METHOD AND COMPUTER SYSTEM FOR DETERMINING PRODUCTION PARAMETERS FOR THE PRODUCTION OF A POLYMERIC PRODUCT

Abstract

The invention relates to a method for determining production parameters (1) for the production of a polymeric product, wherein a prediction model (7) is provided for calculating production parameters (1) based on polymeric product properties (4) on a computer system, which production parameters (1) comprise formulation portions (2) specifying raw material portions for polymeric production and comprise processing parameters (3) specifying process properties during polymeric production, wherein user input is provided to the computer system, which user input comprises user product parameter targets (5) specifying polymeric product properties (4), wherein the computer system applies the user product parameter targets (5) to the prediction model to calculate resultant production parameters (11) for the production of a polymeric product associated with the user product parameter targets (5). The invention also relates to a corresponding computer system.

Claims

1.-15. (canceled)

16. A method for determining production parameters (1) for the production of a polymeric product, wherein a prediction model (7) is provided for calculating production parameters (1) based on polymeric product properties (4) on a computer system, which production parameters (1) comprise formulation portions (2) specifying raw material portions for polymeric production and comprise processing parameters (3) specifying process properties during polymeric production, wherein user input is provided to the computer system, which user input comprises user product parameter targets (5) specifying polymeric product properties (4), wherein the computer system applies the user product parameter targets (5) to the prediction model to calculate resultant production parameters (11) for the production of a polymeric product associated with the user product parameter targets (5).

17. The method according to claim 16, wherein the production parameters comprise machine processing parameters specifying machine process properties during production, preferably, that the machine process properties of the resultant production parameters (11) are applied to provide raw materials according to the formulation portions (2) to a machine for polymeric production, preferably, that the machine process properties comprise user-settable machine process settings and that the resultant production parameters (1) are applied to select machine process settings in a machine for polymeric production, such that a polymeric product is produced by the machine from the raw materials.

18. The method according to claim 16, wherein the prediction model (7) determines a resultant confidence value for the resultant production parameters (11), which confidence value describes an accuracy with respect to the user product parameter targets (5) of the polymeric product, when the user product parameter targets (5) are applied to the prediction model (7).

19. The method according to claim 16, wherein the prediction model (7) calculates a plurality of resultant production parameters (11) for the production of a plurality of polymeric products associated with the user product parameter targets (5) when the user product parameter targets (5) are applied to the prediction model (7), preferably, that the prediction model (7) determines a resultant confidence value for each of the resultant production parameters (11) when the user product parameter targets (5) are applied to the prediction model (7).

20. The method according to claim 19, wherein the user input comprises a user confidence limit for specifying a minimum confidence value for the resultant production parameters (11) and that the resultant confidence value for each of the resultant production parameters (11) is at least equal to the user confidence limit.

21. The method according to claim 16, wherein the user input comprises a user selection of raw materials from a list of raw materials predefined in the computer system, thereby defining combinations of the raw materials for a polymeric formulation, and that the formulation portions (2) specify raw material portions from the user selection of raw materials.

22. The method according to claim 6, wherein the user selection of raw materials comprises an isocyanate and a polyol, in particular also a blowing agent, preferably, further comprises a chain extender, a cross linker, a catalyst for accelerating the formation of polyurethane, a flame retardant, a pigment and/or a surfactant.

23. The method according to claim 16, wherein the user product parameter targets (5) comprise at least one product parameter bracket for a respective polymeric product property (4), which product parameter bracket defines a subrange within a maximum portion range predefined in the computer system for that polymeric product property (4), and that the computer system applies the user product parameter targets (5) to the prediction model (7) such that for a plurality of product parameter values within each product parameter bracket resultant production parameters (11) are calculated, preferably, that the user product parameter targets (5) comprise at least for one polymeric product property (4) a plurality of non-overlapping product parameter brackets.

24. The method according to claim 23, wherein the user product parameter targets (5) comprise a product parameter resolution for each product parameter bracket, which product parameter resolution defines a step value for varying a product parameter value within the respective product parameter bracket and that the computer system applies the user product parameter targets (5) to the prediction model (7) such that the plurality of product parameter values within each product parameter bracket is determined by varying the product parameter values according to the step value, preferably, that the respective product parameter resolution of two product parameter brackets for the same polymeric product property (4) is different.

25. The method according to claim 16, wherein the process properties, preferably the machine process properties, in particular the user-settable machine process settings, comprise a component temperature, a mixing time, a mixing proportion, a tool temperature, a discharge capacity and/or a line speed.

26. The method according to claim 16, wherein the user product parameter targets (5) consist of a user-selected strict subset of a set of selectable polymeric product properties (4), preferably, that the user input comprises a set of user weighting factors, with each weighting factor associated with a user-selected polymeric product property (4) and describing a relative importance for realizing the corresponding user product parameter target (5).

27. The method according to claim 16, wherein, on the computer system a formulation database (14) is provided comprising test entries (15) for a respective polymeric product, wherein each test entry (15) comprises polymeric product properties data associated with that polymeric product and comprises formulation portions data specifying raw material portions used for the production of that polymeric product and comprises processing parameters data specifying process properties during the production of that polymeric product, preferably, that the prediction model (7) is generated by executing a numerical analysis program, further preferably comprising machine learning, on the formulation database (14).

28. The method according to claim 25, wherein the prediction model (7) defines a multivariable functional relationship with the polymeric product properties (4) as input parameters and the production parameters (1) as output parameters, preferably, that when the computer system generates the prediction model (7), the dependency between the input parameters and the output parameters is based on an fitting algorithm to match the prediction model (7) to the test entries (15) of the formulation database (14).

29. The method according to claim 26, wherein when the computer system generates the prediction model (7), the computer system executes a dimension reduction which involves both a formulation portion dimension and processing parameter dimension, preferably, that the dimension reduction comprises a principal component analysis in order to determine a set of principal components with fewer principal components than the production parameters and that at least one determined principal component comprises both a formulation portion dimension and a processing parameter dimension.

30. A computer system for determining production parameters (1) for the production of a polymeric product, the computer system comprising a computer arrangement (6) and a prediction model (7) for calculating production parameters (1) based on polymeric product properties (4) stored on the computer arrangement (6), which production parameters (1) comprise formulation portions (2) specifying raw material portions for polymeric production and comprise processing parameters (3) specifying process properties during polymeric production, wherein the computer arrangement (6) is configured to receive user input comprising user product parameter targets (5) specifying polymeric product properties (4) and further configured to apply the user product parameter targets (5) to the prediction model (7) to calculate resultant production parameters (11) for the production of a polymeric product associated with the user product parameter targets (5).

Description

[0046] Further advantageous and preferred features are discussed in the following description with respect to the Figures. In the following it is shown in

[0047] FIG. 1 an illustration of the functioning principle of an embodiment of the method according to the invention and in

[0048] FIG. 2 a computer system according to an embodiment of the invention.

[0049] The method according to the invention is used to determine production parameters 1 for the production of a polymeric product which in the present example is polyurethane. More generally, a user of the method according to the invention is seeking to identify a recipe for producing a polymeric product, which is here a polyurethane product. The desired recipe is put in terms of the production parameters 1, which are defined here in terms firstly of formulation portions 2, which describe the respective proportion of raw material used in the production of the polyurethane product. The formulation portions 2 thus comprised by the production parameters 1 are illustrated in a simplified way along one axis only, which is here the x-axis, even though the formulation portions 2 are multi-dimensional.

[0050] This recipe is described secondly in terms of processing parameters 3 also comprised by the production parameters 1, which processing parameters determine the properties of a machine, including the settings of the machine either of a smaller scale in a lab or of larger scale in a factory, used to produce a polyurethane product from the ingredients according to the formulation portions 2. Like the formulation portions 2, also the processing parameters 3 are illustrated in a simplified way along one axis only, which is here the y-axis, even though also the processing parameters 3 are multi-dimensional.

[0051] The formulation portions 2 and the processing parameters 3 together form the production parameters 1 defining such a recipe, though it is possible that the production parameters 1 comprise additional information.

[0052] The user strives to identify particular production parameters 1 such that the resultant polyurethane product exhibits physical characteristics in accordance with user product parameter targets 5 that describe polymeric product properties 4 which are here solidified plastic foam product properties. Such polymeric product properties 4 comprise here a wide range of physical or chemical properties of a polyurethane product, such as density, compressive strength, dimensional stability, thermal resistance, fire performance and emissions. In FIG. 1, the totality of these polymeric product properties 4 is illustrated in a simplified way along one axis—the z-axis—only for the sake of clarity, as was done for the formulation portions 2 and the processing parameters 3. In reality, the polymeric product properties 4 correspond to an array of values and are therefore multi-dimensional.

[0053] The user product parameter targets 5 may arise from a specific customer request or may be determined such that the polyurethane product can be used in a specific component or in some specific way for a larger arrangement.

[0054] Like the polymeric product properties 4 that they describe, the user product parameter targets 5 are here illustrated as a single value along the one-dimensional axis corresponding to the polymeric product properties 4, even though in practice they extend in several directions. Moreover and also in deviation from the illustration of FIG. 1 the user product parameter targets 5 do not specify just a single desired value of polymeric product properties 4, but rather a value bracket. In other words, any value within that value bracket is deemed within the specification according to the user product parameter targets 5.

[0055] Associated with the user product parameter targets 5 and comprised by the user input is a confidence limit. It describes a lower limit for the confidence value, i.e. of the minimum projected likelihood that the user product parameter targets 5 are met by the polymeric product properties 4 of a polymeric product produced according to the production parameters 1 determined by the proposed method. The confidence limit as well as any confidence value associated by any set of production parameters 1 may be expressed as a number.

[0056] The user now provides user input to the computer arrangement 6, shown in FIG. 2, of a computer system. That user input comprises the above user product parameter targets 5 specifying particular desired values of solidified plastic foam product properties 4 and the above confidence limit.

[0057] The computer device 6 of the computer system utilizes a prediction model 7 stored on the computer device 6 in order to calculate production parameters 1 based on the polymeric product properties 4 of the user product parameter targets 5. This prediction model 7 is shown in FIG. 1 in a simplified manner as a function with a one-dimensional input, corresponding to the polymeric product properties 4, and a two-dimensional input consisting of the formulation portions 2 and the processing parameters 3. In reality, both the input and the output of the prediction model 7 can be multi-dimensional.

[0058] For the user production parameter targets 5, the computer device 6 of the computer system now calculates the resultant production parameters 8 based on the prediction model 7 and in particular by applying the user product parameter targets 5 to the prediction model 7. In other words, the user product parameter targets 5 forms the input for the function defined by the prediction model 7 and the resultant production parameters 11 form the corresponding output.

[0059] In the present case and as shown in FIG. 1, there is a region 12 in which the user product parameter targets 5 match predicted polymer product properties 4. Based on the above-mentioned confidence limit, which in the present case requires only a relatively modest likelihood of match, the computer arrangement 6 identifies the region 12 as well as a neighborhood 8 of the region 12 in which a match with the user product parameter targets 5 is also likely.

[0060] The plurality of production parameters 11 which correspond to the region 12 and the neighborhood 8 are output by the computer arrangement 6. As can be seen, the output plurality of production parameters 11 present candidate value brackets both for the formulation portions 2, i.e. resultant formulation portions 9, as well as for the processing parameters, i.e. resultant processing parameters 10.

[0061] The computer device 6 comprises a memory unit 13 in which a formulation database 14 is stored with a plurality of test entries 15 for polyurethane formulations, which has been compiled from historic data concerning the production and testing of polyurethane products. The prediction model 7 has been generated by the computer device 6 performing a multivariate analysis, including a dimension reduction method, e.g. principal component analysis, and a fitting step, on the data of the formulation database 14. In FIG. 1, test polyurethane product properties data 16 for some exemplary test entries 15 of the formulation database 14 are shown. As can be seen, the prediction model 7 substantially corresponds to a best fit approximation for the test polyurethane product properties data 16 of the exemplary test entries 1