Method for operating an overhead conveyor for consecutive hanging transport bags and correspondingly equipped warehouse

Abstract

Method for operating an overhead conveyor for successive hanging transport bags, wherein hanging transport bags filled with goods can be conveyed hanging on the overhead conveyor via hanging adapters and the distance between the hanging transport bags can be adjusted by the distance of an engagement of the hanging adapters in the overhead conveyor, which distance is variable in the grid, and in which the hanging transport bags are placed one after the other in an infeed on the overhead conveyor while hanging on hanging adapters, wherein in order to optimize the distance, the distance between the hanging transport bags when being inducted onto the overhead conveyor is set in a distance-optimized manner based on the thickness of the hanging transport bags in the transport direction, for which purpose the thickness of the hanging transport bags is calculated in advance based on the goods, and accordingly equipped warehouse.

Description

[0001] The invention relates to a method for operating an overhead conveyor for consecutive hanging transport bags, wherein hanging transport bags filled with goods can be conveyed suspended on the overhead conveyor via hanging adapters and the distance between the hanging transport bags can be adjusted by the grid-variable distance of an engagement of the hanging adapters in the overhead conveyor, according to claim 1 or a correspondingly equipped warehouse according to claim 12.

[0002] These are so-called overhead conveyors for hanging goods, which are picked up and transported either hanging or in pockets, bags, etc., which engage with adapters and are then moved via overhead conveyor chains. The pockets have flexible walls and can therefore accommodate a wide range of goods.

[0003] From EP 2 789 555 A1, for example, a device for the order-oriented provision of individual goods for a plurality of orders from a warehouse is known, which device comprises at least one intermediate storage unit connected to a warehouse for the intermediate storage of individual goods of at least one order, a collection area connected to the at least one intermediate storage unit for collecting the individual goods of the at least one completed order, and a separation area having a plurality of delivery lines for the order-oriented provision of the individual goods of the at least one completed order.

[0004] In such a system, the fixed spacing of the adapters and the fixed spacing of the flights of the drive chain determine the distance of the transport of successive hanging goods or bags, regardless of how wide the goods are (or their thickness) in the conveying direction and where the goods are to be transported in the system.

[0005] The buffer capacity and performance (i.e., throughput) of such a sorting system are determined by the bag thicknesses and the cumulative effect of these bag thicknesses. Regarding buffer capacity, thinner and more densely packed bags result in a buffer with better space utilization. Regarding throughput, thinner and more densely packed bags result in higher throughput, as more bags can be moved per specified chain length. A better understanding of bag thicknesses therefore allows for optimization of space utilization and how closely bags can be placed next to each other on the sorting chain to improve performance.

[0006] DE 10 2014 201 301 A1 discloses a conveyor device for the automated conveying of individual goods along a conveying direction, wherein the conveyor device is designed as an overhead conveyor device, comprising a plurality of adapters, each having an adapter identification means for receiving the individual goods, wherein each individual item has an individual goods transponder, a transponder reading unit for reading individual goods data from the individual goods transponder, an adapter data reading unit for reading adapter data from the adapter identification means, a control unit which is in signal connection with the transponder reading unit and with the adapter data reading unit for linking the individual goods data read by means of the transponder reading unit and the adapter data read by means of the adapter data reading unit.Here, we work with predefined data to determine the distance between the consecutive adapters in order to enable the reading of the adapter data and the product data.

[0007] DE10 2020 214 613 A1, for example, describes a device for placing transport pockets into a conveyor system of an overhead conveyor system. The device comprises a conveyor rail that can be coupled to the conveyor system and along which the transport pockets can each be conveyed suspended by means of a suspended adapter; a detection unit for detecting the number of suspended adapters arranged on the conveyor rail within a detection range; a control unit that is in signal communication with the detection unit and is configured to variably define a transport distance for conveying the transport pockets by means of the conveyor system; and a separation unit that is in signal communication with the control unit for individually placing the suspended adapters into the conveyor system at the variably defined transport distance. Thus, it is not the thickness of the pockets that is determined, but the number of corresponding suspended adapters per section.

[0008] DE 10 2019 215 304 B3, on the other hand, discloses an analogous system in which a detection unit for detecting the thickness of the individual overhead goods is present in the loading device. The thickness thus determined is used to load the overhead goods with a variable transport distance. DE 10 2019 215 304 B3 assumes that the thickness of the overhead goods is unknown, so each pocket is measured directly upon loading.

[0009] In contrast, the object of the present invention is to create an improved possibility in which the distance of the transport of successive hanging goods or bags can be adjusted in a space-optimized manner and can be taken into account more precisely.

[0010] This object is achieved by the method recited in claim 1 and the warehouse recited in claim 12. Advantageous embodiments emerge from the subclaims and the description.

[0011] According to the invention, it has been recognized that if a predictive model is used to determine the pocket thickness, it is possible to dispense with a single measurement of the pockets during feeding and thus also to dispense with sensors, etc., which simplifies and speeds up the feeding devices.

[0012] The prediction model uses, on the one hand, known data of the goods and, on the other hand, empirical data from a measuring station in which the goods are loaded into a hanging bag under real conditions and the closed hanging bag is then measured.

[0013] In the method according to the invention, for operating an overhead conveyor, the distance between the overhead transport pockets when feeding them onto the overhead conveyor is adjusted in an optimized manner based on the thickness of the overhead transport pockets in the transport direction, for which purpose the thickness of the overhead transport pockets is calculated in advance based on the goods.

[0014] The forecast can be made based on data linked to the goods (identity) of the filled hanging transport bag.

[0015] Such linked data may include properties of the product itself and properties of the hanging bag. Preferably, the properties of the product may include dimensions, shape or geometry, center of gravity, elasticity, softness, contents and packaging type, and orientation.

[0016] Preferably, the properties of the hanging transport bag include its thickness in the transport direction, which is detected by sensors after it has been filled with goods.

[0017] In a preferred embodiment, the data of the goods are linked with those of the hanging transport bag in a model for forecasting.

[0018] Such a model can combine the thickness data of the hanging transport bag with the dimensional data of the goods using a regression model technique or a classification model technique in order to predict the thickness of the hanging transport bag filled with the goods within a probability corridor.

[0019] In this case, the thickness of the suspended transport pocket filled with goods refers to the maximum dimension of the suspended transport pocket in the conveying direction. This represents the minimum distance to the previous or next suspended transport pocket. Based on this minimum distance, the optimized spacing in the overhead conveyor can be set during infeed.

[0020] The model can also use a regression model technique or a classification model technique to link the thickness data of the hanging transport bag with the center of gravity and orientation data of the goods in order to predict the orientation of the goods when filling the hanging transport bag within a probability corridor.

[0021] In this context, orientation refers to the orientation of an elongated item in a hanging transport bag. For example, a shoe box can be oriented lengthwise, crosswise, or vertically when filled in the hanging transport bag.

[0022] The model can also use a Bayesian modeling technique to link the thickness data of the hanging transport bag with the softness data of the product in order to predict the compression of the product when filling the hanging transport bag within a probability corridor.

[0023] In this case, softness data refers to the product's tendency to crumple or compress in the hanging transport bag, thus assuming a different degree of expansion. For example, a sweater may be oriented lengthwise, crosswise, or vertically while being filled in the hanging transport bag and may be compressed by gravity, so that the expansion in the transport direction is greater and thus contributes more to the thickness.

[0024] Preferably, the thickness data of the hanging transport bag are recorded in the closed and / or open state. The hanging transport bag used according to the invention has a first bag wall, a second bag wall and a hanging device arranged at an upper end of one of the bag walls. At a lower end of the bag walls, remote from the hanging device, there are arranged closing elements that can be reversibly coupled to one another and with which the lower ends of the bag walls can be connected to one another to form a bag base that closes the transport bag at the bottom. The closing elements are designed as locking strips that are aligned parallel to one another and can be locked together. The locking strips can be moved from a transport position, which closes the transport bag at the lower end and is locked together, by relative rotation against one another about the longitudinal axis of the locking strips into an unlocking position in which the locking strips can be detached from one another to open the transport bag at the lower end. In other words, the locking bars are designed such that they can be disengaged from the transport position, which closes the hanging transport bag at the lower end and is locked to one another, by rotating them against each other around the longitudinal axis of the locking bars, in order to release the locking bars from one another and open the transport bag at the lower end. Such a transport bag can be opened and reclosed downwards in a particularly simple manner. This also results in a simple unloading process for the hanging transport bag. Such hanging transport bags are known, for example, from EP 3 442 884 A1.

[0025] To load the hanging transport bags, they are opened or held open at the top, and the goods are placed into the hanging transport bag, usually one at a time. The goods can fall from a conveyor belt into the opened hanging transport bag in the loading station. The hanging transport bag is then closed or closes itself under the weight of the goods. Such loading stations are known, for example, from EP 3 442 884 A1.

[0026] Such hanging transport bags can also have a closure at the top. This is what the specification for recording thickness data in the closed and / or open state refers to.

[0027] The model can also use the classification model technique to link the thickness data of the hanging transport bag filled with the goods with the identity of the goods in such a way that the goods are classified according to a higher-level property of their identity.

[0028] In this case, classifiable grouping means that the goods are grouped at a higher level based on their identity (e.g., SKU) and this group is linked to certain thickness properties. For example, a "Nike 2025 basketball shoe box" can be classified into a "shoe box" group. This group can be used by the model, for example, to predict the thickness of a hanging transport bag when filled with a shoe box.

[0029] The invention also relates to a correspondingly equipped warehouse, i.e. a warehouse with an overhead conveyor device for the automated conveying of transport bags loaded with at least one item of goods along the overhead conveyor device, with a controller for controlling the warehouse and with a bag loading station for loading a transport bag with at least one item of goods and subsequently feeding the loaded transport bag into the overhead conveyor device, wherein the distance between the overhead transport bags can be adjusted by the grid-variable distance of an engagement of the overhead adapters in the overhead conveyor. The warehouse further comprises a measuring device which is connected to the controller for the exchange of data. The measuring device is set up to generate data, wherein the data is transmitted by means of sensors and, if applicable,manually controlled examinations comprise certain properties of the at least one product and the data comprise the thickness determined by means of sensors in the transport direction of a closed transport bag filled with the at least one product, wherein the control is programmed in a computer to carry out the above method, to calculate an expected transport bag thickness for a specific product from the data, to control the bag loading station with the expected transport bag thickness in such a way that the distance when feeding in the loaded transport bags is set based on the expected transport bag thickness.

[0030] The controller may be configured to create a data set from the data historically generated by the surveying device for statistical analysis by the controller to determine the expected transport pocket thickness without prior surveying.

[0031] It is preferred if the measuring device is identical to the at least one bag loading station with regard to bag loading. In this context, "identical" means that the stations are mechanically identical, but may be configured differently with regard to sensors, etc., depending on the requirements of the application. Thus, the measuring device will be more elaborately designed with regard to sensors in order to implement the purpose of data collection. It is understood that the measuring device and the bag loading station can also be configured differently.

[0032] In other words, the invention enables a better prediction of the pocket thicknesses or the thickness spectrum, thus achieving better space utilization and increased performance of the overhead conveyor and the entire warehouse system.

[0033] According to the invention, a model is created that incorporates both the host data (data about the goods in a central database on a server) of the goods as well as information from targeted measurements (e.g., dimensions) and recordings (e.g., photos) from a measuring device, which are obtained when filling hanging transport bags. This allows the thickness of the filled hanging transport bags to be linked to the properties of the goods.

[0034] The model can then be used to classify and group goods (e.g., distinguishing between a pencil in a case or a folded T-shirt wrapped in plastic) and estimate their softness (compressibility) (e.g., sweaters, high compressibility factor).

[0035] Based on the goods data from the host, the information from the measuring device and the model results, the thickness of the filled hanging transport bags can be predicted as soon as the goods are filled into the hanging transport bag.

[0036] The measurements, recordings, model predictions and validation measurements are stored in a database so that accuracy can be improved as data collection increases.

[0037] Preferably, the database is designed to be self-learning. In this case, "self-learning" means that the accuracy of the model is continuously improved based on new data. To this end, the described regression modeling technique, classification modeling technique, and / or Bayesian modeling technique can link the thickness data of the hanging bag with the property data of the product in order to predict the orientation, compression, etc. of the product when filling the hanging bag within a probability corridor, thus ultimately predicting the thickness.

[0038] The data can also be fed into a neural network or mapped to the input and hidden layers to calculate the thickness of a hanging bag filled with a specific product. The accuracy of the calculation can be validated by empirical measurements on the measuring device and used as a feedback loop to improve or train the neural network.

[0039] Training is carried out using algorithms and based on data from the so-called measuring device. This device records all packaging-relevant information such as dimensions, orientation, compression factor, contents, and packaging type. Information provided by modern image processing systems is also taken into account in real time. Different criteria can be weighted according to customer requirements.

[0040] The measuring device is equipped with sensors and user interfaces and is used to collect data on goods and validate the model. In principle, only one measuring device will be required per warehouse. If sufficient data is available and sufficient, a measuring device may even be omitted. Goods for which there is low confidence (high probability corridor) in the resulting hanging bag thickness or the goods classification are routed to this station.

[0041] Once confidence in the calculation of the resulting hanging bag thickness has improved (small probability corridor), the measuring device becomes less important and can be used less frequently to ensure the quality of the prediction model. Once confidence in the thickness prediction for all goods is high, it is possible to calculate the thickness prediction entirely without the measuring device. The measuring device can then only be used for new goods.

[0042] An example procedure at the surveying facility can be carried out as follows to collect data and input it into the model: 1. First, an item is taken from a source container and the barcode of the item is scanned, thus retrieving the host data based on the identity of the item. 2. Measurements are then taken and sensor data is generated to confirm or, if necessary, correct the host data. The following measurements are possible: a. Photos from a 2D camera; b. (Optional) Photos from one 3D camera or multiple 2D cameras; c. (Optional) Dimensional measurements using laser, lidar or light curtains; d. (Optional) Weighing and determining the center of gravity; e. (Optional) Manually or automatically operated mechanical guides for measuring softness and compression; f. (Optional) operator input via a computer user interface;

[0043] A robot can also be used to rotate and move the goods to take the necessary measurements. 3. The goods are then grouped and classified, and a clustering model is applied based on the host data, photos and measurements, which are analyzed and used to classify the goods into a cluster group and to calculate the consistency (e.g. softness) of the goods. 4. An operator can optionally perform a review of the model prediction by providing inputs on a user interface that allows them to correct the prediction if necessary. 5. (Optional) Orientation prediction model: Using information from the host data, photos, measurements and the results of the goods classification and clustering model, this model calculates how the goods fall into the hanging bags, that is, the orientation of the goods when filling the hanging bags. 6. Thickness prediction model: Using information from host data, photos, measurements, results of the commodity classification and clustering model, and (optionally) results of the orientation prediction model, this model calculates the thickness and / or thickness range of the hanging bag after it has been filled with the commodity in question. 7. Release of goods: The goods are released and filled into the hanging transport bag. 8. (Verification Option A) The thickness of the opened hanging bag filled with goods can now be verified in a verification measurement before closing. Cameras, laser measurements, or mechanical guides can be used to determine the resulting thickness of the hanging bag and, optionally, the orientation of the goods inside the hanging bag. These measurements are used to validate the orientation prediction and thickness prediction models. 9. The hanging transport bag is then closed or closes automatically. 10. (Verification Option B) The thickness of the closed hanging bag filled with goods can now be verified in a verification measurement after closing (analogous to the above). These measurements are used to validate the orientation prediction and thickness prediction models. Typically, at least verification option A will be selected. Both can also be selected. 11. The hanging transport bag filled with the goods is then fed in and then transported to buffering and sorting. 12. Data storage and model updating take place after or in parallel with each step. For this purpose, all measurements, photos, and model predictions are stored in a database. The data, together with the prediction and validation data, are used to update the models. By constantly repeating the steps and monitoring the alignment of the goods and the resulting measurements while taking the parameters into account, the accuracy of the models improves over time. Therefore, in principle, only one master measuring device is required, running in parallel with the other normal pocket loading stations in a sorting stream. The host data is enriched with new and / or updated information.

[0044] The invention addresses the problem that it is difficult to measure the size of goods in hanging transport bags from the outside to obtain a thickness parameter, as hanging transport bags are very soft and pliable. The measured data is therefore inaccurate, especially for small goods, and often incorrectly evaluated.

[0045] The inventive evaluation of goods using models (possibly using AI) has several advantages: - No separate measuring station required in the system (cost savings); - No resulting performance bottleneck due to a measuring station; - It is not necessary to measure each bag repeatedly, the system is able to learn the goods; - Data is collected across multiple projects / systems and product clusters are getting better and better; - The thickness of the hanging transport pockets and thus the required length of the order buffer can be predicted before the order is executed; - Order lead times can be more accurately predicted before the order is executed. Scheduling decisions can be made to ensure customer priorities are met. - Better decision model as to whether the order needs to be split up into several stations or can be packed at one packing station also allows a decision as to whether one or more boxes need to be erected, leading to greater sustainability.

[0046] Further details of the invention emerge from the following description of embodiments with reference to the drawing, in which Fig. 1 a schematic view of the system for processing a plurality of picking orders according to the invention; Fig. 2 a more detailed view of the retrieval area from a circulating storage system; Fig. 3 is a flow chart of the material flow in the system according to the invention; Fig. 4 a schematic perspective view of a loading station or measuring device according to the invention; Fig. 5 is a flowchart of the steps on a measuring device according to the invention; Fig. 6 is a flowchart of the steps at a loading station according to the invention; Fig. 7 a flowchart of a model creation; Fig. 8 a flowchart of classification and clustering; Fig. 9 shows a flowchart of an alignment prediction, a thickness prediction and a KPI prediction.

[0047] The figures show a system, designated overall by 1, for processing a large number of picking orders with several goods.

[0048] The system generally comprises a storage area A with an automatic container replenishment warehouse 2 and a manual central warehouse 7, an adjoining reloading area B with a plurality of reloading stations 3 for reloading goods from the containers for hanging transport into overhead conveyor pockets H, a storage area C supplied thereby with a buffer storage 4 with circulating storage 8 for the overhead conveyor pockets and an adjoining packing area 5 with a plurality of packing stations and a shipping area 6.

[0049] The manual central warehouse 7 is used to supply the replenishment warehouse 2 with mixed containers. It consists of shelving units separated by aisles through which several order pickers 20 move, picking goods from the compartments into containers on their trolleys. The order pickers are guided by systems such as pick-to-voice or pick-to-light, or manual order lists, etc.

[0050] In the manual central warehouse 7, the order-oriented order pickers 20 manually fill containers mixed with different goods from multiple orders by moving through their respective zones I, II, and III and processing the assigned goods lists. The assigned goods originate either from individual goods orders or orders that are scheduled for processing together in the circulation buffer 8 within approximately 15-minute slots. To do this, the multiple order pickers 20 move in parallel through the manual central warehouse 7 and fill containers they have carried in their zones I, II, and III with goods from different orders, each of which is assigned to a slot or time period of a scheduled processing time (approximately 1 hour in the future). The resulting mixed containers are grouped across all zones and order pickers according to the 15-minute time periods in order to jointly complete orders that are to be made available during this time.

[0051] The containers are carried by the order pickers 20 on a trolley, which is then unloaded after completion of the run, with the containers being fed into the replenishment warehouse 2 via a suitable station and conveyor technology.

[0052] In contrast to containers containing only single goods, these mixed containers are always completely emptied or reloaded at the transfer stations and are then returned via a cycle either to the manual central warehouse and / or to the goods receiving area for reuse.

[0053] The replenishment warehouse 2 is supplied with single-goods containers from a goods receiving area 14 and with mixed containers from the manual central warehouse 7. The replenishment warehouse supplies a plurality of transfer stations 3 with containers for transferring the goods from the containers into overhead conveyor pockets for suspended transport. For this purpose, each transfer station 3 is connected to an aisle of the replenishment warehouse via inbound and outbound conveyors 15.

[0054] The replenishment warehouse 2 is a multi-aisle, multi-level rack warehouse operated by single-level storage and retrieval machines, so-called shuttles 17, which supply and retrieve containers from the storage locations in the shelves 18 from the aisles 19. The containers can then be transported to and from the transfer stations 3 via lifts 21 and connected conveyors 15. Each aisle is connected to a transfer station 3.

[0055] The buffer storage 4, which is connected to the transfer stations 3 via an overhead conveyor loop 13, comprises a plurality of parallel circulating storages 8 for the overhead transport means.

[0056] Transfer stations 3 are manual or automatic stations where the goods are individually transferred from the containers into overhead conveyor pockets. For this purpose, the suspended overhead conveyor pockets are provided open so that goods can be inserted manually or automatically.

[0057] The transfer stations 3 thus supply the buffer storage 4 with the individually filled overhead conveyor pouches via the overhead conveyor loop 13. There, they are stored in parallel recirculating storage units 8 until the goods of the respective order are completed. They are then transported via a collection line 24 to a packing area 5, where they are repackaged into shipping containers according to the order. From there, the orders or shipping containers ultimately reach a shipping area 6. It is understood that there is an empty pouch return line from the shipping area, which is not shown. When the filled overhead conveyor pouches are loaded into the overhead conveyor loop 13 at the transfer stations 3, they are tracked via assigned windows.

[0058] From the overhead conveyor loop 13 in the buffer storage 4, the overhead conveyor pockets are fed into the respective recirculating storage 8 in a controlled manner via switches 11. Each recirculating storage 8 consists of an infeed conveyor 25, which is connected to the inlet-side switch 11a and via which new overhead conveyor pockets are stored when the assigned window reaches there. A stopper 28 is also provided in the infeed conveyor 25 so that the overhead conveyor pockets can be buffered specifically to a free space in the adjoining rotating recirculating conveyor 26. Accordingly, a further controlled switch 11b is provided at the interface between the infeed conveyor 25 and the recirculating conveyor 26. Analogous to the infeed conveyor 25, an outfeed conveyor 27 is provided at the outlet of the recirculating conveyor 26 to remove required overhead conveyor pockets from the recirculating storage 8.Accordingly, a goods identification device 12 and another controlled switch 11c are provided at the interface between the discharge conveyor 27 and the circulating conveyor 26 to selectively discharge the desired overhead conveyor pocket onto the discharge conveyor 27. The discharge conveyor 27 has two sections 29, 30 separated by stoppers 28 before it flows into the collection section 24 via another switch 11d. The first section 29 is a reserve section to ensure continuous movement of the overhead conveyor pockets. The second section 30 is a collection section for grouping overhead conveyor pockets.

[0059] Preferably, the infeed conveyor 25 and the outfeed conveyor 27 are located on opposite sides of the circulating conveyor 26. In a circulating storage 8 there is selective access to each individual overhead conveyor pocket (no FIFO).

[0060] Individual orders 22, i.e., orders that only comprise a single item of goods, can be handled separately to avoid slowing down the overall system. This is indicated by dashed arrows. Such individual orders can thus be transported directly from the manual central warehouse 7 to the transfer stations 3. Direct transport to the packing area 5 or a dedicated transfer station 3* is also conceivable. This transfer station supplies a bypass 23 that leads directly to the collection line 24, bypassing the buffer 4. At the transfer station 3* or its switch 11, individual orders originating from the normal transfer stations 3 in the overhead conveyor loop 13 can also be selectively diverted into the bypass 23.

[0061] The entire system 1 and its areas as well as all systems and facilities therein are controlled by a central control system 16.

[0062] The controller 16 is configured to assign a plurality of orders to be picked to each revolving storage unit 8 and to create a complete list of all goods and their quantity per revolving storage unit 8 from these orders. Based on the complete list, the goods contained therein are fed to the respective revolving storage unit 8, with the overhead conveyor pockets being channeled into the respective revolving storage unit 8 via the respective switch 11. The thickness of the loaded overhead conveyor pockets plays a role in the control and dimensioning of the routes.

[0063] Based on the incoming orders, the control system 16 distributes the orders and thus the goods to specific revolving storage areas 8, from which master lists are generated. These master lists control the provision of goods from the replenishment warehouse 2 to the transfer stations 3 and finally the transfer of the goods into the revolving storage areas 8 according to the master list, taking into account the thickness of the loaded overhead conveyor pockets. This also involves the supply of the replenishment warehouse 2 with mixed containers from the manual central warehouse 7 and single-goods containers via the goods receiving area 14.

[0064] The control system 16 is configured to regularly check each circulating storage unit 8 after a new product has been fed in, using the product identification device 12 at the exit of the circulating conveyor 26, to determine whether all the products required to fulfill an original order are included in the overall list, and then to control the switch 11c using the product identification device 12 there to discharge the product of the order, which product then passes onto the discharge conveyor 27 and later onto the collecting line 24 for feeding to the packing area 5.

[0065] The control system 16 is also configured to regularly check each circulating storage unit 8 after a new product has been fed in, using the product identification device 12 at the exit of the circulating conveyor 26, to determine whether all the goods required to fulfill another order to be given priority are already contained in the respective circulating storage unit 8, and then to control the switch 11c using the product identification device 12 there to discharge the product of this other order, which then passes onto the discharge conveyor 27 and later onto the collecting section 24 for feeding to the packing area 5.

[0066] Alternatively, it is also checked whether all goods required to fulfill another order that is to be prioritized are already distributed in two or more circulation storage units 8, and then the switches 11c of the respective circulation storage units 8 are controlled with the aid of the goods identification devices 12 there to divert the goods of this other order onto the collection line 24 for delivery to the packing area 5. The goods then originate from more than one circulation storage unit 8.

[0067] This procedure allows new, urgent express orders to be given higher priority, meaning they can be processed sooner. Any missing items for the original order are then added back to the overall list and delivered within the 15-minute timeframe.

[0068] Using the flowchart of the Fig. 3 this procedure is explained again in general terms.

[0069] First, a large number of orders are loaded into the system, which may originate from an e-commerce ordering system (Step I). These orders are distributed among the circulating storages 8 (step II) and from this, a total list of goods for each circulating storage 8 is generated based on the multitude of goods contained in the orders (step III). Using this list, the rest of the system is controlled to supply the goods to the respective circulation storage 8 in a timely manner (step IV).

[0070] Each circulating storage unit 8 requests the goods from the entire list, which are provided under the control of the controller 16, starting with the supply of the replenishment warehouse 2 from the goods receiving area 14 or the manual central warehouse 7, the removal of the corresponding containers and reloading in the reloading stations 3 and conveying to the entrance 9 of the respective circulating storage unit 8 (step V).

[0071] After a suspended conveyor pocket is fed through the switch 11a at the entrance 9 and the infeed conveyor 25 and the switch 11b, the controller 16 regularly checks the respective circulating conveyor 26 of the circulating storage 8 using the goods identification device 12 at the switch 11c to determine whether all the goods required to fulfill an order are present (step VI or VIC). If this is not the case, the system waits for further goods.

[0072] However, if all goods or overhead conveyor pockets required to fulfill an order are present, they are diverted via switch 11c onto the discharge conveyor and via switch 11d onto the collection line 24 (step VII) to be conveyed to packing area 5 (step VIII). If necessary, stoppers 28 are used in areas 29 and 30.

[0073] It is understood that, within the scope of the invention, a "finished" job in this sense can also be a new job with a higher priority, which is therefore to be given priority. This is also checked in step VI or substep VIA. It is also checked whether a new job with a higher priority can be fulfilled if other circulating memory contents are taken into account (step VIB).

[0074] In Fig. Figure 4 shows an example of a transfer station 3 or measuring device 3*, in which goods (from a container, see above) are individually transferred into overhead conveyor pockets via a conveyor belt. This is therefore a loading station. The removal from the container is not shown. The corresponding process is shown in Fig. 5 shown.

[0075] If the product is unknown to the system, it first checks (Step I) whether new model data is available and, if necessary, loads it (Step IA). The product is then examined in the loading station or measuring device 3*, and the obtained data is fed into the system. The following steps are explained using a blue, foil-wrapped, medium-sized sweatshirt as an example of a soft product and a red-gray shoebox as an example of a non-compressible product.

[0076] If the transfer station 3 is additionally equipped with sensors, it can serve as a measuring device 3* within the meaning of the invention and can simultaneously generate data, as described below, for model creation, improvement and corresponding prediction of the orientation of the goods during filling and the thickness.

[0077] Transfer station 3 and measuring device 3* are therefore mechanically identical. They differ only in terms of their sensor technology and additional functionality. It is clear that transfer station 3 and measuring device 3* can also be designed differently.

[0078] The measuring device 3* comprises a 2D camera 40, a lidar sensor 41, a scale 42, and automatically actuated mechanical guides 43 for measuring softness and compression. All of this data is fed into the controller 16.

[0079] First, a product is taken from a container and the barcode of the product is scanned with the 2D camera 40 as a scanner (step II of the Fig. 5 to the process in the measuring facility, as well as at a normal transfer station 3 - accordingly, this also applies to normal loading according to the procedure according to Fig. 6) and thus the host data is retrieved from the controller 16 via the identity of the goods.

[0080] For the sweatshirt, for example, this results in - SKU ID #343

[0081] This is used to retrieve customer data, e.g. • Color: Blue • Size: Medium • Dimensions: 30 x 25 x 5 • Weight: 0.25kg

[0082] For the shoe box, for example, this results in - SKU ID #541

[0083] This is used to retrieve customer data, e.g. • Color: Red / Gray • Size: Large • Dimensions: 36 x 23 x 13 • Weight: 1.2kg

[0084] The goods are then placed on the conveyor belt 128, aligned and measured as follows in order to generate sensor data to confirm the host data or, if necessary, to correct or supplement it.

[0085] The following measurements (step III) may be performed: a. Photos from the 2D camera 40; b. Dimensional measurements using Lidar sensor 41; d. Weighing and determining the centre of gravity using the scale 42; e. Measurement of softness and compression using the automatically operated mechanical guides 43;

[0086] Photos are taken and the goods are measured with the Lidar sensor.

[0087] For the sweatshirt, the laser measures 30.5 cm x 25.3 cm x 7.0 cm and for the shoebox, the dimensions are 36.5 cm x 23.3 cm x 13.0 cm. Through user input, the sweatshirt is given the class “shirt,” whereas the shoebox is given the class “box.”

[0088] The model uses a Bayesian model technique to combine the thickness data of the hanging transport bag with the softness data of the product in order to predict the compression of the product when filling the hanging transport bag within a probability corridor and to classify the product by group (step IV in Fig. 5).

[0089] For the sweatshirt, for example, this results in: -Classification model Softness: 9 / 10 Object: rectangular prism Material: Fabric in polybag -Regression model Length: 31.5cm Width: 25.5 cm Height: 5.3 cm Elasticity: 8.4 -Cluster model Cluster: Men's Shirts (Cluster 177 of 3355)

[0090] For the shoe box, for example, this results in: Classification model Softness: 1 / 10 Subject: rectangular prism Material: Box Regression model Length: 36.0cm Width: 23.1cm Height: 13.5cm Elasticity: 1.4 - Cluster model Cluster: Shoes for Women (Cluster 1431 of 3355)

[0091] Thus, a grouping classification of the goods and the application of a clustering model based on the host data, photos and measurements are carried out, which are analyzed and used to classify the goods into a cluster group and to calculate the consistency (e.g. softness) of the goods.

[0092] In step VA, manual verification of the data can be performed if desired in step V. For example, for the sweatshirt, the user can correct the height data to 7.1 cm. For the shoebox, this step is skipped.

[0093] The model also uses a regression model technique to link the thickness data of the hanging bag with the center of gravity and orientation data of the goods to predict the orientation of the goods when filling the hanging bag within a probability corridor, if desired (step VIA). Thus, using information from the host data, photos, measurements, and the results of the model for classifying and clustering goods, it is possible to calculate how the goods fall in the hanging bags—i.e., the orientation of the goods when filling the hanging bags.

[0094] For example, the compressible sweatshirt can be predicted to be placed in the pocket in the following position: Position x, y, z = [2.3 cm; 10.5cm ; 3.3cm] Roll, pitch, yaw: [30°, 2°, 5°] With the shoebox, this step is skipped.

[0095] The controller 16 is also programmed to link the data of the goods with those of the hanging transport bag in a model for prediction, for which purpose the thickness data of the hanging transport bag are linked with the dimensional data of the goods by means of a regression model technique in order to thereby predict the thickness of the hanging transport bag filled with the goods within a probability corridor (step VII).

[0096] For example, for the sweatshirt, the thickness prediction model results in Bag thickness: 7.5 cm Lower limit of thickness: 7.1 cm Upper limit of thickness: 8.5 cm

[0097] For example, for the shoe box, the thickness prediction model yields Bag thickness: 14.1 cm Lower limit of thickness: 13.5cm Upper limit of thickness: 24.0cm

[0098] The goods are then released and filled into the hanging transport bag H (step VIII).

[0099] The thickness of the opened hanging transport bag H filled with the goods can now be checked in a verification measurement before closing (steps IX, IXA).

[0100] For this purpose, laser sensors 44 are used to determine the resulting thickness of the hanging transport bag and, optionally, the orientation of the goods inside the hanging transport bag. These measurements are used to feed and validate the models for the orientation prediction and the thickness prediction and are also transmitted to the controller 16 for this purpose.

[0101] For the sweatshirt, for example, the measured thickness is 8.1 cm.

[0102] The hanging transport pocket H is then closed (step XA) or closes automatically and can be fed into the hanging conveyor loop 13.

[0103] Alternatively or additionally, the thickness of the closed hanging transport bag H filled with the goods can now be checked in a verification measurement (steps X, XI, XIB).

[0104] For example, the sweatshirt has a measured thickness of 8.0 cm and the shoebox has a measured thickness of 14.3 cm.

[0105] If no verification should be carried out, the hanging transport bag is removed (steps X, XI, XIA, XII).

[0106] The thickness data is now linked to the respective hanging transport bag and the specific goods contained therein, and the control system can therefore use the thickness to determine the track lengths and control switches, etc.

[0107] The data from the measurements and verifications are made available to the controller 16 (steps III-VII). This controller is programmed accordingly to link the product data with that of the hanging transport bag in a model for forecasting.

[0108] Verification measurements compare or verify whether the model's predictions are correct or accurate. If not, the data can be used to adaptively improve the model, and if so, the probability corridor can be reduced. All data, together with the prediction and validation data, is used to update the models. By constantly repeating the steps and monitoring the alignment of the goods and the resulting measurements while taking the parameters into account, the accuracy of the models improves over time. All data can be used in each subsequent step as needed.

[0109] Therefore, in principle, only one master measuring device is required, running in parallel with the other normal pocket loading stations of a sorting stream. The host data is enriched with new and / or updated information.

[0110] In the first step I of model generation (cf. Fig. 7) the data from the measuring device 3*, which was received by the control system and stored in a database, is retrieved.

[0111] Subsequently, in step II, the data are initially linked together in a classification, regression, and cluster analysis as part of the model generation process. This link is then saved again in the database for later use (see also Fig. 8).

[0112] In the classification ( Fig. 8A) A Bayesian modeling technique, or alternatively a support vector machine, a decision tree, a random forest, or a neural network, is used to group the goods and apply a clustering model based on the host data, photos, and measurements, which are analyzed and used to divide the goods into a classification group. In addition to the thickness data of the loaded hanging bag, the softness of the goods, the type of geometry of the goods, and the material of the goods are also taken into account to create the classification groups. The use of neural networks, which are particularly well suited for processing image data, is particularly interesting.

[0113] In regression ( Fig. 8B) is carried out by means of a regression model technique, implemented e.g. by means of a decision tree, a linear or logistic regression or a neural network, a linking of the product properties based on the host data, photos and thickness measurements of the hanging transport bag with the geometric dimensional data (bounding box) and the elasticity of the product.

[0114] In cluster analysis ( Fig. 8C), the goods are grouped based on the goods' characteristics (host data, photos, etc.) and the thickness measurements of the hanging bag using the K-Means algorithm. K-Means, Gaussian Mixing, Neural Network Clustering, and Hierarchical Clustering can also be used.

[0115] Then in step III of Fig. 7 (cf. Fig. 9A) a partial model is calculated in which, by means of a regression model technique - implemented, for example, by means of a decision tree, a linear or logistic regression or a neural network - information from the host data, photos, measurements and the results of the model for classifying, regressing and clustering goods is calculated as to how the goods will fall or be aligned in the hanging transport bags, i.e. the alignment of the goods when filling the hanging transport bags.

[0116] Then in step IV of Fig. 7 (cf. Fig. 9B) a calculation of a partial model in which a regression model technique - implemented, for example, by means of a decision tree, a linear or logistic regression, or a neural network - incorporates information from the host data, photos, measurements, and the results of the model for classification, regression, and clustering, as well as the above-mentioned orientation of goods, such as whether the thickness of the hanging transport bag filled with the goods lies within a probability corridor.

[0117] Then in steps V and VI of Fig. 7 (cf. Fig. 9C) A calculation of a partial model in which information from the host data, photos, measurements, and results from the model classification, regression, clustering, orientation, and thickness models are incorporated into a regression model, which is implemented, for example, using a decision tree, linear / logistic regression, or a neural network. The regression model predicts the system's KPI data. For example, it can predict the achievable throughput in the system or the required size of an order in a buffer for orders or hanging transport bags.

[0118] For example, based on the known data for the sweatshirt or shoe box, a throughput of 788 items per hour could be predicted. This prediction can be made by retrieving all SKU IDs, quantities, locations, customer data, etc. in the system; retrieving all historical model results (item classification, regression, clustering, placement prediction, thickness prediction); passing all retrieved data to the KPI prediction model and calculating the prediction for the system KPI (e.g., throughput in items / hour).

[0119] This prediction can then be compared with actual measured values, if available. In this case, the actual throughput was measured at 811 items per hour. The deviation can be used to improve the model.

[0120] In the subsequent steps VII, VIII and IX, a query is made as to whether there is new data that needs to be taken into account, whether there is new KPI data or whether a specified period of time has passed since the last update of the model.

[0121] This results in a model that can be used at normal transfer stations 3 according to Fig. 6 can be used to simply calculate the thickness of the hanging transport bag when filled with a known specific product. Accordingly, Fig. 6 the same steps as in Fig. 5 are designated by the same reference numerals.

[0122] Thus, the thickness prediction can be calculated based on information from the database of the controller 16, i.e., from host data, photos, measurements, results from the model for classifying and clustering goods, and results from the model for predicting orientation. This thickness of the filled hanging transport bag can then be used in the control system, as explained above.

[0123] This results in (step II) for the now known sweatshirt in the transfer station 3 after scanning:- SKU-ID #343

[0124] This is used to retrieve customer data, e.g. • Color: Blue • Size: Medium • Dimensions: 30 x 25 x 5 • Weight: 0.25kg

[0125] The model data is then provided (Step III): -Classification model Softness: 9 / 10 Object: rectangular prism Material: Fabric in polybag -Regression model Length: 31.5cm Width: 25.5 cm Height: 5.3 cm Elasticity: 8.4 -Cluster model Cluster: Men's Shirts (Cluster 177 of 3355)

[0126] Then, if desired (query VI), the orientation of the goods when filling the hanging transport bag can be predicted within a probability corridor (VIA).

[0127] Otherwise, jump directly to the thickness prediction (VII). For the sweatshirt, this results in: Bag thickness: 8.0 cm Lower limit of thickness: 7.1 cm Upper limit of thickness: 8.5 cm

[0128] The goods are then filled into the bag (step VIII) and the bag is closed (XA) and transported to a sorter etc. (XIA).

[0129] The pocket thickness can then be used by the control system to determine distances, track lengths, etc. and to control switches, etc.

Claims

[1] Computer-implemented method for operating a hanging conveyor for successive hanging transport bags, wherein hanging transport bags filled with goods can be conveyed on the hanging conveyor via hanging adapters and the distance between the hanging transport bags can be adjusted by the distance of an engagement of the hanging adapters in the hanging conveyor, which distance is variable in the grid, and in which the hanging transport bags are placed one after the other in an infeed on the hanging conveyor, hanging on the hanging adapters, characterized by that the distance between the hanging transport pockets when feeding onto the overhead conveyor is adjusted to optimise the distance based on the thickness of the hanging transport pockets in the transport direction, for which purpose the thickness of the hanging transport pockets is calculated in advance based on the goods. [2] Method according to claim 1, characterized by that the advance calculation is based on data linked to the goods (identity) of the filled hanging transport bag. [3] Method according to claim 2, characterized by that the linked data are properties of the product itself and properties of the hanging transport bag. [4] Method according to claim 3, characterized by that the characteristics of the goods include dimensions, shape or geometry, centre of gravity, elasticity, softness, contents and packaging type, and orientation. [5] Method according to claim 3 or 4, characterized by that the properties of the hanging transport bag include its thickness in the transport direction, which is detected by sensors after it has been filled with goods. [6] Method according to claims 4 and 5, characterized by that the data of the goods are linked with those of the hanging transport bag in a model for forecasting. [7] Method according to claim 6, characterized bythat the model links the thickness data of the hanging transport bags with the dimensional data of the goods by means of a regression model technique and / or a classification model technique and / or a cluster model technique in order to predict the thickness of the hanging transport bags filled with the goods within a probability corridor. [8] Method according to claim 7, characterized by that the model links the thickness data of the hanging transport bags with the center of gravity and orientation data of the goods by means of a regression model technique and / or a classification model technique and / or a cluster model technique in order to predict the orientation of the goods when filling the hanging transport bags within a probability corridor. [9] Method according to claim 7 or 8, characterized bythat the model uses a regression model technique to link the thickness data of the hanging transport bags with the softness data of the goods in order to predict the compression of the goods when filling the hanging transport bags within a probability corridor. [10] Method according to one of claims 5 to 9, characterized by that the thickness data of the hanging transport bags are recorded in the open and / or closed state. [11] Method according to one of claims 6 to 10, characterized by that the model uses the classification model technique to link the thickness data of the hanging transport bag filled with the goods with the identity of the goods in such a way that the goods are classified according to a higher-level characteristic of their identity. [12] A warehouse with an overhead conveyor device for the automated conveying of transport bags loaded with at least one item of goods along the overhead conveyor device, with a control system for controlling the warehouse and with a bag loading station for loading a transport bag with at least one item of goods and subsequently feeding the loaded transport bag into the overhead conveyor device, wherein the distance between the overhead transport bags can be adjusted by the distance, which is variable in the grid, of an engagement of the overhead adapters in the overhead conveyor, characterized byin that the warehouse has a measuring device which is connected to the controller for exchanging data, and the measuring device is set up to generate data, wherein the data comprise certain properties of the at least one product by means of sensors and optionally manually controlled examinations, and the data comprise the thickness, determined by means of sensors, in the transport direction of a closed transport bag filled with the at least one product, wherein the controller is programmed in a computer to carry out the method according to one of the preceding claims, to calculate an expected transport bag thickness for a certain product from the data, to control the bag loading station with the expected transport bag thickness in such a way that the distance when feeding in the loaded transport bags is set based on the expected transport bag thickness. [13] Warehouse according to claim 12, characterized bythat the control system is configured to create a data collection from the data historically generated by the measuring device for statistical analysis by the control system in order to determine the expected transport pocket thickness without prior measurement. [14] Warehouse according to claim 12 or 13, characterized by that the measuring device is identical to the at least one bag loading station with regard to bag loading.

Images