Method for selecting a computer-supported analysis model

Abstract

The invention relates to a method for selecting a computer-supported analysis module for use in the analysis of optical anomalies of a product produced on a manufacturing line, comprising the following steps: - providing at least one computer-supported analysis model - determining at least one property of the product and / or at least one production parameter - selecting an analysis module, assigned to a product type, from a plurality of analysis models assigned to product types taking into consideration the at least one determined property of the product and / or the at least one determined production parameter - examining the product using the selected analysis model.

Description

[0001] Description

[0002] Procedure for selecting a computer-aided analysis model

[0003] The invention relates to a method for selecting a computer-aided analysis model for use in the analysis of optical anomalies of a product produced on a production line.

[0004] Such computer-aided analysis models are used, for example, in the detection of visual anomalies with the aid of a visual quality control system. In this case, the production lines or assembly lines being monitored may operate with rapid product changeovers. In such a case, different products are manufactured by the same production line at different times.

[0005] Accordingly, it may be necessary to adapt the analysis model used to the respective product. Depending on the product currently being produced, different optical parameters may indicate a defect or be harmless with regard to the quality of the product being inspected.

[0006] The quality control system that detects the anomalous data is usually located at the end of the production line or lines. The anomalies can be different for each product. For example, in a bakery use case, different visual criteria may indicate a quality defect for different products such as baguettes, rolls, and croissants. The possible anomalies can even be ambiguous between different products, i.e., an anomaly in one product may be acceptable for another. Therefore, when using a central analytics model based on artificial intelligence or machine learning, it cannot be trained for all products. Individual models can be trained for each product. However, since products are frequently changed in a production line, efficient selection of the right model is important.It is possible to create a schedule that specifies when a changeover from one product to another product takes place. This method is error-prone because it does not determine which product is currently on the production line.

[0007] Two approaches to solving the problem are known from the state of the art. Firstly, the product model can be selected manually by an operator. The scalability of this approach is limited when new products are added to the list of possible products, and it is also prone to errors. Secondly, an object recognition model can be used, which optically recognises the product on the production line based on a captured image and loads the corresponding model. In this case, an incorrect prediction from object recognition can also lead to the retrieval of an incorrect model, so this method is also prone to errors.

[0008] The task is to enable an efficient selection of the appropriate analysis model for detecting anomalies in an object under test.

[0009] According to the invention, a method is provided which allows such an efficient selection of the appropriate analysis model. The inventive method for selecting a computer-aided analysis model for use in the analysis of optical anomalies of a product produced on a production line basically comprises the following steps:

[0010] - Providing at least one computer-aided analysis model,

[0011] - Determining at least one property of the product and / or at least one production parameter,

[0012] - Selecting an analysis model assigned to a product type from a plurality of analysis models assigned to product types, taking into account the at least one determined property of the product and / or the at least one determined production parameter, and

[0013] - Examine the product using the selected analysis model.

[0014] According to the invention, additional information about the product is utilized, thus providing an approach for integrating models from different products. The information obtained about the product and / or a production parameter allows conclusions to be drawn about the type of product currently being produced.

[0015] The method according to the invention has the advantage that the products produced on the production line can be reliably and efficiently assigned to a suitable analysis model. Alternatively or in addition to an optical analysis, for example based on image recognition software that analyzes an image captured by a camera, it is possible to use the at least one determined property and / or the at least one determined production parameter when selecting the analysis model.

[0016] A production line can be understood in particular as a semi-automatic or automatic production facility, which may, for example, be equipped with a conveyor belt. On the production line, several or many products of the same type can be produced consecutively in series production.

[0017] In the context of the present description, a product type is understood to mean, in particular, a plurality of similar or identical products. In particular, the product type can correspond to a product model, for example. The product can be examined, in particular, optically, for example by taking one or more images of the product and subsequently analyzing the images. This allows optically visible anomalies or defects to be detected. The method according to the invention can therefore be used, for example, in the context of quality assurance in a production plant.

[0018] The unique product ID, for example a product identification number or another code assigned to the product, is ideally unique, but it is not reliable because it can change over time or be adjusted by the operator. Therefore, according to the invention, other parameters are used alternatively or in addition to the product ID when assigning the product type to the analysis model to be used. The analysis model relevant for a product type can then be selected based on the incoming parameters. These parameters can normally be discretized. It may be that not all parameters assigned to the product type are relevant for detecting anomalies, but are required for later steps, such as packaging.

[0019] It can advantageously be provided that the product is examined immediately after a change in the product type of the products produced on the production line. It is possible for the method to be carried out only for the first product produced after a change in the product type. When the product type produced changes, the parameters used to analyze the products or the analysis model used usually have to be adapted. If the method according to the invention is carried out for the first product produced after such a change in the product type of the products produced on the production line, the analysis model to be used at that time can be reliably selected. It may be sufficient to carry out the method only for the first product produced after the change in the product type.However, it is also possible to carry out the procedure at regular intervals without changing the product type in order to ensure that a change in the product type has not taken place without the appropriate analysis model having been selected.

[0020] It can be provided that the method according to the invention is carried out when it is detected that a change in product type has occurred. For this purpose, a visual inspection of each produced product can be carried out. It is also possible for a corresponding control signal to be sent to the device by means of which the method according to the invention is carried out or by means of which the products are analyzed. Such a control signal can be sent, for example, by a control system.

[0021] According to a further development of the invention, the selected analysis model remains activated until it is detected that a new product type is being produced on the production line. In other words, the selected analysis model can remain activated as long as there is no reason to assume that the product type being produced has changed. This enables a resource-saving process.

[0022] It is advantageous for the selected analysis model to be an analysis model based on artificial intelligence and / or machine learning. Such methods are often used in the analysis of optical anomalies and in particular in the quality assurance of products produced on a production line, since such analysis models allow optical anomalies or quality defects to be determined particularly efficiently. The at least one analysis model provided can have been trained to analyze a predetermined product type. Such a method is common for analysis models based on artificial intelligence and / or machine learning and makes it possible to achieve a desired level of accuracy in the analysis.

[0023] It is possible that a non-optical method is used to determine at least one property of the product and / or at least one production parameter. Such a method can be, for example, weighing, temperature measurement or another non-optical measuring method. In this case, when determining the at least one property of the product and / or the at least one production parameter, one or more of the following properties can be determined: a size of the product, a weight of the product, a shape of the product, a color of the product and / or a temperature of the product. These properties of the product offer a variety of possibilities for identifying the product type, thus enabling a reliable selection of the analysis model to be used.

[0024] It is also possible that, when determining at least one property of the product and / or at least one production parameter, a production parameter such as a conveyor belt speed and / or a production temperature is determined. This information also makes it possible to determine the product type to be analyzed with high accuracy.

[0025] It is possible that, in addition to the determined product property and / or in addition to the production parameter, a product identification code is determined. This results in a process with redundant information retrieval, thus increasing the reliability of the process. The product identification code can be determined, for example, by querying, reading, or receiving the identification code.

[0026] A further development of the method according to the invention provides that, in addition to the determined property and / or in addition to the determined production parameter, an image of the product taken by a camera is used when selecting the analysis model. Here, too, expanding the information provided to the system results in increased accuracy in selecting the analysis model to be used.

[0027] It is advantageous for the analysis model to include at least one autoencoder. Autoencoders are often used for unsupervised optical anomaly detection. An autoencoder is trained on healthy data and outputs reconstructed data. Since the model is trained only on healthy data, the anomalous data cannot be reconstructed well, resulting in a higher reconstruction error.

[0028] According to a development of the invention, a binary decision tree is used when selecting the analysis model, on the basis of which decision tree a plurality of properties of the product to be tested and / or a plurality of production parameters are included when selecting the analysis model. A binary decision tree can be understood in particular as a sequence of queries about properties, for which two answers are possible for each query. For example, the answer options can be "yes / no", "larger / smaller" or similar answer options. Each answer option can be assigned a subsequent query until finally the last query in the sequence of queries is assigned an individual analysis model, which is then used.It is possible to include an assignment rule that assigns an analysis model to each combination of values ​​of a plurality of input parameters when selecting the analysis model. Such an assignment can be made, for example, in the form of the binary decision tree described above. However, it is also possible to implement the assignment, for example, using a possibly multidimensional matrix or another assignment rule.

[0029] It is advantageous for the assignment rule to be saved as binary code. This type of implementation of the assignment rule is particularly advantageous when a binary decision tree, as described above, is used to select the analysis model. However, all queries can also be carried out in parallel without a tree structure. Each possible combination of binary values ​​can then be assigned an analysis model to be used. It is also possible, however, to assign an error message or an instruction to repeat the procedure to certain combinations of values ​​to which no product type can be reliably assigned.

[0030] It can advantageously be provided that the binary code is based on an expandable format so that additional product types can be included in the assignment rule. In particular, additional bits, which may be zero by default, can be present at the end of a binary data set. If the assignment rule is to be made more complex or expanded, the additional bits can be activated and filled with content.

[0031] According to a preferred embodiment, the method is carried out with the aid of a computer model based on artificial intelligence and / or machine learning, wherein the computer model has been trained to assign input parameters to product types. This allows for a high degree of accuracy and reliability in assigning the analysis models to be used to the products under investigation.

[0032] A further development stipulates that if a product type is to be produced on the production line to which no analysis model has yet been assigned, the following steps are carried out:

[0033] - Check whether a suitable analysis model is selected without adjustments, and

[0034] - If no suitable analysis model is selected without adjustments, train the computer model until a desired accuracy in assigning the analysis model to the product type is achieved.

[0035] The assignment of the appropriate analysis model to the parameters belonging to the new product can be conveniently carried out using a Kl model or a machine learning model.

[0036] A further development of the invention provides for multiple product types to be assigned to the same analysis model. Multiple analysis models can be combined into a group or groups of analysis models.

[0037] Product instances or product types can be grouped, and a model can be trained for each product type. For example, a bakery might have different types of pretzels, such as pepper, salt, or sesame pretzels, all belonging to the same category, and their quality is determined only by their shape, which is specified in the parameters used to select the analysis model. Similar instances can also be grouped based on these parameters. This allows the system to adapt more efficiently to new instances of the same product category, and the number of trained models can be reduced.In a continuation of this approach, it is possible for all product types to be assigned to the same analysis model, wherein the analysis model is designed in such a way that it uses different examination criteria when examining the product based on the at least one determined property of the product and / or based on the at least one determined production parameter. This makes it possible to reduce the number of analysis models kept and ideally to restrict them to a single analysis model. In other words, all analysis models assigned to different product types can then be integrated into one analysis model. The input for the analysis model can be an image and a set of parameters. Since such a set of parameters can be modeled as a combination of attributes, this set of parameters can also be encoded in a binary format.To accommodate new parameters, empty spaces for future parameters can be included in the binary format. These can be appended to the end of the parameter list and initially have the value zero.

[0038] In a further development of the invention, it is provided that the selection of an analysis model (34, 36, 38) assigned to a product type is carried out by the following steps:

[0039] - Taking a picture ( 56 , 58 ) of the product

[0040] - entering the image (56, 58) and the at least one property and / or the at least one production parameter into an analysis model, wherein the analysis model has previously been trained with structurally identical data.

[0041] This method is characterized by the fact that, strictly speaking, no analysis model is selected. Instead, a single analysis model is available. However, its input data is expanded to include not just an image of the product to be analyzed as the input variable, but also additionally determined product properties and / or determined production parameters. If the analysis model is based on artificial intelligence and has been trained with appropriate data, a high level of accuracy in detecting optical anomalies in various product types can be achieved even with just a single analysis model.The options for using the various parameters—in other words, the product properties and / or production parameters—are essentially identical to those described previously in the process for selecting an analysis model. Likewise, the parameter sets can be stored and processed in the formats already described.

[0042] The object is further achieved by a computer program comprising instructions which cause a computer to carry out a method according to the invention, and by a storage medium on which such a computer program is stored.

[0043] The above-described properties, features, and advantages of this invention, as well as the manner in which they are achieved, will become clearer and more readily understood in connection with the following description of the embodiments, which are explained in more detail in conjunction with the drawings.

[0044] Figure 1 is a schematic representation of a production facility,

[0045] Figure 2 is a flow chart illustrating a method according to the invention,

[0046] Figure 3 is a schematic representation of an analysis model, Figure 4 is a possibility for selecting an analysis model, and

[0047] Figure 5 shows an alternative way of coding parameters.

[0048] Figure 1 shows a schematic representation of a production facility in the form of a production line 2. The production line 2 is set up to manufacture different products or product types. As an example, a first product type 4, a second product type 6, a third product type 8, a fourth product type 5, a fifth product type 7 and a sixth product type 9 are shown here. The number of reproducible product types naturally depends solely on the type and configuration of the production line 2. After production, the produced products are fed to an analysis system 10 or analyzed by it. In particular, the analysis system 10 can carry out an examination to determine optical anomalies.

[0049] Figure 2 shows a flow chart to illustrate a first embodiment of a method according to the invention.

[0050] In a step S100, at least one computer-aided analysis model is provided. Typically, a plurality of analysis models, for example, more than five, more than ten, or more than 100 analysis models, will be present or provided. The analysis models may be analysis models based on artificial intelligence, such as those commonly used to determine optical anomalies, for example, within the framework of a quality assurance process.

[0051] In a step S 102, at least one property of the product produced on the production line and / or at least one production parameter is determined. The property, the properties, the production parameter, and / or the production parameters are preferably selected such that they allow a conclusion to be drawn from the determined values ​​regarding the product type of the product being examined.

[0052] In a step S 104, an analysis model assigned to a product type is selected from a plurality of analysis models assigned to product types, taking into account the at least one determined property of the product and / or the at least one determined production parameter.

[0053] Finally, in a step S 106, the product is examined using the selected analysis model.

[0054] Figure 3 shows a schematic representation of an analysis model 10 , associated parts of the production line and the computer system by means of which the analysis model 10 is used.

[0055] A central processing unit 17 can be seen, with the help of which the analysis of the products is carried out. The processing unit 17 has access to several other components from which it receives information. These can be, for example, a conveyor belt 12, a camera 14 and / or a further processing unit 16, which can be, for example, a programmable logic controller (PLC). The processing unit 17 receives, in the illustrated embodiment mediated by the further processing unit 16, access to a parameter set 18. Values ​​for a first parameter 20 and a second parameter 22 are stored there. The number of parameters can in principle be as large as desired and depends on the number of different product types as well as on the specific properties of the product data.

[0056] The computing unit 17 then determines the analysis model to be used, taking into account the determined product properties and / or the determined production parameters, by comparing them with the values ​​for the first parameter 20 and the second parameter 22. The analysis models are symbolized in Figure 3 by the first autoencoder 24, the second autoencoder 26 and the third autoencoder 28. The first autoencoder 24 is assigned to a first product type 4, the second autoencoder 26 is assigned to a second product type 6, and the third autoencoder 28 is assigned to a third product type 8. The different product types 4, 6 and 8 are also assigned different values ​​or value combinations for the first parameter 20 and the second parameter 22, so that the computing unit 17 is able to select the correct analysis model for analyzing the respective product on the basis of these values.If all analysis models have the same shape and architecture (i.e. the input data for all models has the same size), only the weights of the corresponding analysis model are loaded, otherwise the model architecture is also loaded.

[0057] Figure 4 shows one possibility of how the different analysis models 34, 36, and 38 can be selected based on the determined properties and / or based on the determined production parameters. In the illustrated embodiment, a binary decision tree is used for this purpose. The first analysis model 34 links a first product type 4 with the first autoencoder 24, the second analysis model 36 links the second autoencoder 26 with the second product type 6, and the third analysis model 38 links the third autoencoder 28 with the third product type 8.

[0058] At a first fork 52, a first determined production parameter or a first determined property of the product to be analyzed is evaluated. In the exemplary embodiment shown in Figure 4, this is a temperature, which can be, for example, a production temperature or a temperature of the product. If the temperature value in the exemplary embodiment shown is below 100 °C, the left branch of the fork 52 in the figure is taken. Accordingly, the product to be analyzed is classified as belonging to the first product type 4. The first autoencoder 24 is therefore used to determine potential anomalies.

[0059] If the temperature is above 100°C, the right branch of fork 52 is taken and the selection process continues at the second fork 54. Here, a second parameter, in the illustrated embodiment, a conveyor belt speed, is analyzed. If the determined conveyor belt speed is less than 40 cm / s, the left branch of the second fork 54 is taken and the product to be examined is classified as belonging to the second product type 6. Therefore, the second autoencoder 26 is used to determine potential anomalies.

[0060] If the determined conveyor belt speed is greater than 40 cm / s, the right branch of the second fork 54 is taken, and the product to be examined is classified as belonging to the third product type 8. Therefore, the second autoencoder 28 is used to detect potential anomalies.

[0061] Figure 5 shows an alternative way of coding parameters 46, 48, which can then be used to select the required analysis model. The first parameter 46, which in the illustrated embodiment is a temperature, and the second parameter 48, which in the illustrated embodiment is again a conveyor belt speed, are shown as determined variables. Unlike in the embodiment shown in Figure 4, no binary decision tree is used here, but rather different value ranges of the parameters 46, 48 are stored in a binary code. In the illustrated embodiment, the first 4 bits of the binary codes 44, 46 are assigned to different temperature ranges, and the second 4 bits of the binary codes 44, 46 are assigned to different conveyor belt speed ranges.For convenience, the relevant bit is set to 1 and the non-relevant bits are set to 0.

[0062] In the illustrated embodiment, the first bit corresponds to a temperature range below 50°C, the second bit corresponds to a temperature range of 50-100°C, the third bit corresponds to a temperature range of 100-150°C, and the fourth bit corresponds to a temperature range above 150°C. Thus, the first four bits are assigned to the temperature, and the second four bits, in other words bits 5 to 8, are assigned to the conveyor belt speed. The fifth bit corresponds to a conveyor belt speed of below 10 cm / s, the sixth bit corresponds to a conveyor belt speed of 10 cm / s to 20 cm / s, the seventh bit corresponds to a conveyor belt speed of 20 cm / s to 30 cm / s, and the eighth bit corresponds to a conveyor belt speed of over 30 cm / s.

[0063] For the first product type 4, a temperature between 50 °C and 100 °C and a conveyor belt speed of less than 10 cm / s were determined, so that the corresponding first binary code 42 is "0100100". For the second product type 6, a temperature between 100 °C and 150 °C and a conveyor belt speed between 10 cm / s and 20 cm / s were determined, so that the corresponding second binary code is "00100100".

[0064] In the illustrated embodiment, a common autoencoder 40 has been trained for all configured product types, into which an image 56, 58 captured by a camera and the respectively determined parameter set or binary code 42, 44 are input. Corresponding input values ​​have already been used previously to train the autoencoder 40, so that in the illustrated embodiment, a single autoencoder or a single analysis model can be used to analyze the respective product. Although the invention has been illustrated and described in detail by means of preferred embodiments, the invention is not limited by the disclosed examples, and other variations can be derived therefrom by a person skilled in the art without departing from the scope of the invention.

[0065] Reference symbol list

[0066] 2 production line

[0067] 4 first product type

[0068] 5 fourth product type

[0069] 6 second product type

[0070] 7 fifth product type

[0071] 8 third product type

[0072] 9 sixth product type

[0073] 10 Analysis system

[0074] 12 Conveyor belt

[0075] 14 Camera

[0076] 16 "Computer" PLC

[0077] 17 Computing unit

[0078] 18 Parameter set

[0079] 20 first parameters

[0080] 22 second parameter

[0081] 24 first autoencoder

[0082] 26 second autoencoder

[0083] 28 third autoencoder

[0084] 30 first data channel

[0085] 32 second data channel

[0086] 34 first analysis model

[0087] 36 second analysis model

[0088] 38 third analysis model

[0089] 40 autoencoders

[0090] 42 binary code

[0091] 44 binary code

[0092] 46 parameters

[0093] 48 parameters

[0094] 50 parameter set

[0095] 52 first fork

[0096] 54 second fork

[0097] 56 Image

[0098] 58 Image

Claims

Patent claims 1. A method for selecting a computer-aided analysis model (34, 36, 38) for use in the analysis of optical anomalies of a product produced on a production line (2), comprising the steps - Providing at least one computer-aided analysis model (34, 36, 38) - Determining at least one property of the product and / or at least one production parameter - selecting an analysis model (34, 36, 38) assigned to a product type from a plurality of analysis models (34, 36, 38) assigned to product types (4, 6, 8), taking into account the at least one determined property of the product and / or the at least one determined production parameter - Examine the product using the selected analysis model (34, 36, 38) .

2. Method according to claim 1, characterized in that when determining the at least one property of the product and / or the at least one A non-optical process is used to determine the production parameter.

3. Method according to claim 1, characterized in that when determining at least one property of the product, one or more of the following properties are determined: a size of the product, a weight of the product, a shape of the product, a color of the product, and a temperature of the product.

4. Method according to one of claims 1 or 3, characterized in that when determining at least one production parameter, a speed of a conveyor belt (12) and / or a production temperature is determined.

5. Method according to one of claims 2 to 4, characterized in that in addition to the determined property of the product and / or in addition to the production parameter, an identification code of the product is determined.

6. Method according to one of the preceding claims, characterized in that when selecting the analysis model (34, 36, 38) a binary decision tree is used, on the basis of which a plurality of properties of the product to be tested and / or a plurality of production parameters are included when selecting the analysis model (34, 36, 38).

7. Method according to one of claims 1 to 5, characterized in that an assignment rule which assigns an analysis model to each combination of values ​​of a plurality of input parameters is included in the selection of the analysis model (34, 36, 38).

8. Method according to one of the preceding claims, characterized in that several product types (4, 6, 8) are assigned to the same analysis model.

9. Method according to one of the preceding claims, wherein the selection of an analysis model (34, 36, 38) associated with a product type is carried out by the following steps: - Taking a picture (56, 58) of the product - inputting the image (56, 58) and the at least one property and / or the at least one production parameter into an analysis model, wherein the analysis model has previously been trained with structurally identical data.

10. A computer program comprising instructions which, when executed by a computer, cause the computer to carry out a method according to any one of the preceding claims.

11. Machine-readable storage medium on which the computer program according to the preceding claim is stored.

Images