of the dispensing station 50 such as the control circuitry, the sensors, the wireless network interface 58, tag reader 10 and the dis- pensing prevention means 65. Power may also be provided by a connector and a regular power outlet.

Beverage dispensing station 50 preferably has one or more sensors. The sensor ofthe beverage dispensing station 50 may for example be a weight determining means 55, an accelerometer 56, or a beverage flow sensor. ln a preferred embodiment the dispensing station 50 has at least a weight determining means 55 and an accelerometer The beverage dispensing station 50 may have at least two sensors, a fist sensor that is able to wake up the control circuitry from a sleep state to a wake state and second sensor that determines the amount of beverage that is consumed. The first type of sensor may be able to wake up processor 61 with the use of an interrupt pin of processor 61, by provid- ing a signal to the interrupt pin. ln a preferred embodiment the first type of sensor is an accelerometer 56, and the second type of sensor is weight determining means Returning to Figs. 5 to 12, the dispensing station 50 preferably has a tag reader 10. The tag reader 10 is arranged to detect the identity of the beverage container 2 that is present at the beverage dispensing station 50. Preferably the tag reader 10 is arranged to detect a beverage container 2 that is supported by or mechanically connected to the dispensing station 50. Preferably the tag reader 10 of a dispensing station 50 only detects the tag of a certain beverage container 2 if the beverage container 2 is supported by the dispensing station 50 or mechanically connected to the dispensing station 50 such that beverage can be dispensed from it.

The tag readers of Figs. 10 and 11 shows two different examples of how the tag reader 10 can be arranged in a beverage dispensing station 50 (schematically shown from above) and only detects a beverage container 2 supported by the dispensing station 50. ln Figsand 11 the beverage container 50 rests on dispensing station For example, the tag reader 10 may be NFC tag reader 10 that reads at a short distance and in a certain direction, forming a field of detection 57 directed to the site where the beverage container 2 is located on the dispensing station 50, see Fig. 10. A tag 9 outside the field of detection 57 is not detected by tag reader 10. The tag reader 10 of Fig. 11 has an even shorter range of detection, and only detects the tag 9 when the beverage con- tainer 2 is oriented in the correct manner so that the tag 9 faces the tag reader 10. This may be achieved with the use of a directional fit between the beverage container 2 and the dispensing station 50. For example, the correct orientation may be achieved with use of keying between the beverage container 2 and the dispensing station 50. The arrange- ments of Figs. 10 and 11 are examples only.

The tag reader 10 may be activated in various manners by the control circuitry ofthe dis- pensing station 50. For example, the tag reader 10 may read the tag 9 of the beverage container 2 at least every predetermined time period such as at least once every hour or every day. The tag reader 10 may also be activated in various manners, such as by the ac- celerometer 56, or weight determining means 55, as described herein.

Furthermore, ifthe weight determining means 55 detects a weight but no tag 9 can be de- tected it may indicate that a false beverage container 2 is supported by dispensing station The beverage dispensing station 50 is in data communication with server 3, preferably us- ing at least partly wireless data communication. Hence the dispensing station has a wire- less communication interface 58. The wireless data communication may preferably be provided by the dispensing station 50 acting as user equipment in cell network 16, such as a 3G, 4G or 5G cell network 16. However, other standards such as LoRa may be used.Hence the dispensing station 50 may have an antenna as a part ofthe wireless network interface 58 suitable for cell network communication, particular for acting as user equip- ment. Beverage dispensing station 50 may provide data from a sensor to the server 3, such as data from the tag reader 10, weigh determining means 55 accelerometer 56, tem- perature sensor 59 or flow sensor to server 3, as described below, using the data connec- tion.

The control circuitry ofthe dispensing station 50 may have a sleep state and a wake state, where the energy consumption is substantially lower in the sleep state. The control cir- cuitry may be woken up from the sleep state for example by a sensor, in particular the first sensor. The control circuitry may go to sleep after a time of inactivity, for example af- ter more than 15 minutes of inactivity. The control circuitry ofthe dispensing station 50 may be configured to wake up from sleep when dispensing is detected, for example by the accelerometer 56 or the weight determining means 55. Weight determining means may be connected to an amplifier that amplifies the signal for this purpose.

Fig. 13 describes a communication session between a dispensing station 50 and server 3 involving a tag reader 10 reading the tag 9 of a beverage container 2. A tag reader 10 that is a part of the filling station 4 or an independent tag reader 10 carries out a communica- tion session with server 3 in a similar manner. ln step 100 the tag reader 10 in particular a tag reader of a dispensing station 50, detects the tag 9 of an individual beverage container 2. This involves reading the identity of the tag 9 which is also the identity of the beverage container 2. The identity is stored in the memory 60 of the dispensing station 50, prefera- bly together with a time point for detection. Preferably other sensor data, such as data from weight determining means 55 or accelerometer 56 is also determined and stored, see below. Step 100 may be initiated by for example first sensor, such accelerometer 56 or weight determining means 55, that wakes up the control circuitry of the dispensing sta- tion 50 from sleep or otherwise provides a signal to control circuitry that causes tag reader 10 to read the tag 9. Step 100 may also by initiated by the end-user interacting with the beverage dispensing station 50, for example switching on the beverage dispensing station with use of an on/off switch.ln step 101, the identity ofthe container 2 is provided to the server 3. Preferably the bev- erage dispensing station 50 uses the wireless network interface 58 for establishing a data connection to server 3. Preferably the identity of the dispensing station 50 is also provided to the server 3. The communication may be encrypted. ln step 102 the data is stored in the dataset 15 and associated in the dataset 15 with the appropriate beverage container 2 and dispensing station 50. Preferably the identity of the beverage container 2 is associated with the identity of the dispensing station 50 in the da- taset. The server 3 may then verify the identity of the beverage container 2. For example, that the beverage container identity is present in dataset 15 and not present at any other beverage dispensing station 50. An error message may be sent to a server user or a end- user if an error is detected, or the dispensing prevention means 65 may be activated, see below.

The data may also comprise a message from dispensing station 50 indicating that the dis- pensing station 50 works as intended. Such a check may be carried out by control circuitry ofthe beverage dispensing station 50 at wake up or as scheduled. ln one embodiment the transferred data comprises the identity of the dispensing station 50, the identity of the beverage container 2 and a systems ok message. This may for example be provided as a text message (SMS message) from beverage dispensing station.

A communication session between the dispensing station 50 and the server 3 may com- prise one or mor ofthe following: dispensing station 50 identity, present container 2 iden- tity, battery status, dispensing events, weight, consumption data, other sensor data such as data from accelerometer, temperature data, or a message that indicates no change since previous session. A ”no change” message is an efficient way of saving data traffic and power. Sensor data such as accelerometer data, temperature data or data from weight determining means 55 may be associated with a time point for detection which is stored in the memory 60 ofthe beverage dispensing stationCommunication between dispensing station 50 and server 3 may be carried out at any suitable schedule. For example, the dispensing station 50 may be configured to connect to the server 3 at least every predetermined time interval such as, for example, at least every hour, every day or every week.

Table 1 below show a highly simplified and schematic example of dataset 15 showing the identities of three different containers 2 and three different dispensing stations 50. As mentioned, the dataset 15 may include many other types of data that are left out in this example. Table 1 shows how the same table is used for beverage containers 2 and bever- age dispensing station 50, but two or more tables with cross references may off course be used as the dataset 15 may be structured in any suitable manner.

Activate Identity of Remain- Last dis- disp. pre- ldentity of bever- dispensing ing pensing vention Filling age container station amount Battery status event means station 10:55. Nov 26, 2020, 5647 2354 40% 73% 10:55 no 1 16:03, Nov 1225 0122 83% 14% 10, 2020 no 1 07:32, Oct, none 2355 n/a 99% 22, 2020 no 2 1002 none 100% n/a n/a n/a 2 Table2355 is an example of a beverage dispensing station that does not presently have a bever- age container 2. Beverage container 1002 is an example of a beverage container 2 that is presently not associated with any beverage dispensing station. ln various embodiments the dispensing station 50 comprises an accelerometer 56. The ac- celerometer 56 may be any type of useful accelerometer. The accelerometer may for ex- ample be of the type used in mobile phones, such as the iPhone. The accelerometermay be able to detect dispensing or movement of the dispensing station such as jolts orvibrations or changes in orientation or impact, such as ifthe dispensing station 50 falls to the floor from a table. The accelerometer 56 may also be able to detect tampering. Accel- erometer readings above a certain threshold may indicate impact, caused by the dispens- ing station falling. Such accelerometer readings may trigger a self-test of dispensing sta- tion 50, and also a report to the server 3. The accelerometer 56 may have a threshold for detecting movement so that background movements such as traffic does not trigger a sig- nal from the accelerometer A signal from the accelerometer 56 may be used to wake up the control circuity from sleep, for example a signal to an interrupt pin on the processor 61 ofthe control circuitry. A signal from the accelerometer 56 may be used to active the wireless network interface 58 or to activate the weight determining means A signal from the accelerometer 56 may be configured to cause the tag reader 10 to read tag 9, for example as shown in steps 100-102 of Fig. 13. The accelerometer 56 provides a signal to the control circuitry that in turn activates the tag reader 10, which reads the tag 9 ofthe beverage container in step 100, and stores the identity in the memory 60 ofthe control circuitry of the beverage dispenser 50. For example, this may be used to detect if a beverage container 2 has been replaced. lf a beverage container 2 has been replaced, the identity of the new container 2 read by tag reader 10 is associated with the identity of the beverage dispensing station 50 in the dataset Accelerometer data may be stored in the memory 60 ofthe dispensing station 50, prefera- bly together with a time point for detection. ln various embodiments, the dispensing station 50 has weight determining means 55 for determining the weight of a beverage container 2 that is supported by the dispensing sta- tion 50. The weight determining means 55 may be of any suitably type. For example, a strain gauge type or a load cell may be used. Any suitable type of weight sensor may be used. For example, a metal beam that supports the beverage container 2 is fitted with strain gauge that detects the strain in the beam. The weight determining means 55 may also comprise a piezoelectric crystal that provides voltage when a mass change provides a change in stress of the crystal.

The weight determining means 55 may be used to determine the amount of beverage in container 2. This information may be used by server 3 to determine when a container 2 should be replaced. For example, when the weight of beverage container 2 supported by a dispensing station 50 is below a threshold weight, a new full container 2 is automatically shipped to the user of the dispensing station 50. Server 3 may comprise logic that com- pares the current filling level (weight) of a beverage container 2 with a threshold weight and automatically provides an order to filling station 4 to produce and ship a new bever- age container 2 when the weight is below the threshold weight.

Information about consumption in dispensing stations 50 may also be used in an aggre- gated manner to track consumption of beverage from a plurality of dispensing stations For example, the consumption for a subset 100a of dispensing station can be compared with the consumption of a different subset 100b (Fig. 2) of dispensing stations. Subsets may be selected based on geographical location, demography, or the like.

Beverage consumption data or weight data may also be used as described elsewhere herein for example in relation to machine learning, see below. lt should be noted that in- stead of weight determining means 55 a flow meter in the beverage dispensing station may be used.

Beverage consumption may be determined by control circuitry in dispensing station 50 or by server 3. For example, the dispensing station 50 may provide weight data to server 3 which calculates the consumption. Typically, a weight after the latest dispensing event is subtracted from the weight after the previous dispensing event, to determine the differ- ence in weight. The difference is used as the beverage consumption of the dispensing event. Or the dispensing station 50 may do the calculation and provide consumption data to the server 3. Or the dispensing station 50 may only report that the weight is below a threshold weight. Weight data may be filtered such that only differences above a certainthreshold is used in order to filter out noise. lt may be an advantage ifthe consumption or filing level is determined by dispensing station 50 rather than server 3 in order to reduce data traffic in system 1 and data handling by server The following is an example of calculation of beverage consumption.

Initial total weight 11 kg Less tare weight: (container possibly tap shown in Fig.7): 1 kg Tared weight 10 kg 1st measurement: 9.8 kg * consumption: 0.2 kg.

Tare weight may be prestored in memory 60 of beverage dispensing station 50. The use of a tare weight is optional, it may not be needed if it can be assumed that the initial weight is known, for example from the filling station Hence the weight data may be provided to the server 3 such as weight data, amount of beverage remaining in container 2, amount of beverage dispensed, or that that the weight is below a threshold weight has been passed or by triggering the delivery of a new con- tainer 2. All this is referred to as ”weight data or beverage consumption data” herein.

The weight determining means 55 may be configured to determine the weight at least every predetermined time interval. For example, the weight determining means 55 may be configured to determine the weight at least every week, every day, every hour, every minute or every second.

The dispensing station 50 may be configured to determine the weight or consumption when dispensing has occurred. Dispensing may be detected with a separate sensor, such as a flow sensor, or a mechanical switch that detects the position of tap 52. ln a preferred embodiment dispensing is detected with accelerometer 56. Hence a signal from theaccelerometer may activate the weight determining means 55. First sensor may also acti- vate the weight determining means The accelerometer 56 is preferably arranged in dispensing station 50 such that dispensing beverage from the beverage container 2 activates the accelerometer 56. The signal from an accelerometer 56 may be used to wake up the control circuitry to activate the weight determining means 55, for example by using an interrupt pin on a processor 61 of control circuitry. The accelerometer 56 may for example detect movement ofthe beverage dis- pensing station 50, beverage flow, movement of tap 52 or a cup detector. ln a preferred embodiment the accelerometer 56 detects movements of the dispensing station 50 itself or the beverage container 2 attached to, or supported by, the dispensing station 50. For example, the accelerometer 56 may be attached to the housing 51 in order to detect vi- bration or movement of housing 51 that indicates dispensing.

System 1 may be configured to determine a dispensing event. A ”dispensing event” com- prises at least information about an amount of beverage that has been dispensed from a certain beverage container 2 presently supported by or connected to a beverage dispens- ing station 50 or weight data for the beverage container 2. That the beverage container 2 is present at beverage dispensing station may be detected by tag reader 10 of beverage dispensing station 10. Weight determining means 55 or a flow meter may be used for de- termining an ”dispensing event”. The dispensing event may be stored in the memory 60 of the control circuitry of the dispensing station 50. The dispensing event may be stored to- gether with a time point for dispensing. The dispensing event may trigger a timer. lt should be noted that ”dispensing” means, herein, the fact that dispensing occurs or likely occurs and may be detected with for example an accelerometer 56 and may not amount to a ”dispensing event”.

Weight data may be provided from the dispensing station 50 to the server 3 at any suita- ble interval, such at least every predetermined time interval, such as, for example, at least every week, every day, every hour or every minute. lt is an advantage if communicationtakes place as little as possible in order to save the battery 63 of dispensing station 50 and keep data traffic at a minimum. ln a preferred embodiment the weight determining means 55 is activated by the first sen- sor. The dispensing station 50 may be configured to activate the weight determining means 55 after a predetermined waiting time. The predetermined time may be at more preferably at least 10 seconds and even more preferably at least 1 minute and even more preferably at least 5 minutes. This may be advantageous because the weight determining means 55 preferably determines the weight of the beverage container 2 once dispensing has finished, which may take few seconds. Also, it may be desirable to store dispensing into two mor more several glasses as one dispensing event, in order to reduce data han- dling in system The beverage dispensing station 50 may be further configured to establish a data connec- tion and provide weight data or consumption data to the server 3 after waking up, prefer- ably after a predetermined waiting time after waking up. The waiting time may be calcu- lated from wake up of the control circuitry or from the determination of the weight deter- mining means 55 and may be for example at least 10 seconds and even more preferably at least 1 minute and even more preferably at least 5 minutes.

The weight data or beverage consumption data is preferably provided to the server 3 to- gether with the identity of the beverage container 2 and the identity of the dispensing sta- tion 50. The dispensing station 50 may be configured to activate the tag reader 10 and read the tag 9 of the beverage container 2 each time the control circuitry is woken up from sleep, each dispensing event or each time the weight is determined, or each commu- nication session.

Fig. 14 shows a method involving a beverage dispensing station 50. lt should be noted that the method shows a specific embodiment and that the steps of Fig.14 may comprise various separate embodiments of the current disclosure, and that the various embodi- ments may be combined in any suitable manner. For example, the first and second waitingsteps are separate embodiments. Furthermore, it should also be noted that that the tag reader 10 may be activated in any of steps 201 to 204 to read the tag 9 of the beverage container 2. The control circuitry may also carry out other checks of the circuitry of dis- pensing station 50 such as check battery status, detect temperature etc. in any of steps 201 to ln step 200 the first sensor wakes up the control circuitry from sleep. The fist sensor is preferably is an accelerometer 56 that detects movement. 202 is an optional first waiting step that awaits completion of dispensing. As described herein, the control circuitry may have a timer. The predetermined time may be at least 10 seconds, more preferably at least 1 minute and even more preferably at least 5 minutes. A longer waiting time ensures that dispensing has finished and may also be used to catch a dispensing event made by a second person, which saves energy and computing power. However, it may be useful to have short waiting time to have more granularity in the dis- pensing data. ln step 203 the control circuitry activates the second sensor which preferably is weight de- termining means 55. The weight ofthe beverage container 2 is determined and stored and preferably beverage consumption is calculated and stored in memory 60. The weight or consumption data may be stored as a beverage dispensing event. The event may be stored together with a time point for dispensing. The second sensor may also be flow sen- sor, in which case the first waiting step 202 preferably is omitted.

Step 204 is an optional second waiting step. During step 204 any additional consumption events are stored in memory 60. The additional dispensing events may be detected, for example, by the accelerometer 56 again activating the weigh determination means 55. The waiting period is more preferably at least 10 seconds and even more preferably at least 1 minute and even more preferably at least 5 minutes. ln step 205 the control circuitry provides the beverage consumption data or weight data to the server 3. Hence any previously non-reported dispensing evens are reported to server 3. Other data may be provided to the server 3 such as the identity of the beverage container 2, the identity ofthe dispensing station 50, and other sensors data. ln the same session, other data may be transferred to the server 3 such as temperature, battery sta- tus, system ok message regarding beverage dispensing station, etc. The server 3 stores the data in the dataset 15. The beverage dispensing station 50 may also, in the communica- tion session, receive information from server 3, such as activation ofthe dispensing pre- vention means 65, software updates etc. The wireless network interface 58 may be acti- vated at this step or in a previous step, for example step in step The method of Fig. 14 may use one or both or none of the optional waiting steps. The wireless network interface 58 may be arranged to go to sleep after a certain time of inac- tivity.

The dispensing events are received by the server 3 and stored in the dataset 15. The server 3 hence can keep track of consumption of beverage. A threshold stored in server 3 or in dispensing station 50 may be used to trigger the automatic delivery of a new bever- age container 2 to the end- user of a certain dispensing station 50. The threshold for deliv- ery may be determined based on shipping time and the speed of consumption. The speed of consumption may be calculated using dispensing events in dataset 15 for the dispensing station 50 or a plurality of dispensing stations 50. For example, the average consumption for a time period may be used.

The beverage dispensing station 50 may have a temperature sensor 59 for detecting the temperature of the beverage container 2 or ambient temperature in the vicinity of the beverage dispensing station 50. The temperature may be used for triggering delivery of a new container 2. For example, if a beverage container 2 with beverage has been subject for high temperature for at time that is longer than a threshold time, this may trigger acti- vation ofthe dispensing prevention means 65 or delivery of a new beverage container Temperature data may be stored in memory 60 of the control circuitry. The dispensingstation 50 may be configured to determine the temperature at least every predetermined time interval such as for example at least every hour, or at least every day. Temperature data may be stored in the memory 60 ofthe dispensing station 50, preferably together with a time point for detection. Temperature readings may also be triggered by a dispens- ing event, first sensor, accelerometer 56 or weight determining means The dispensing station 50 may have a dispensing prevention means 65. The dispensing prevention means 65 prevents, to some extent, the dispensing of beverage from the bev- erage container 2 currently supported or connected to the dispensing station 50. The bev- erage dispensing prevention means 65 is remotely controlled by server 3 via the control circuitry of the dispensing station The dispensing prevention means 65 may provide a warning signal to a user of the dis- pensing station 50. The warning signal may be of any type such as visible, audible or tactile signal. The warning signal may for example be a light on the exterior of the dispensing sta- tion, such as a LED 64. For example, a red light, or a flashing light.

The dispensing prevention means 65 may also lock the dispensing station 50 such that beverage cannot dispensed from it, such as locking the tap 52. For example, when the tap 52 includes a handle 66 the handle 66 may be locked in a closed position.

The dispensing prevention means 65 may be activated contents of a certain beverage con- tainer 2 is unsuitable for consumption, as detected manually or in an automatic manner by system 1. For example, a server user may know that certain batches of beverage have a non-pleasant taste or is contaminated in some manner. System 1 may be able to detect faulty containers 2 with any suitable method, for example the methods describes herein with reference to the machine learning method shown in Fig 17, temperature sensor 59, second tag 18 or error detection logic. For example, the fact that a that a non-authorized container is attached to the dispensing station, for example if the server 3 detects that that a beverage container identity received from a dispensing station 50 is not present inthe dataset 15, or the fact that a second tag 18 (tampering device) indicates that a bever- age container 2 has been tampered with.

I\/|oreover, temperature data from dispensing station 50 may be used to determine that the contents of an individual beverage container 2 is unsuitable for consumption. Temper- ature data from temperature sensor 59 may be reported to server 3 which may use the data to determine that a beverage container 2 should be flagged for non-consumption. For example, ifthe temperature detected by temperature sensor 59 has been above a threshold temperature for a certain threshold time.

Thus, one or a plurality of beverage containers 2 may be flagged for non-consumption in the dataset 15. This may be done manually or automatically as described above. The non- consumption flag may automatically trigger activation of dispensing prevention meansin the beverage dispensing stations 50 where flagged containers 2 are present.

The flag in dataset 15 is detected by server 3 in step 400 of Fig. 15. For example, logic at server 3 may repeatedly scan dataset 15 for flagged containers 2. The server 3 identifies, in the dataset 15, the dispensing station 50 where a flagged container 2 is currently in use (step 401). The server 3 then uses the wireless network interface 58 of the dispensing sta- tion 50 to provide an activation signal to the dispensing prevention means 65 of the dis- pensing station 50, (step 402) which then activates the dispensing prevention means 65 (step 403). The server 3 may have to wait for a communication event from dispensing sta- tion 50 in case the server 3 cannot be able to initiate communication with dispensing sta- tion As discussed above, the beverage container 2 has an identification tag 9 permanently at- tached to it. The permanent tag 9 is also referred to as ”first tag 9” herein. The first tag 9 is preferably an RFID or NFC tag. The first tag 9 stores the identity of the beverage containerln various embodiments, the beverage container 2 may have a second tag 18, which pref- erably is an RFID or an NFC tag. ln a preferred embodiment both first tag 9 and second tags 18 are NFC tags. The second tag 18 is an antitampering device, arranged to detect if a beverage container 2 has been tampered with. Preferably each of the tags 9, 18 comprises at least one NFC antenna. Typically, the second tag 18 is disposable.

The second tag 18 has at least two states and can be attached to a beverage container 2 in such a manner that opening the opening 12 ofthe beverage container 2 puts the second tag 18 in the second state in a non-reversible manner. At least the first state can be de- tected by a tag reader 10. The second state can be a state where the second tag 18 cannot be detected by tag reader 10 or where the signal from second tag 18 is altered in some way in relation to the first state.

The second tag 18 is attached to the opening 12 of a filled beverage container 2 in such a manner that opening the opening 12 of beverage container 2 necessitates altering the state of the second tag 18 in a non-reversible manner. For example, removing the lid 13 necessitates putting the second tag 18 in the second state. The second tag 18 may for ex- ample have an antenna that breaks when the opening 12 is opened, for example when the lid 13 is removed. For example, the antenna is broken in two parts. For example, the sec- ond tag 18 may be in the form of a sticker with an NFC antenna that is placed across the lid 13 and a part of container 2 so that removing the lid 13 necessities breaking the NFC antenna (Fig. 3b).

Regarding the term ”non-reversible” it should be noted that, in some cases, a highly skilled expert with access to the correct tools and equipment may be able to restore the second tag 18 to the fist state, but it is still to be regarded as ”non-reversible” because an average end user is not able to do that.

The second tag 18 may for example be in the form of a sticker or a loop. Part of the sec- ond tag 18 may be made from a polymer material into which an NFC antenna is inte- grated. An example of such a system is the CircusW' tamper loop provided by AveryDennison. The second tag 18 may be of a type that is completely destroyed such that an NFC tag reader 10 does not receive any signal from the second tag 18. Or the second tag 18 may be of a type where the signal is altered but still present in the second state. This may be achieved by the second tag 18 having two antennas, one which is broken in the second state and one which is still functioning in the second state. This increases the relia- bility because the still-functioning antenna serves as a control that, for example, tag reader 10 works as intended.

The second tag 18 may store an identity of the second tag 18. The identity of the second tag 18 may be stored together with the identity of the beverage container 2 (first tag iden- tity) in the dataset 15. Hence any tampering information will be associated to the identity ofthe beverage container 2. For example, when a beverage container 2 is delivered, it is useful to known that the container 2 was delivered in a non-tampered state, and that the specific container 2 is now connected to the beverage dispenser 50. This is particularly useful when a second tag 18 that is completely destroyed (quiet) in the second state be- cause otherwise it is difficult to known that that a container 2 that was delivered in a non- tampered state to the end-user is now connected to the beverage dispenser 50. For ma- chine learning-assisted tracking of beverage containers 2 it is also useful to know if a sub- set of containers (associated with a certain beverage filling station 4, for example) are tampered with.

The second tag 18 may be applied to a container 2 with the use of the method shown in Fig. 16. ln step 700 an empty beverage container 2 is provided. The beverage container 2 has an opening 12 that can be closed as described above. The beverage container 2 has a first tag 9 permanently attached, and the first tag 9 stores an identity of the beverage con- tainer 2. Preferably the container 2 is clean and the ”washed” state. ln step 701 the bever- age container 2 is filled so that it contains at least some beverage. The container 2 is then closed, for example by attaching the lid 13. ln step 702 the second tag 18, in the first state, is attached to the beverage container 2 in such a manner that opening the opening 12 of the beverage container 2 non-reversibly puts the second tag 18 in the second state. This is preferably done in a controlled environment so that no tampering occurs in between. ln step 703 the state of the second tag 18, which typically is now in the first state, is stored in the dataset 15. The state is stored together with the identity of the first NFC tag 9, i.e. the identity ofthe beverage container 2. ln step 704, which may be carried out before, after or simultaneously with step 703, an identity of the second tag 18 is stored, in the dataset 15, together with the identity of the beverage container 2. The beverage container 2 may then be shipped to end-users.

A tag reader 10, for example a portable tag reader used by delivery service may be used to confirm that a non-tampered container is delivered to an end user. The tag reader 10 may provide this information to server 3 which stores the information in the dataset. This pro- vides information that a non-tampered container 2 has been delivered to the end-user. Delivery may be confirmed by the end-user by various means such as signature, for exam- ple an electronic signature. lf the second tag 18 is the second state, a new containermay be automatically shipped to the end user by server The end user may then install the container 2 at the dispensing station 50, such that the tag reader 10 of the dispensing station can read the first tag 9 and the second tag 18. The tag reader 10 of the dispensing station 50 may then read the identity of the first tag 9 and confirm that the second tag 18 is in the second state, and report this to the server 3. The tag reader 10 of the dispensing station 50 may read the second tag 18 and store the state in the memory 60 ofthe dispensing station 50. The tag reader 10 of dispensing station 50 may store the identity of the first tag 9 and the identity of the second tag 18 (if second tag 18 has an identity) and store these together with the state (first/second) in memory 60. lf the second tag is not detected and the second tag 18 is of the type that is automatically destroyed by the tag reader 10, the dispensing station may store information that the sec- ond tag 18 is in the second state if no tag is detected. Hence, the tag reader 10 attempts to read second tag 18 but does not get a reply and thereby assumes that second tag 18 is in the second state.

The method may comprise the step, if a tampered container 2 has been delivered, to flag the container 2 for non-use in the dataset 15. Hence there may be logic in server 3 thatdetermines that allows only containers 2 that have been delivered in a non-tempered state (in the first state) to be associated with dispensing station 50. lf such a container 2 is present at a dispensing station 50 the server 3 will receive the identity as described else- where, as a dispensing station 50 reports the identity of the beverage container 2 to the Se FVeF. lf a server 3 detects that tampered container 2 is present at a dispensing station 50, the server 3 may be configured to take different actions. For example, the dispensing preven- tion means 65 may be activated. A message may be sent to the end-user about the tam- pering, for example that the end-user should not consume the contents.

Tag reader 10 of dispensing station 50 is preferably able to read the second tag 18 in the same manner as it can read the identification tag 9. For example, the embodiments shown in Figs. 10-11 may be used for the second tag 18 as well as the first tag Server 3 may comprise or be in digital communication with machine learning service 20. I\/|achine learning service 20 may apply machine learning to the dataset 15 to obtain a de- cision rule. I\/|achine learning service 20 is able to detect patterns to produce a decision rule that is useful for managing the system 1, for example with regard to: a) demand for beverage, b) demand of logistics, i.e., distribution of beverage containers 2, c) errors or bottlenecks in system 1 or various components of system 1, e) determine end of life for beverage containers and f) detecting beverage containers 2 that should not be consumed by an end-user. A user of server 3 may use machine learning to forecast behaviour of sys- tem 1 and their (sub)-users.

There are many possibilities. Here are a few hypothetical examples: 1. There is always a surge in beverage demand on Wednesdays. This can be used to increase production of filled beverage containers 2 beforehand. 2. Beverage containers keep breaking down at one certain filling station 4. This infor- mation can be used to visit this filling station 4 to take care ofthe problem.3. A certain user pattern for beverage containers 2, for example number of filling cy- cles, length of transportation, or other parameter, makes them break earlier than other containers. Beverage containers 2 that fulfil these criteria, for example a number of filling cycles, or total length of transport, may be detected and taken out of use before they break. 4. One individual beverage container 2 departs from a pattern. This may indicate that the container 2 has been tampered with or is a fake container, provided by a non- authorized source. For example, a fake container ID has been created or the ID of a ”true” container 2 has been cloned and the container 2 has then been inserted in the system 1, for example been attached to a beverage dispenser 50. Hence, the decision rule may also be used for detecting fraudulent behaviour.

Parameters that may be useful and to be included in the data set 15 include: number of filling cycles, length of transportation, length of time that container 2 spends at an end- user, date of manufacture of beverage container 2, production batch number of beverage container 2, time spent in storage, data from sensors (such as weight determining means 55 or accelerometer 56) of dispensing stations 50, manually or automatic recorded errors or failures of containers 2 and dispensing stations 50, means of transportation and combi- nations ofthese. Any data in dataset 15 may be used. Information for machine learning may also be provided from external information providers, for example information about weather, temperature at end-users, or during transport or during storage, humidity etc.

These are just non-limiting examples, as there are many possibilities. ln particular the decision rule may be used to determine a statistical model and to detect an behaviour of an individual beverage container 2 that departs from the statistical model.

This may indicate that the beverage container 2 has been tampered with, that the identity ofthe beverage container 2 is fake, or that an identity of a beverage container 2 has been cloned or stolen.

The following are hypothetical examples used to illustrate how machine learning may be used to identify strange behaving beverage containers 50:- Abnormal shipping speeds for beverage container 2 to dispensing station 50 (too short times may indicate multiple IDS, too long time may indicate that the bever- age dispenser has been tampered with) - Beverage container 2 is absent from system for a prolonged time (may indicate tampering) - Abnormal accelerometer data (may indicate tampering) However, it should be kept in mind that the decision rule may be considered as a ”black box” and it is difficult and maybe not even meaningful to speculate in which parameters are predictive for indicating tampering or other faults in beverage containers I\/|achine learning predictions are improved when the dataset 15 comprises information about a large number of beverage containers 2 that have been accumulated over time. I\/|achine learning is improved when the containers 2 are similar or identical, or at least comparable. The same is true of filling stations 4 or dispensing stations 50. I\/|achine learn- ing may be provided from a machine learning service 20 such as Amazon or may be pro- vided in the server 3. Any useful type of machine learning may be used, such as for exam- ple machine learning implementing Bayes' theorem.

Applying machine learning may result in a statistical model for behaviour for the plurality of beverage containers. The decision rule may be that any container 2 that departs from the behaviour should be taken out of service. There may be a threshold for how much a container 2 may depart from the statistical model before being taken out of service.

With reference to Fig. 17 a method may comprise the steps of step 800: applying machine learning to the dataset 15 to obtain a decision rule. The decision rule can be updated by applying machine learning at any useful interval, such as for example daily, weekly or monthly. ln step 801 the decision rule is applied to the dataset 15. This is also carried out at any suitable time interval but typically more often than the decision rule is updated, such as hourly or daily. The decision rule may be applied to a part of dataset 15. For ex- ample, machine learning may be applied on data for beverage containers 2 that has beentaken out of use whereas the decision rule is applied to beverage containers 2 that are currently in service only. The dataset 15 is typically updated often, preferably in real time. So, the dataset 15 will change over time. ln step 802 a decision is taken. lt may be a deci- sion with regard to an individual beverage container 2, for example that a certain bever- age container 2 should be taken out of service, or that the contents ofthe beverage con- tainer 2 should not be consumed. An individual container 2 may be flagged in the dataset for taking out of service. This may trigger manual alarm at the server 3 to a server user.

This may also trigger activation ofthe dispensing prevention means ln the same manner machine learning may be applied to the dataset 15 to predict errors of beverage dispensing stations System 1 may comprise one or more beverage fillings stations 4 for filling beverage con- tainers with beverage. The beverage filling station 4 may for example be located in a suita- ble building. However, in certain embodiments, beverage filling station 4 may wholly or partly be located in a filling station container 21. This provides for fast expansion of the system, where one or more filling station containers 21 fitted with filling stations 4 are de- ployed.

The filling station 4 may comprise a beverage production unit. The beverage production unit may be computer controlled such that a computer 19 controls tanks, dispensers, valves, pumps, heaters, and coolers in a predetermined sequence to provide an automatic brewing or beverage production process. Computer 19 may also control a cleaning process when the beverage has been prepared or on a predetermined schedule (Clean-in-place).

The beverage production unit may be adapted to provide any type of beverage such as po- table water, beer or soda, or cider. ln one preferred embodiment the beverage is beer. ln one preferred embodiment the beverage is potable water.

The beverage production unit may comprise water purification unit 5. The water purifica- tion unit 5 may use any suitable technology such as filters, heating, UV irradiation or addition of chemicals such as chlorine. The beverage production unit may also comprise a desalination unit for removing salt from the water.

The beverage production unit may also provide a unit for adding various ingredients such as flavouring, such as syrup for providing a soda, and carbon dioxide. Beverage may be pro- duced continuously or in a batch process and stored in tank Beverage production unit preferably has at least one sensor for sensing for example, flow, temperature, pressure, or level in the beverage production unit. ln general, the sensor may be able to provide data to computer 19 and to server 3. Sensor data and production data may be stored in dataset 15. Computer 19 of beverage filling station 4 may be connected to a data network communication device which enables data communication with server 3. Computer 19 may also comprise memory, processor and bus. Computer 19 is preferably connected to tag reader 10 of beverage filling station The memory of computer 19 of the beverage filling station 4 may have at least one set of instructions in digital form for producing a beverage. The computer 19 may have access to a plurality of instructions, one for each type of beverage, such as different types of beer. The digital instructions (recipes) may be provided from server 3 to computer 19. Computer 19 may have a user interface, which a user may use to, for example, start or stop brewing procedures. The user interface may comprise a display and input means, such as a keyboard, a mouse, joystick, touchscreen, or the like. Alternatively, a user of beverage filling station 4 may be able to use a mobile phone as interface for example by connecting to computervia server 3 or directly connecting to computer Computer 19 may be able to receive an order from server 3 to produce beverage, for exam- ple if server 3 detects or predicts a demand for beverage. Hence computer 19 may be able to activate the beverage production unit, for example to produce beer or potable water. ln a preferred embodiment the beverage filling station 4 is arranged such that a tag reader is arranged in relation to the washing station 7 to detect the tag 9 of a beveragecontainer 2 only when the beverage container 2 has passed through the washing station 7. For example, the tag reader 10 is placed in close proximity to the exit of the washing sta- tion 7. This provides automatic tracking of washed containers 2 and makes sure that only washed containers 2 are flagged with the ”washed” state.

With reference to Fig. 18 the method may comprise the steps of 900, detecting the bever- age container 2, step 901 the tag reader 10 providing the identity ofthe beverage container 2 to the server 3. This is done via computer 19 which is in network communication with server 3. ln step 902 the server 3 sets the state of the beverage container 2 to ”washed” in the dataset 15. The server 3 may set the state to ”washed” because the tag reader 10 is associated with this state at sever 3 such that an identity of the tag reader 10 is associated with the washed state in the dataset 15. Or, the ability to set the washed state is stored with the tag reader 10, which provides a washed state message to the server 3 together with the identity of the beverage container Similar rules may be applied anywhere in the system where they are useful and may for example involve two or more tag readers ln an even more preferred embodiment, system 1, in particular filling station 4 provides at least three tag readers 10: one that provides an ”unwashed” state (or ”entered station” state), one that provides ”washed” state, and one that provides a third state that may be ”filled/ready to deliver” state.

As mentioned, server 3 may have logic that detects errors such as erroneous transfer from ”unwashed” state to the ”filled” state without passing ”washed” state. When an error has occurred, an individual beverage container 2 may be flagged in dataset 15 to be taken out of service or to be reprocessed, for example to be taken back to beverage filling station 4 to be washed and filled again. Also, beverage dispensing prevention means 65 of dispensing station 50 may be activated ifthe containers 2somehow is delivered to a dispensing stationThe filling station 4 may be contained in a filling station container 21. The filling station container 21 is preferably ofa size such that the beverage filling station 4 can be transported using a truck or trailer, or similar. Hence the filling station container 21 may be movable. For example, it may have the dimensions of a prefabricated shipping container. The filling station container 21 may have the dimensions of a standard intermodal shipping container. Standard ISO shipping containers are 8ft (2.43m) wide, 8.5ft (2.59m) high and come in two lengths; 20ft (6.06m) and 40ft (12.2m). The filling station container 21 may be made of steel or other strong material such as aluminium, glass fiber, plastic, plywood, or similar material. The filling station container 21 is preferably self-supporting. Filling station container 21 may have any suitable shape where a rectangular block shape or approximately such a shape is preferred. The rectangular block may have dimension that are approximately: height 2-3 m, width 2-3 m and length 3-12 m. lt may be possible for a user to enter into the filling station container 21, for example by walking into the filling station container, in order to service the beverage preparation unit or to make input into computer The filling station container 21 may be adapted to be placed on the ground or on a floor. The filling station container 21 may have an essentially flat underside for placing the filling station container 21 on flat ground or on a floor, but filing station container 21 may be provided with feet or rails for placing on an uneven surface.

Filling station container 21 may have a suitable arrangement for lifting the filling station container 21 by a crane or otherwise moving the filling station container 21 Hence, the filling station container 21 can be transported to the intended location and easily become con- nected and provide beverage in a short time.

Filling station container 21 is preferably suitable for being placed outdoors. Hence the filling station container 21 may have suitable proofing against one or more ofthe following: heat, cold temperatures, rain, lightning, snow or strong winds, in order to protect the various components of the filling station container 21 or provide a controlled environment for bev- erage production. The walls (including roof and bottom) of filling station container 21 may have thermal insulation. The walls may be rainproof and windproof. The bottom of thefilling station container 21 is suitable for being placed on the ground and may be proofed against rain and water seepage. The walls of the filling station container 21 may have a total thickness, including insulation, of at least 2, cm more preferably at least 4 cm and most preferably at least 6 cm. ln one embodiment the filling station container 21 is a regular intermodal shipping container provided with internal thermal insulation. Filling station con- tainer 21 may be provided with arrangements for proving a controlled temperature and humidity inside the filling station container 21, such as a heater, cooler, air conditioner, humidifier or dehumidifier. The computer may be arranged to control the internal condi- tions of filling station container 21 and keep for example temperature or humidity within a predetermined range.

The filling station container 21 is preferably connectable to external power supply, a water supply and sewage and may preferably be provided with suitable pipes and wiring to enable easy connectivity. Walls of filling station container 21 may have suitable bushings for vari- ous connections or ports for various utilities such as electricity, water, sewage and data communication. The filling station container 21 may have a pump for providing suction from a water source such as, the sea or a lake or a river to provide water. Walls of filling station container 21 may have suitable bushings for various connections or ports for various utilities such as electricity, water, sewage, and data communication.

The beverage filling station container 21 is preferably connectable to external power supply, a water supply and sewage and may preferably be provided with suitable pipes and wiring to enable easy connectivity.

Filling station container 21 may in particular comprise water purification unit 5, beverage production unit, beverage tank 6, one or more tag readers 10, computer 19 and washing station lt is understood that the present methods, systems and devices are partly computer-imple- mented, using digital computer equipment. The various embodiments and components de- scribed herein such as server 3, beverage dispensing station 50, tag reader 10, filling station4, computer 19, machine learning service 20 and communication between these compo- nents uses digital computer technology for storing and handling digital information and sig- nals as well as suitable hardware and software, including for example suitable digital pro- cessors, digital memories, input means, output means, buses and communications inter- faces. A user may be able to make input using for example a keyboard, a mouse or a touch screen. Output may be provided on for example a display.

The various components, such as server 3 and computer 19 may each have an operating system. The server 3 may have a user interface that a server user can use to add new bev- erage containers 2, analyse data etc. With reference to Fig. 19 each of tag reader 10, server 3, computer 19 and machine learning service 20 may comprise control circuitry comprising a memory 80, a processor 81 a bus 82 and a communication interface Server 3 may be one physical server or may be a virtual server. Function of sever 3 may hence be distributed across several physical entities. Data, such as dataset of dataset 15 may be stored in a datastore or in server 3. Dataset 15 may be stored in a distributed data- base comprising several nodes.

The methods herein can be implemented any suitable combination of software and hard- ware. Any suitable programming language may be used for the software units and methods described. Data communication in system 1 may be implemented using suitable networking technologies and protocols, inducing cellular communication such as 3G, 4G and 5G, LoRa, Wi-Fi or Bluetooth, or Ethernet. Data communication can be wireless, or wire bound. Infor- mation may be exchanged over a wide area net such as internet 17. Data communication in system 1 may be encrypted.

The identity of the each of the beverage containers 2 and the dispensing stations 50 may comprise any suitable combination of numbers, letters or other symbols suitable for digital data processing. Each of filling station 4 in system 1 may also have an identity.

Communication in system 1 and update of dataset 15 may be carried out using any suitable schedule. Tag readers 10 and dispensing station 50 may provide information to rest of sys- tem 1 in particular sever 3, for example immediately when a beverage container 2 is de- tected by tag reader 10 or when a sensor of dispensing station 50 detects a value. Commu- nication sessions may alternatively be scheduled to be carried out at a suitable interval, for example at least every predetermined interval such as at least every second, at least every minute, at least every hour or every day. lt is realized that everything which has been described in connection to one embodiment is fully applicable to other embodiments, as compatible. Hence, the invention is not limited to the described embodiments, but can be varied within the scope of the enclosed claims. While the invention has been described with reference to specific exemplary embodiments, the description is in general only intended to illustrate the inventive concept and should not be taken as limiting the scope of the invention. The invention is generally defined by the claims.

Claims

Claims A beverage distribution system (1) comprising: a plurality of reusable and portable beverage containers (2) where each ofthe plurality of beverage containers (2) carries a unique identification tag (9), the system further comprising a plurality of identification tag readers (10), where at least some tag readers (10) are located at different geographical locations, the tag readers (10) being configured to read the identification tags (9), said tag readers (10) being connected to a server (3), said server (3) being configured to store a state of each ofthe plurality of beverage containers (2) in a digitally stored dataset (15), where the state is that the beverage container (2) is present at one ofthe tag readers (10), the data set (15) further comprising data about manually or automatic recorded errors or failures of beverage containers (2) where the beverage distribution system (1) is configured to apply machine learning to the data set (15) comprising the presence ofthe plurality of beverage containers at one of the tag readers (2) and the errors and failures of the beverage containers (2) to produce a decision rule that comprises a statistical model for the behaviour of the plurality of beverage containers (2) in the beverage distribution system (1), for making a decision about taking an individual beverage container (2) out of service, where the decision rule comprises a behaviour pattern for the plurality of beverage containers (2) and where the decision rule is arranged to determine an end-of life for the beverage container (2), where system is configured to apply the decision rule at least every predetermined time period and where a beverage container (2) that departs from the behaviour pattern is flagged in the dataset (15) for being taken out of service. The beverage distribution system (1) of claim 1 where the state is stored together with a time point for reading the identification tag (9). . The beverage distribution system (1) of any one of any one of claims 1 or 2 further comprising at least one beverage dispensing station (50) comprising one of the identification tag readers (10), said beverage dispensing station (50) being adapted to connect to the beverage container (2) such that beverage can be dispensed from said beverage container (2), where the dispensing station (50) is configured to use the tag reader(10) to detect that a beverage container (2) is present at said beverage dispensing station (50) . The beverage distribution system (1) according to c|aim 3 where the beverage dispensing station (50) has a dispensing prevention means (65), and the server (3) is configured to activate the dispensing prevention means (65) when a beverage container (2) is flagged for taking out service. . The beverage distribution system (1) of any one of claims 3 or 4 where said beverage dispensing station (50) has a sensor the dispensing station (50) being configured to co||ect data from the sensor and provide it, using a wireless data connection, to the server (3), the server (3) further being configured to include the sensor data in the dataset (15). . The beverage distribution system (1) of c|aim 5 where the sensor is a weight determining means (55) or an accelerometer (56). . The beverage distribution system (1) according to any one of claims 1 to 6 where the beverage distribution system (1) comprises a beverage fi||ing station (4) and at least one identification tag reader (10) ofthe beverage distribution system (1) is comprised in the beverage fi||ing station (4). A method in beverage distribution system (1) said beverage distribution system (1) comprising a plurality of reusable and portable beverage containers (2) where each ofthe plurality of beverage containers (2) carries a unique identification tag (9), the beverage distribution system (1) further comprising a plurality of identification tag readers (10), where at least some tag readers (10) are located at different geographical locations, the tag readers (10) being configured to read the identification tags (9), said tag readers being connected to a server (3), said server (3) being configured to store a state of each ofthe plurality of beverage containers (2) in a digitally stored dataset (15), where the state is that the beverage container (2) is present at one of the tag readers the dataset (15) further comprising data about manually or automatic recorded errors or failures of beverage containers (2), the method comprising a) applying machine learning to the data set (15) comprising the presence of the plurality of beverage containers at one ofthe tag readers (2) and the errors and failures of the beverage containers (2) to produce a decision rule, where the decision rule is configured to make a decision about taking an individual beverage container (2) out of service, where the decision rule is comprises a statistical model for the behaviour of the plurality of beverage containers (2) in the system, and is arranged to determine an end-of life for a beverage container, b) applying the decision rule to the plurality of beverage containers (2), where the decision rule is applied at least every predetermined time period, where a beverage container (2) that departs from the behaviour pattern is flagged in the dataset (15) for being taken out of service.