Apparatus for generating control data for controlling a polyurethane foam recycling process

Abstract

An apparatus (110) for controlling a foam recycling process is disclosed. A providing unit (111) provides measurement data associated with the foam product. The measurement data is indicative of a spectrum received from foam product. A SAN model providing unit (112) provides a machine learning based SAN model for determining an SAN content in the foam product. The SAN model has been trained utilizing historical training data to determine the SAN content in a foam product based on the measurement data. A determination unit (113) determines the SAN content in the foam product based on the SAN model and the measurement data. A control data generation unit (114) generates control data for controlling the polyurethane foam product recycling process based on the determined SAN content of the foam product.

Description

[0001] Apparatus for generating control data for controlling a polyurethane foam recycling process

[0002] FIELD OF THE INVENTION

[0003] The invention refers to an apparatus, a method and a computer program product for generating control data for controlling at least a part of a polyurethane foam recycling process for recycling polyurethane foam products. Further, the invention refers to a system for generating control data for controlling at least a part of a polyurethane foam product recycling process comprising the apparatus. Moreover, the invention refers to an apparatus, a method and a computer program product for training a machine learning SAN model utilized in the apparatus.

[0004] BACKGROUND OF THE INVENTION

[0005] The recycling of polyurethane foam products, for instance, of a mattress, is a complex process, wherein the quality of the final recycled product strongly depends on accurately controlling the production process.

[0006] SUMMARY OF THE INVENTION

[0007] Generally, end-of-life polyurethane foams included in polyurethane foam products, for instance, mattresses or upholstery, in most cases consist of a mixture of different polyurethane foam types and further can comprise a plurality of additional additive materials. Producing a high-quality recycled product from these waste polyurethane foam products in a chemical recycling process is therefore a challenging task. In particular, the content of sty- rene-acrylonitrile (SAN) particles in the polyurethane foam products can have a major influence on the chemical recycling process and thus also on the quality of the final recycled product. For example, in the presence of SAN particles, a commonly utilized polyurethane foam recycling process is not energetically optimized for the complete conversion of the polyurethane foams. In particular, for different SAN particle content it would be more useful for the chemical recycling process to utilize respective different water content or different residence times in a respective reactor. Moreover, also process steps after the conversion step can be affected by a high SAN content. A high SAN content can then even lead to damaging or fouling respective operation units like heat exchangers or distillation columns. Thus, it could be advantageous for the quality of the final recycled product to know the SAN content in a polyurethane foam product in advance.

[0008] However, currently, no automatic methods exist that allow to determine an SAN content accurately enough before the chemical recycling process and it is common to sort the polyurethane products to be recycled by hand training the respective personnel accordingly, for instance, to distinguish between different products of different vendors for which respective SAN contents are known or to even estimate the SAN content by the texture, feel and look of the polyurethane foam product itself. Clearly, this kind of sorting based on subjective evaluations of the polyurethane foam products of respectively trained personnel is an error- prone, time-consuming, and costly process that still in many cases does not allow to get a final recycled product with a desired high quality. Accordingly, it would be very helpful if an at least semiautomatic process could be provided that allows to determine an SAN content in a polyurethane foam product in a fast, easy to handle, and objective manner that at the same time allows for a sufficient accuracy in the determination of the SAN content for ensuring that a final recycling product can be produced based on the polyurethane foam products with the desired quality.

[0009] In this context the inventors have found that it is advantageous to utilize measurement data indicative of a spectrum from a polyurethane foam product as input to a respectively trained machine learning based model that is trained to utilize the measurement data to predict the respective SAN content of a polyurethane foam product. Since the invention utilizes measurement data indicative of a spectrum from a polyurethane foam product, wherein the polyurethane foam product is subjected to light of a predetermined wavelength range, the measurement data can be acquired very easily, for instance, with a handheld or installed light providing measurement device. Moreover, utilizing this easy measurement in a machine learning based model that is trained to utilize the measurement data to predict the respective SAN content of a polyurethane foam product, allows to determine the SAN content in a fast and suitable accurate manner. Thus, controlling at least a part of the polyurethane foam product recycling process based on the such determined SAN content that has been determined objectively, fast, easily and in a suitable accurate manner, allows for a production of a final recycled product with a desired quality.

[0010] In a first aspect, an apparatus is presented for generating control data for controlling at least a part of a polyurethane foam recycling process for recycling polyurethane foam products, wherein the polyurethane foam product is provided to a recycling process, wherein the provided polyurethane foam product optionally comprises SAN particles and optionally comprises different polyurethane foam types, wherein the apparatus comprises a) a measurement data providing unit for providing measurement data associated with at least a part of the polyurethane foam product, wherein the measurement data is indicative of a spectrum received from the part of the polyurethane foam product when the part of the polyurethane foam product is subjected to light of in predetermined wavelength range, b) an SAN model providing unit for providing a machine learning based SAN model for determining an SAN content in the part of the polyurethane foam product, wherein the SAN model has been trained utilizing historical training data comprising historical measurement data of different polyurethane foam products and a respective SAN content of the polyurethane foam products to determine the SAN content in a polyurethane foam product based on the measurement data, c) an SAN content determination unit for determining the SAN content in the at least the part of the polyurethane foam product based on the SAN model and the measurement data, d) a control data generation unit for generating control data for controlling at least a part of the polyurethane foam product recycling process based on the determined SAN content of the part of the polyurethane foam product.

[0011] Generally, the apparatus can be realized in form of any hardware and / or software, in particular, as any general or dedicated computer. For example, the apparatus can be realized as a computer comprising one or more processors performing the functions defined by the units of the apparatus. However, the apparatus can also be realized in form of a distributed computing network in which one or more communicatively coupled computers comprising one or more processors are utilized for performing the functions as defined by the units of the apparatus. In particular, the units of the apparatus refer to functional units that can be realized in form of one or more processors of one or more communicatively coupled computers.

[0012] The polyurethane foam recycling process can refer to any chemical recycling process that allows for recycling of polyurethane foam. In particular, the polyurethane foam recycling process can comprise a plurality of steps that lead to a production of a respective final recycled product from the provided polyurethane foam products. In particular, the polyurethane recycling process can comprise a pre-processing step, a chemical recycling step and a post-production step. The pre-processing step can, for instance, refer to or comprise the providing of the polyurethane foam products, a sorting of the polyurethane foam products, a pre-processing of the polyurethane foam products, for instance, a chopping or grinding of the polyurethane foam products, etc. The chemical reaction step can comprise any chemical lysis reaction that allows for a recycling of the polyurethane foam product, for instance, the chemical lysis reaction utilized by the recycling process can be a alcoholysis, in particular, a hydrolysis, a glycolysis, a ammonolysis or an aminolysis or combination thereof. The post-processing step can further comprise the steps that are performed after the chemical processing step, for instance, a cooling of a lysis result, a filtration of the lysis result, a portioning of a final recycled product, a washing of a lysis result, etc. Generally, the control data can be configured for controlling any part of the polyurethane foam recycling process, for instance, any of the steps of the polyurethane foam recycling process.

[0013] The polyurethane foam product can be any product comprising or consisting of polyurethane foam. In particular, the polyurethane foam product can comprise SAN particles. However, the polyurethane foam product can also comprise no SAN particles, wherein in this case a determined content of the SAN particles would be zero. Thus, the apparatus can be applied not only to polyurethane foam products that comprise SAN particles, but also to polyurethane foam products that do not comprise SAN particles. Further, the polyurethane foam products can comprise different polyurethane foam types. Generally, a polyurethane foam type is defined by the utilized isocyanate and the molar mass distribution of the utilized polyol. Optionally, a polyurethane foam type can also be defined by utilized additives, stabilizers, etc. Moreover, also different polyols can be utilized for different polyurethane foam types. However, the polyurethane foam product can also comprise or consist of only one polyurethane foam type. Preferably, the polyurethane foam product is at least a part of a mattress or upholstery.

[0014] Generally, the polyurethane foam product can be provided to the polyurethane foam recycling process in a plurality of different forms. For example, the polyurethane foam product can be provided in its complete form, for instance, the complete mattress can be provided to the polyurethane foam recycling process, wherein in this case, respective pre-processing steps are applied before applying the chemical recycling process. However, the polyurethane foam product can also be provided in an already pre-processed form, for instance, comprising only the polyurethane foam parts that are already chopped into smaller pieces that in many cases are easier to transport. In most cases, the to-be recycled polyurethane foam product is a waste product, i.e. a product at the end of its life that has previously been used accordingly. However, the to-be recycled polyurethane foam product can also be a new product, for instance, from overproduction or due to a faulty production. Moreover, the to-be recycles polyurethane foam product can also result from cuttings during the production of a polyurethane foam product.

[0015] In particular, the polyurethane foam product or the part of the polyurethane foam product can refer to foam articles, foam pieces, foam elements or foam flakes. Foam pieces refer to input foam-containing material collected at collection and / or pre-sorting facilities. Whole foam pieces, e.g. mattresses, seats, etc. can be referred to as foam articles. Foam pieces are shredded forming foam elements. Shredding the foam pieces results in foam elements with a maximum dimension typically in the range of 100 mm to 500 mm. Foam elements are milled forming foam flakes. Milling the foam elements results in foam flakes with a maximum dimension typically smaller than 120 mm preferably smaller than 90 mm, even more preferably smaller than 70 mm.

[0016] The measurement data providing unit is then configured for providing measurement data associated with at least a part of the polyurethane foam product. In particular, the measurement data providing unit can be a receiving unit, for instance, an interface configured for receiving data, that is configured for receiving the measurement data from a storage unit on which the measurement data is already stored and for then providing the measurement data. However, the measurement data providing unit can also be realized in form of a storage unit or can be configured to have access to a respective storage unit on which the measurement data is already stored. Moreover, the measurement data providing unit can also be realized as a respective measurement unit, for instance, as an infrared sensor, configured for acquiring the measurement data and then providing the same.

[0017] Generally, the measurement data is associated with at least a part of the polyurethane foam product, in particular, comprises or is indicative of measurements resulting from the respective part of the polyurethane foam product. In particular, the measurement data is indicative of a spectrum received from the part of the polyurethane foam product when the part of the polyurethane foam product is subjected to light in a predetermined wavelength range. The part of the polyurethane foam product can be subjected to light by utilizing a respective light source providing light in the predetermined wavelength range, wherein the provided spectrum of the light will change when interacting with the polyurethane foam product. Generally, the predetermined wavelength range can be in the visible range and / or the infrared range. It has been found by the inventors that in particular the spectrum of infrared light, preferably, near infrared light, is indicative of an SAN content of the respective polyurethane foam product, i.e. different SAN contents result in different infrared spectra received from the polyurethane foam product. Thus, it is preferred that the predetermined wavelength range is part of the near infrared. The spectrum can generally be received as a reflected or refracted spectrum. However, since in most cases the polyurethane foam is not permeable for light in a visible or infrared wavelength range, it is preferred that the spectrum is a reflected spectrum received by a reflection of the light from a surface of the polyurethane foam product. Preferably, the part of the polyurethane foam product is subjected to near infrared light with wavelengths between 950 and 1700 nm. More preferably, in a range between 600 nm to 800nm. However, it has been found that also infrared light in the mid infrared range can be utilized. Moreover, also visible light wavelength ranges can be utilized. In this case it is preferred that the spectrum comprises an image of a part of the polyurethane foam product, in particular, a microscopic image. Thus, the spectrum can be a distribution of measured light intensities with respect to the wavelength, i.e. an intensity spectrum, or can be a distribution of intensity of a respective wavelength in space, i.e. an image in a predefined wavelength range. Preferably, the spectrum is an intensity spectrum.

[0018] Generally, the measurement data is indicative of the spectrum, in particular, can directly comprise the result of the measurement of the spectrum. However, the result of the measurement of the spectrum can also first be processed, for instance, filtered to reduce noise, before being provided as measurement data, for instance, by the measurement data providing unit. For subjecting the polyurethane foam product with light and for measuring the resulting light spectrum, generally, the same device can be utilized that comprises a light source and also a respective light sensor. However, also different devices can be provided. The respective light source and / or light sensor can be a handheld device or a device integrated, for instance, into an automated pre-processing unit in which the polyurethane foam is pre-processed, without human intervention. For example, the light source and / or sensor can be provided such that they illuminate polyurethane foam products moving on a conveyor belt.

[0019] The apparatus further comprises an SAN model providing unit configured to providing a machine learning based SAN model. The SAN model providing unit can also be a receiving unit for receiving the SAN model, for instance, via a respective interface such as an interface to the storage or an input unit and to then provide the SAN model. However, the SAN model providing unit can also be the storage unit itself or the respective input unit itself. Generally, the SAN model provided by the SAN model providing unit can be trained already and then stored on a respective storage, for instance, to which the SAN model providing unit has access. However, the providing of the SAN model can also comprise initiating a training or retraining of a respective machine learning based model for a specific case based on respective training data, wherein the SAN model is then provided after the training process is completed.

[0020] The SAN model is configured for determining an SAN content in at least the part of the polyurethane foam product that has been subjected to the infrared light. For most cases, the SAN content of a part of the polyurethane foam product will also be the content in other parts of the polyurethane foam product, in particular, the whole polyurethane foam product, since it is common that a homogeneous polyurethane foam product is already provided to the recycling process. However, if a heterogeneous polyurethane foam product is provided to a recycling process, the same procedure as described here, i.e., the measurement of the spectrum and the determination of the SAN content, can also be applied to different parts of a polyurethane foam product. The provided SAN model is then adapted to determine the SAN content of a respective polyurethane foam product based on the measurement data. In particular, the SAN model is a machine learning based model that has been parameterized, i.e. trained, based on historical training data, to determine the SAN content based on the measurement data. Generally, the SAN model can be based on any known machine learning algorithm, like a neural network, a regression model, or a classification algorithm, etc. However, it has been found by the inventors that in particular, an SAN model comprising a vector support machine provides accurate results while the training and also the application efforts, in particular, with respect to computational resources, is in an acceptable range to still allow for a fast determination of the SAN content, in particular, for a real-time determination of the SAN content that allows for a real-time controlling of the polyurethane recycling process, for instance, for a real-time sorting of respective parts of a polyurethane foam product. However, also other machine learning based models like neural networks can be trained for this task. The provided SAN model has generally been trained before being provided by the SAN model providing unit by utilizing historical training data comprising historical measurement data of different polyurethane foam products and respective SAN contents of the plurality of different polyurethane foam products. For example, for different polyurethane foam products an SAN content can be measured in a laboratory or polyurethane foam products with known SAN content can be utilized. Respective measurement data can then be acquired in the same way as will be the case during the application of the model, for instance, by also using a handheld device or a respective process-integrated device and a respective trainable model can then be trained by parameterizing the respective parameters of the model utilizing known training methods. For example, supervised but also unsupervised training methods can be utilized, respectively.

[0021] The SAN content determination unit is then configured to determine based on the provided SAN model and the measurement data the SAN content in at least the part of the polyurethane foam product for which the measurement data has been acquired. In particular, the acquired measurement data is provided as input to the SAN model, wherein the SAN model then provides as output the respective SAN content. Generally, the determined SAN content can refer to a specific SAN content value, like the amount of SAN particles per weight. However, the determined SAN content can also be determined as a percentage value based on a reference of a Standard SAN content of known mixtures of respective polyurethane foam types. Preferably, the SAN content is provided as the amount of SAN particles per volume. In particular, the determined SAN content can also be a simple indicator whether the SAN content lies above or below one or more predetermined thresholds. In the last case the SAN content determination can also be regarded as a classification of the polyurethane foam product into classes defined by the SAN content.

[0022] The control data generation unit can then generate control data for controlling at least the part of the polyurethane foam product recycling process based on the determined SAN content of the part of the polyurethane foam product. For example, predetermined rules can be utilized that determine how respective process parameters are to be set based on the SAN content, wherein the control data can then be generated such that it causes a setting of the respective parameters. Preferably, the control data is configured for controlling a sorting process being part of the polyurethane foam product recycling process such that different polyurethane foam products and / or parts of a polyurethane foam product are sorted based on the SAN content of the respective polyurethane foam product and / or parts of the polyurethane foam product. For example, the control data can cause information indicative of the determined SAN content of a polyurethane foam product to be provided to an interface utilized by an employee that sorts the respective polyurethane foam products based on their SAN content. The interface can then indicate to the employee how the polyurethane foam product has to be sorted based on the determined SAN content, for instance, by providing a visual or audio signal. Additionally or alternatively directly the determined SAN content can be provided to the employee via the interface so that the employee can decide or control the sorting of the polyurethane foam product. However, the control data can also be configured to cause an automatic sorting of the polyurethane foam product based on the SAN content. For example, a mechanic sorting using pressurized air or other machine sorting mechanisms can be controlled utilizing the control data such that respectively provided polyurethane foam products are sorted automatically based on the SAN content. For example, one or more thresholds can be provided with respect to the SAN content such that the polyurethane foam products are sorted based on whether they are above or below a respective SAN content threshold. This allows already in a sorting process to provide a more homogeneous input to the respective chemical recycling process such that the chemical recycling process can be adapted more optimally to the respective homogeneous material with respect to the SAN content.

[0023] Additionally or alternatively, the control data is configured for controlling process parameters of a chemical lysis reaction being part of the recycling process based on the SAN content. In particular, reaction temperatures, reaction pressure, residence time, and / or reagents concentration, can be adapted based on the SAN content. For example, also in this case predefined rules can be utilized that determine the process parameters with respect to the SAN content. In particular, it is preferred that based on the SAN content it is determined to utilize one of two possible process parameter sets for the chemical recycling process with respective two sets of parameters. Also in this case, a respective threshold for the SAN content can be provided based on which it is decided whether a polyurethane foam product is subjected to the first set of process parameters or to the second set of process parameters based on the SAN content of the respective polyurethane foam product.

[0024] In an embodiment, the apparatus further comprises a) a classification model providing unit for providing a foam classification model for classifying the at least a part of the polyurethane foam product into a class of one or more foam classes, wherein the foam classes depend on the polyurethane foam types and / or further additives in a polyurethane foam product, and wherein the classification model is a machine learning based model trained utilizing historical data to classify a polyurethane foam product into a class of the one or more classes based on the measurement data, b) a classification unit for classifying the at least a part of the polyurethane foam product based on the classification model and the measurement data, c) wherein the control data generation unit is configured for generating control data for controlling at least a part of the polyurethane foam product recycling process further based on the classification of the part of the polyurethane foam product. Generally, the classification model providing unit can be a receiving unit for receiving the classification model, for instance, from a storage or from an input unit, but can also be a storage unit itself or comprise access to a respective storage unit on which the classification model is already stored for providing the same. The classification model is a machine learning based model that has been trained such that it is configured to classify a polyurethane foam product, in particular, at least the part of the polyurethane foam product for which the measurement data is provided, into a class of the one or more classes based on the measurement data that is also utilized for determining the SAN content. Thus, the same measurement data cannot only be utilized for determining the SAN content, but provided as input to another model, here, the foam classification model, for classifying the polyurethane foam products. In a preferred embodiment, the classification model comprises a support vector machine.

[0025] Generally, the polyurethane foam product classes, i.e. the foam classes, depend on the polyurethane foam types and / or further additives present in the polyurethane foam product. The classes can be predefined and the foam classification model can be trained to classify a polyurethane foam into respective predefined foam classes. Predefined foam classes can, for example, be based on possible differences between the classes with respect to the optimal recycling process parameters. However, the classes can also be identified during the training itself, and thus refer to classes that have not been predefined. Depending on the definition of the classes, the foam classification model can be configured to classify a polyurethane foam product only in one class, respectively, or in a plurality of classes. For example, if two of the classes refer to “a polyurethane foam product with flame protection” and another class refers to “standard polyurethane foam product”, the polyurethane foam product can be classified into both classes if it refers to a standard polyurethane foam product with flame protection. However, if, for instance, the classes for classifying the polyurethane foam products refer to only two classes like “standard” and “not standard”, a polyurethane foam product can only be classified into one of these classes. Generally, the classes can depend on the polyurethane foam types utilized in the polyurethane foam product and thus on the mixture of polyurethane foam types present in the polyurethane foam product. Additionally or alternatively, the foam classes can depend on any further additives provided in the polyurethane foam product, for instance, on functional additives that are utilized to adapt one or more functions of the polyurethane foam product. An example for such additives are additives that are utilized to provide a flame protection to the polyurethane foam product. Further possible additives can be utilized for classifications are water, silicone-stabilizers, like siloxanes with polyetherol-sidechains, crosslinker, like glycerin, diethanolamine, sorbitol, chain-extenders, like Butandiol, Methyl-propanediol, monools, like phenoxyethanol, antioxidants, flame retardants, inorganic fillers like chalk, defoamers, catalysts, colourants. Moreover, also other aspects of the polyurethane foam product can be taken into account as classes, for example, dirt, dust, humidity, microbial or fungal contaminations. For example, a respective threshold can be provided for any of these contaminations and a polyurethane foam product lying above the respective threshold with at least one of these contaminations can be classified as contaminated.

[0026] For training the classification model, historical data can be utilized comprising measurement data of a plurality of polyurethane foam products and a respective classification for the respective polyurethane foam products. For example, the classification can be provided by an expert as predetermined classification or can be determined during the training process based on the measurement data. Then, any known training method can be utilized for training the classification model based on the historical data. Preferably, the one or more classes comprise at least one of the following classes: HR-MDI, HR-TDI, HR-TDI CM, HyperSoft-MDI, HyperSoft-TDI, Standard, Standard-CME, Standard-HR-mix-TDI, VE- MDI,more preferably, TDI-Standard and no-Standard. The term “TDI” indicates that the polyurethane foam comprises toluene diisocyanate. The term “CM” stands for combustion modified and “CME” for combustion modified ether. The class “Standard-TDI” refers to a flexible foam comprising polyols of different molecular weight in the range 3000 to 3500 and toluene diisocyanate. The term “MDI” indicates that the polyurethane foam comprises 4,4-diphenylmethane diisocyanate. The term “HR” refers to high resilient foams comprising polyols of different molecular weight in the range 5000 to 12000 with MDI and / or TDL The term “VE” refers to viscoelastic foams comprising polyols of different molecular weight in the range 700 to 2000. A further possible classification can refer to the classes TDI- Standard, HR or VE. Moreover, the classes can also consist of “TDI-standard” and “no standard” in an embodiment.

[0027] The classification unit is then configured to classify at least the part of the polyurethane foam product based on the classification model and the measurement data. In particular, the classification unit is configured to provide the measurement data as input to the classification model, wherein the classification model then outputs the one or more respective classes into which the polyurethane foam product can be classified by the classification model. The control data generation unit is then configured to generate control data for controlling at least the part of the polyurethane foam product recycling process further based on the classification of the part of the polyurethane foam product. In particular, in addition to sorting the polyurethane foam products in a pre-processing step based on the SAN content, the polyurethane foam products can also be sorted based on the classification. This allows for an even higher homogenization of the polyurethane foam products provided to the recycling process and thus allows to strongly adapt the recycling process, in particular, the chemical recycling process to the respective input polyurethane foam products, for instance, not only to the SAN content but also to the respective type of polyurethane foam products. This can lead to a higher quality of the resulting final recycled product. Moreover, the control data generation unit can also generate control data that can control, for instance, the chemical recycling process, in particular, the chemical lysis process to adapt this process to the polyurethane foam product classes provided to the recycling process. For example, one or more predetermined parameter sets for the respective chemical lysis process can be set by the control data based on the respective SAN content and further based on the respective polyurethane foam product class in the recycling process.

[0028] In a further aspect of the invention, a system is presented for generating control data for controlling at least a part of a polyurethane foam product recycling process, wherein the system comprises a) a measurement device configured for illuminating at least a part of a polyurethane foam product with light in a predetermined wavelength range and measuring a spectrum received from the part of the polyurethane foam product illuminated with the light for generating and providing measurement data indicative of the measurement, and b) an apparatus according to any of the preceding claims. In a further aspect of the invention, an apparatus is presented for training a machine learning SAN model utilizable in the apparatus as described above, wherein the apparatus comprises i) a historical data providing unit for providing historical data for training the SAN model, wherein the historical data comprises a) measurement data of a plurality of polyurethane foam products with different SAN contents, and b) a respective SAN content for each of the plurality of polyurethane foam products, wherein the measurement data is indicative of a spectrum received from at least a part of a respective polyurethane foam product when the part of the respective polyurethane foam product is subjected to light in a predetermined wavelength range, ii) a model providing unit for providing a trainable machine learning based SAN model, iii) a training unit for training the trainable SAN model based on the historical data such that the trained SAN model is configured to determine the SAN content of a polyurethane foam product based on the respective measurement data of the polyurethane foam product, and iv) a trained model providing unit for providing the trained SAN model.

[0029] In a further aspect, a computer-implemented method is presented for generating control data for controlling at least a part of a polyurethane foam recycling process for recycling polyurethane foam products, wherein the polyurethane foam product is provided to a recycling process, wherein the provided polyurethane foam product optionally comprises SAN particles and optionally comprises different polyurethane foam types, wherein the method comprises a) providing measurement data associated with at least a part of the polyurethane foam product, wherein the measurement data is indicative of a spectrum received from the part of the polyurethane foam product when the part of the polyurethane foam product is subjected to light in a predetermined wavelength range, b) providing a machine learning based SAN model for determining an SAN content in the part of the polyurethane foam product, wherein the SAN model has been trained utilizing historical training data comprising measurement data of different polyurethane foam products and a respective SAN content of the polyurethane foam products to determine the SAN content in a polyurethane foam product based on the measurement data, c) determining the SAN content in the at least a part of the polyurethane foam product based on the SAN model and the measurement data, d) generating control data for controlling at least a part of the polyurethane foam product recycling process based on the determined SAN content of the part of the polyurethane foam product.

[0030] In a further aspect, a computer-implemented method is presented for training a machine learning SAN model utilizable in the apparatus as described above, wherein the method comprises i) providing historical data fortraining the SAN model, wherein the historical data comprises a) measurement data of a plurality of polyurethane foam products with different SAN contents, and b) a respective SAN content for each of the plurality of polyurethane foam products, wherein the measurement data is indicative of a spectrum received from at least a part of a respective polyurethane foam product when the part of the respective polyurethane foam is subjected to light in a predetermined wavelength range, ii) providing a trainable machine learning based SAN model, iii) training the trainable SAN model based on the historical data such that the trained SAN model is configured to determine the SAN content of a polyurethane foam product based on the respective measurement data of the polyurethane foam product, and iv) providing the trained SAN model.

[0031] In a further aspect, a polyurethane recycling process is presented for recycling a polyurethane foam product, wherein the polyurethane recycling process is controlled utilizing control data generated by the apparatus as described above and / or the method as described above.

[0032] In a further aspect, a computer program product is presented for generating control data for controlling at least a part of a polyurethane foam product recycling process for recycling a polyurethane foam product, wherein the computer program product comprises program code means for causing an apparatus as described above to carry out the method as described above.

[0033] In a further aspect, a computer program product is presented for generating control data for training a machine learning SAN model utilizable in the apparatus as described above, wherein the computer program product comprises program code means for causing an apparatus as described above to carry out the method as described above.

[0034] It shall be understood that the apparatuses as described above, the methods as described above and the computer program products as described above have similar and / or identical preferred embodiments, in particular, as defined in the dependent claims.

[0035] It shall be understood that a preferred embodiment of the present invention can also be any combination of the dependent claims or above embodiments with the respective independent claim.

[0036] These and other aspects of the invention will be apparent from and elucidated with reference to the embodiments described hereinafter. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] In the following drawings:

[0038] Fig. 1 shows schematically and exemplarily an embodiment of a system for recycling polyurethane foams comprising an apparatus for generating control data for controlling at least a part of the polyurethane recycling process,

[0039] Fig. 2 shows schematically and exemplarily a flow chart of a method for generating control data for controlling at least a part of a polyurethane recycling process,

[0040] Fig. 3 shows schematically and exemplarily a flow chart of a method for training an SAN model for determining an SAN content.

[0041] DETAILED DESCRIPTION OF EMBODIMENTS

[0042] Fig. 1 shows schematically and exemplarily a system 100 for recycling a polyurethane foam product, in particular, a mattress or upholstery product or a part thereof. The system 100 comprises industrial assets utilizable for performing a polyurethane recycling process, for instance, an industrial plant 130. Further, the system 100 comprises an apparatus 1 10 for generating control data for controlling at least a part of the polyurethane foam recycling process performed in the industrial plant 130. Optionally, the system 100 can further comprise a training apparatus for training an SAN model utilizable by the apparatus 1 10.

[0043] The following polyurethane foam product recycling process is described as being performed as part of an industrial plant 130. However, parts of the process can also be performed by a network of industrial plants at the same or at different industrial plant sites. In an exemplary polyurethane foam product recycling process, first a to be recycled polyurethane foam product 140 is provided to the industrial plant site 130. In particular, the to be recycled polyurethane foam product 140 can be part of a mattress or an upholstery. Generally, the polyurethane foam product 140 can be provided in any form suitable for the respective recycling process, for instance, can be provided as complete product such that pre-processing steps like the sorting of different materials, for instance, of a mattress, are performed at the industrial plant 130 or can be provided as already only comprising the respective polyurethane foam parts, for instance, in form of chunks of a predetermined size. The provided polyurethane foam product 140 is then provided to a measurement unit 133 of the industrial plant site 130. The measurement unit 133 can be realized in form of a handheld device utilized by a respective employee or can also be part, for example, of an automatic measurement asset. For example, the measurement unit 133 can be placed above some kind of movement unit like a conveyor belt on which the respective polyurethane foam product is moved during the measurement. This automatic measurement is in particular preferred for cases in which the polyurethane foam product is provided in smaller pieces, for example, as shredded flakes. The measurement unit 130 is configured to subjecting at least a part of the polyurethane foam product to light in a predetermined wavelength range. In a preferred embodiment described in the following the light is infrared light and the light source is a respective infrared source. Preferably, the infrared light refers to near infrared light. Further, the measurement unit 133 comprises means for detecting light, in particular, in form of a spectrum, received, for example, reflected or refracted, from the polyurethane foam product 140 when subjected to the infrared light. For example, the measurement unit 133 can comprise an infrared detector for measuring the infrared spectrum reflected from the polyurethane foam product 140. The measured spectrum can then, for instance, be digitalized or formatted in any other way that allows for a further processing of the measurement data and then provided to a respective storage unit, not shown, or directly to apparatus 110. Based on a respective result of an SAN content determination or a respective classification of the polyurethane foam product, as will be described hereafter, with respect to the apparatus 110, the polyurethane foam product 140 or parts thereof can be transferred to a storage facility 141 for storing the polyurethane foam product or the polyurethane foam product can be provided to a sorting facility 142 that is configured for sorting the polyurethane foam product based on the respective results of the determination of the SAN content or the classification of the polyurethane foam product based on the apparatus 110. In this case, polyurethane foam products that can be recycled, for instance, by the current process performed by the industrial plant site or with the currently utilized process parameters of the recycling process can be directly transferred further to the industrial asset 131 performing the recycling process. Polyurethane foam products for which the determined SAN content and / or the classification indicates the recycling at a later point, for instance, with different processing parameters, can then also be transferred from the sorting facility 142 to the storage unit 141 . The polyurethane foam recycling process is then performed by respective industrial assets 131 that are controlled by control means 132. Control means 132 can generally be part of a respective industrial plant control and management system and are configured to manipulate one or more process parameters determining the polyurethane recycling process performed by the industrial assets 131 .

[0044] A respective exemplarily polyurethane recycling process will be described in the following. An example of typical pre-processing steps for foam recycling, in particular for polystyrene foams, are summarized in the website https: / / www.recyclefoam.org / about-foam-recycling. Foam residues can either be placed in the same container as other recyclables or can be taken to a dedicated drop-off center. From there, the recyclables can be delivered to a sorting facility or material recovery facility (MRF) where the foam is either separated from other recyclables (mechanically or manually) or is kept separated from the other recyclables. The foam pieces are delivered to a grinder or shredder from where they are transferred to a densifier or press, where the loose foam elements are compacted for storage or transportation to the recycling facility. Another example of the steps of a pre-processing method for a polystyrene foam mix can be found in the website https: / / www.homefor- foam.com / recycling / . An exemplary chemical processing that can follow the pre-processing steps described above is described in the scientific publication by M. Grdadolnik et al. titled Insight into Chemical Recycling of Flexible Polyurethane Foams by Acidolysis, ACS Sustainable Chem. Eng. 2022, 10, 3, 1323-1332 describes an exemplary chemical recycling process for polyurethane foams. Other examples of chemical recycling processes are disclosed, for instance in documents DE 102016122275 A1 , DE 102013106364 A1 , WO 2021023889 A1 , and US 20220251328 A1. These respective chemical processes can be performed utilizing respective chemical operation units, like a reactor a filtration unit, a distillation unit, etc. The resulting one or more recycling products can then be subjected to respective post-processing steps, including quality checking, packaging, etc. The above described polyurethane recycling process then results in a respective final recycled product 143.

[0045] The apparatus 1 10 is configured to provide control signals, for example, to the controlling unit 132 and / or to the sorting facility 142, for controlling the respective polyurethane foam product recycling process. The apparatus 110 can be realized, for instance, as part of the control unit 132 but can also be realized as a standalone computer system that is communicatively coupled, for instance, with the control unit 132. Generally, the apparatus 110 can be realized in form of any computing device comprising respective processors that can perform the functions defined by the units of the apparatus 110. For example, the apparatus 110 can be realized in form of a general or dedicated computing device or network of computing devices. The apparatus 110 comprises a measurement data providing unit 111 , an SAN model providing unit 1 12, an SAN content determination unit 113 and a control data generation unit 114. Optionally, the apparatus 110 can further comprise, not shown in Fig. 1 , a classification model providing unit and a classification unit.

[0046] The measurement data providing unit 11 1 is configured to provide the respective measurement data of the measurement unit 113. For example, the measurement data providing unit 111 can be an interface that allows to receive the data from the measurement unit 133 and to provide the received data from the measurement unit 133, for instance, to the SAN content determination unit 1 13. However, the measurement data providing unit 111 can also be configured to access a storage unit on which the measurement data of the measurement unit 133 is already stored, for instance, in a long-term or short-term storage.

[0047] The SAN model providing unit 112 is configured to provide a machine learning based SAN model for determining an SAN content in the part of the polyurethane foam product for which the measurement data has been provided. Generally, the SAN model is configured for determining the SAN content of a respective polyurethane foam product based on the provided measurement data. The SAN model is machine learning based and has been trained utilizing historical training data comprising a) historical measurement data of different polyurethane foam products and b) a respective SAN content of the polyurethane foam products. For example, the SAN model can be trained utilizing training apparatus 120, described in the following.

[0048] Training apparatus 120 comprises a historical data providing unit 121 , a model providing unit 122, a training unit 123 and a trained model providing unit 124. The historical data providing unit 121 is configured for providing historical data for training the SAN model. For example, the historical data can be stored on a respective storage unit not shown in Fig. 1 . Generally, the historical data comprises a) measurement data of a plurality of polyurethane foam products with different SAN contents, and b) the respective SAN content for each of the plurality of polyurethane foam products. For example, the historical data can comprise data that has been acquired from polyurethane foam products with known SAN contents. For example, the SAN contents of a plurality of polyurethane foam products can be measured utilizing known measurement methods and from the same polyurethane foam products respective measurement data, e.g. infrared spectra, can be acquired. However, in some cases, the SAN content can also be generally known, for instance, provided and guaranteed by the vendor or producer of the polyurethane foam product. In this case, only the respective measurement data of the polyurethane foam product has to be acquired. The historical data can then be stored on a respective storage unit.

[0049] The model providing unit 122 is configured for providing a trainable machine learning based SAN model. In particular, a respective trainable machine learning based SAN model can be stored on a storage unit already and be provided by the model providing unit. However, the model providing unit can also be configured to receive the trainable machine learning based SAN model, for instance, via an input unit or user interface where a user can input or select a respective trainable machine learning based SAN model. The trainable SAN model refers to a respective algorithm comprising parameters that are determined during the training process of the trainable SAN model and thus reflect in the trained SAN model the functional relation between the measurement data and the SAN content. Thus, the training of the trainable SAN model can also be regarded as a parameterization, i.e. a setting of the parameters, of the trainable SAN model. Preferably, the trainable SAN model comprises a support vector machine, since the inventors have found that such an algorithm allows for a suitable accuracy in the determination of the SAN content with acceptable utilization of computational resources. However, also other known machine learning algorithms can be utilized, for instance, neural network algorithms, linear or nonlinear regression algorithms, classification algorithms, etc.

[0050] The training unit 123 is then configured for training the trainable SAN model utilizing historical data. In particular, any known training algorithms, for instance, supervised or unsupervised training algorithms, can be utilized for training the SAN model based on the historical data. In particular, the SAN model is trained utilizing the historical data such that the trained SAN model is configured to determine the SAN content of a polyurethane foam product based on the respective measurement data of the polyurethane foam product with a predetermined accuracy. The trained model providing unit 124 is then configured for providing the trained SAN model. For example, the trained SAN model can be provided to a storage unit from which the SAN model providing unit 1 12 can access the SAN model or if the SAN model has currently been trained or retrained, the trained model providing unit 124 can directly provide the SAN model to the trained SAN model providing unit 1 12. In a preferred embodiment, the machine learning based SAN model is a support vector machine with a radial basis function. It has been found that such a model can be trained successfully utilizing between 100 and 200 historical data samples.

[0051] The SAN content determination unit 113 is then configured for determining the SAN content in at least the part of the polyurethane foam product. In particular, the SAN content determination unit is configured to utilize the provided SAN model and the provided measurement data to determine the respective SAN content. For example, the SAN content determination unit 1 13 can be configured to provide the measurement data as input to the respective SAN model such that the output refers to the SAN content. Optionally, the SAN content determination unit 113 can also be configured for pre-processing the respective measurement data, for instance, by filtering out strong outliers that might impact the results of the SAN content determination, by removing parts of a spectrum that have been found to be irrelevant for the determination of the SAN content, etc. However, such pre-processing can also be omitted if the SAN model has been trained on the raw data or can be performed by a respective part of the SAN model itself. It has been found by the inventors that the results of the SAN content determination allow for a good accuracy if the pre- processing of the measurement data, e.g. the infrared spectra, comprises first a detrending step in order to linearize the respective data, further a normalization, for instance, utilizing a standard normal variant algorithm, and a calculation of a second derivative from the respective measurement data. Moreover, it has been found that the most relevant part of the spectrum for determining the SAN content lies in the near infrared, in particular, between a wavelength of 1100 nm to 1600 nm measured with NIR spectroscopy. Although the inventors have been found that this pre-processing was in particular suitable for the preferably utilized support vector machine, when utilizing other machine learning algorithms also other pre-processing steps or even an omitting of pre-processing steps can result in a suitable accuracy in the performance of the respective machine learning model.

[0052] The control data generation unit 114 is then configured to generate control data for controlling at least a part of the polyurethane foam product recycling process performed by industrial plant 130 based on the determined SAN content of the part of the polyurethane foam product 140. For example, the control data can be configured for controlling the sorting of the polyurethane foam products, for instance by a sorting facility 142. In particular, it can be useful to setup the chemical recycling process performed by the process assets 131 for a specific SAN content and to then sort the polyurethane foam products until a batch of polyurethane foam products with the respective SAN content can be provided to the recycling process. Polyurethane foam products with a different SAN content can then be transferred to a respective storage unit 141 and recycled in another batch with the respective SAN content. Moreover, the sorting can also refer to mixing polyurethane foam products with different SAN content, for example, from the storage unit 141 , such that a respective predetermined SAN content is provided by a batch of polyurethane foam products processed in the chemical recycling process at the same time. This allows to also recycle polyurethane foam products, for instance, with a high SAN content that are often difficult to recycle, when mixed together with polyurethane foam products with a very low or no SAN content at all. Additionally or alternatively, the control data can be configured for controlling process parameters of the chemical recycling process, in particular, of the chemical lysis reaction. For example, the control data can be provided to the controlling unit 132 and cause the control unit 132 to adapt a respective process parameter of the chemical lysis reaction in accordance with the controlling data. For example, predetermined rules can be utilized that indicate how process parameters are to be changed based on the SAN content.

[0053] Preferably, the controlling of the sorting of the polyurethane foam products before being provided to the chemical reaction process and the controlling of the chemical reaction process are combined. For example, it is preferred that the control data causes a sorting of the polyurethane foam products into two batches, wherein a respective first batch comprises a low SAN content, i.e. an SAN content below a predetermined threshold, and the second batch comprises a high SAN content, i.e. an SAN content above a predetermined threshold, wherein the control data then further causes that a respective recycling process is utilized for each batch. This allows to produce a final recycling product 143 with a desired quality even in the presence of different SAN contents.

[0054] In a preferred embodiment, the apparatus 110 further comprises a classification model providing unit that is configured for providing a foam classification model. The foam classification model is configured for classifying the polyurethane foam product 140 into different polyurethane products classes. Generally, a polyurethane foam product can comprise more than one polyurethane foam type and / or can comprise further additives to the polyurethane foam. Thus, classes can be defined that classify the polyurethane foam products based on the polyurethane foam types and / or further additives in the polyurethane foam product. Generally, the polyurethane foam types and / or further additives in the polyurethane foam product can also influence the recycling process and thus influence the quality of the recycled product. Accordingly, a classification allows in the same way as a determination of the SAN content, to sort the polyurethane foam products based on the respective predetermined classes such that only recycling processes are applied to a batch of polyurethane foam products in the same class that are suitable for this class and / or the processing parameters of the chemical recycling process can be adapted to the determined polyurethane foam class in order to allow the production of a recycling product with a desired quality.

[0055] Also the classification model is preferably a machine learning based model that is trained utilizing historical data comprising a) measurement data of a plurality of respective polyurethane foam products and b) the respectively determined class for each of the different polyurethane foam products. For example, for training the classification model, the same polyurethane foam products can be utilized that have been utilized for training the SAN model, wherein in this case the measurement data is already provided. However, also a different batch of polyurethane foam products can be utilized for the training, wherein in this case first the measurement data has to be acquired. The class of a respective polyurethane foam product can then be determined, for instance, based on information of a producer or vendor of the polyurethane foam product indicating the respective polyurethane foam types utilized in the polyurethane foam product or also based on respective laboratory measurements of the polyurethane foam types. Also the classification model is preferably based on a support vector machine algorithm and the training can be performed based on the historical data in the same way as described above with respect to the SAN model. A classification unit can then be configured to utilize the classification model to classify a part of the polyurethane foam product based on the measurement data. The control data generation unit can then be further configured to generate the control data not only based on the SAN content, but also based on the classification of the polyurethane foam product. As described above, for instance, the control data can be configured for causing a respective sorting of the polyurethane foam product or a part of the polyurethane foam product or can be configured to cause the utilization of a respective recycling process, for instance, by utilizing respective process parameters. Preferably, the classes utilized by the classification model refer to a standard class and a non-standard class. However, the classes can also be more diversified, for instance, the following classes or at least one or more of the following classes can be utilized HR-MDI, HR-TDI, HR-TDI CM, HyperSoft-MDI, HyperSoft- TDI, Standard, Standard-CME, Standard-HR-mix-TDI, VE-MDI,more preferably, TDI- Standard and no-Standard. The term “TDI” indicates that the polyurethane foam comprises toluene diisocyanate. The class “Standard-TDI” refers to a flexible foam comprising polyols of different molecular weight in the range 3000 to 3500 and toluene diisocyanate. The term “MDI” indicates that the polyurethane foam comprises 4,4-diphenylmethane diisocyanate. The term “HR” refers to high resilient foams comprising polyols of different molecular weight in the range 5000 to 12000 with MDI and / or TDI . The term “VE” refers to viscoelastic foams comprising polyols of different molecular weight in the range 700 to 2000. A further possible classification can refer to the classes TDI-Standard, HR or VE.

[0056] Fig. 2 shows schematically and exemplarily a flow chart of a method for generating control data for controlling at least a part of a polyurethane recycling process for recycling polyurethane foam products. In particular, the method can be performed by the apparatus 110 as described with respect to Fig. 1 . In particular, the method 200 comprises a first step of providing measurement data associated with at least a part of the polyurethane foam product as described, for instance, with respect to the measurement data providing unit 111 of Fig. 1 . In a next step, a machine learning based SAN model is provided for determining the SAN content in a part of the polyurethane foam product, for instance, in accordance with the principles described above with respect to the SAN model providing unit 112 and the training apparatus 120. In a further step, the SAN content is determined for the polyurethane foam product based on the SAN model and the measurement data. In a last step, control data is generated for controlling at least the part of the polyurethane foam product recycling process based on the determined SAN content of the part of the polyurethane foam product.

[0057] Fig. 3 shows schematically and exemplarily a training method 300 for training a machine learning based SAN model, for instance, in accordance with the principles described with respect to the training apparatus 120 in Fig. 1 . In a first step, the method 300 comprises providing historical data for training the SAN model, wherein the historical data comprises a) measurement data of a plurality of polyurethane foam products with different SAN content and b) a respective SAN content for each of the plurality of polyurethane foam products. Further, the method 300 comprises a step of providing a respectively trainable machine learning based SAN model. In a next step, the trainable machine learning based SAN model can be trained utilizing the historical data such that the trained SAN model is configured to determine the SAN content of a polyurethane foam product based on the respective measurement data of the polyurethane foam product. In a last step, the trained SAN model can then be provide.

[0058] Other variations to the disclosed embodiments can be understood and effected by those skilled in the art in practicing the claimed invention, from a study of the drawings, the disclosure, and the appended claims.

[0059] For the processes and methods disclosed herein, the operations performed in the processes and methods may be implemented in differing order. Furthermore, the outlined operations are only provided as examples, and some of the operations may be optional, combined into fewer steps and operations, supplemented with further operations, or expanded into additional operations without detracting from the essence of the disclosed embodiments.

[0060] In the claims, the word "comprising" does not exclude other elements or steps, and the indefinite article "a" or "an" does not exclude a plurality.

[0061] A single unit or device may fulfill the functions of several items recited in the claims. The mere fact that certain measures are recited in mutually different dependent claims does not indicate that a combination of these measures cannot be used to advantage.

[0062] Procedures like the providing of the measurement data and the SAN model, the determining of the SAN content, the generating of control data, etc. performed by one or several units or devices can be performed by any other number of units or devices. These procedures can be implemented as program code means of a computer program and / or as dedicated hardware.

[0063] A computer program product may be stored / distributed on a suitable medium, such as an optical storage medium or a solid-state medium, supplied together with or as part of other hardware, but may also be distributed in other forms, such as via the Internet or other wired or wireless telecommunication systems.

[0064] Any units described herein may be processing units that are part of a classical computing system. Processing units may include a general-purpose processor and may also include a field programmable gate array (FPGA), an application specific integrated circuit (ASIC), or any other specialized circuit. Any memory may be a physical system memory, which may be volatile, non-volatile, or some combination of the two. The term “memory” may include any computer-readable storage media such as a non-volatile mass storage. If the computing system is distributed, the processing and / or memory capability may be distributed as well. The computing system may include multiple structures as “executable components”. The term “executable component” is a structure well understood in the field of computing as being a structure that can be software, hardware, or a combination thereof. For instance, when implemented in software, one of ordinary skill in the art would understand that the structure of an executable component may include software objects, routines, methods, and so forth, that may be executed on the computing system. This may include both an executable component in the heap of a computing system, or on computer- readable storage media. The structure of the executable component may exist on a computer-readable medium such that, when interpreted by one or more processors of a computing system, e.g., by a processor thread, the computing system is caused to perform a function. Such structure may be computer readable directly by the processors, for instance, as is the case if the executable component were binary, or it may be structured to be interpretable and / or compiled, for instance, whether in a single stage or in multiple stages, so as to generate such binary that is directly interpretable by the processors. In other instances, structures may be hard coded or hard wired logic gates, that are implemented exclusively or near-exclusively in hardware, such as within a field programmable gate array (FPGA), an application specific integrated circuit (ASIC), or any other specialized circuit. Accordingly, the term “executable component” is a term for a structure that is well understood by those of ordinary skill in the art of computing, whether implemented in software, hardware, or a combination. Any embodiments herein are described with reference to acts that are performed by one or more processing units of the computing system. If such acts are implemented in software, one or more processors direct the operation of the computing system in response to having executed computer-executable instructions that constitute an executable component. Computing system may also contain communication channels that allow the computing system to communicate with other computing systems over, for example, network. A “network” is defined as one or more data links that enable the transport of electronic data between computing systems and / or modules and / or other electronic de- vices. When information is transferred or provided over a network or another communications connection, for example, either hardwired, wireless, or a combination of hardwired or wireless, to a computing system, the computing system properly views the connection as a transmission medium. Transmission media can include a network and / or data links which can be used to carry desired program code means in the form of computer-executable instructions or data structures and which can be accessed by a general-purpose or specialpurpose computing system or combinations. While not all computing systems require a user interface, in some embodiments, the computing system includes a user interface system for use in interfacing with a user. User interfaces act as input or output mechanism to users for instance via displays.

[0065] Those skilled in the art will appreciate that at least parts of the invention may be practiced in network computing environments with many types of computing system configurations, including, personal computers, desktop computers, laptop computers, message processors, hand-held devices, multi-processor systems, microprocessor-based or programmable consumer electronics, network PCs, minicomputers, main-frame computers, mobile telephones, PDAs, pagers, routers, switches, datacenters, wearables, such as glasses, and the like. The invention may also be practiced in distributed system environments where local and remote computing system, which are linked, for example, either by hardwired data links, wireless data links, or by a combination of hardwired and wireless data links, through a network, both perform tasks. In a distributed system environment, program modules may be located in both local and remote memory storage devices.

[0066] Those skilled in the art will also appreciate that at least parts of the invention may be practiced in a cloud computing environment. Cloud computing environments may be distributed, although this is not required. When distributed, cloud computing environments may be distributed internationally within an organization and / or have components possessed across multiple organizations. In this description and the following claims, “cloud computing” is defined as a model for enabling on-demand network access to a shared pool of configurable computing resources, e.g., networks, servers, storage, applications, and services. The definition of “cloud computing” is not limited to any of the other numerous advantages that can be obtained from such a model when deployed. The computing systems of the figures include various components or functional blocks that may implement the various embodiments disclosed herein as explained. The various components or functional blocks may be implemented on a local computing system or may be implemented on a distributed computing system that includes elements resident in the cloud or that implement aspects of cloud computing. The various components or functional blocks may be implemented as software, hardware, or a combination of software and hardware. The computing systems shown in the figures may include more or less than the components illustrated in the figures and some of the components may be combined as circumstances warrant.

[0067] Any reference signs in the claims should not be construed as limiting the scope.

[0068] The invention refers to an apparatus for controlling a foam recycling process. A providing unit provides measurement data associated with the foam product. The measurement data is indicative of a spectrum received from foam product. A SAN model providing unit provides a machine learning based SAN model for determining an SAN content in the foam product. The SAN model has been trained utilizing historical training data to determine the SAN content in a foam product based on the measurement data. A determination unit de- termines the SAN content in the foam product based on the SAN model and the measurement data. A control data generation unit generates control data for controlling the polyurethane foam product recycling process based on the determined SAN content of the foam product.

Claims

Claims:1 . Apparatus for generating control data for controlling at least a part of a polyurethane foam recycling process for recycling polyurethane foam products, wherein the polyurethane foam product is provided to a recycling process, wherein the provided polyurethane foam product optionally comprises styrene-acrylonitrile (SAN) particles and optionally comprises different polyurethane foam types, wherein the apparatus comprises: a measurement data providing unit for providing measurement data associated with at least a part of the polyurethane foam product, wherein the measurement data is indicative of a spectrum received from the part of the polyurethane foam product when the part of the polyurethane foam product is subjected to light in a predetermined wavelength range, an SAN model providing unit for providing a machine learning based SAN model for determining an SAN content in the part of the polyurethane foam product, wherein the SAN model has been trained utilizing historical training data comprising historical measurement data of different polyurethane foam products and a respective SAN content of the polyurethane foam products to determine the SAN content in a polyurethane foam product based on the measurement data, an SAN content determination unit for determining the SAN content in the at least the part of the polyurethane foam product based on the SAN model and the measurement data, a control data generation unit for generating control data for controlling at least a part of the polyurethane foam product recycling process based on the determined SAN content of the part of the polyurethane foam product.

2. The apparatus according to claim 1 , wherein the control data is configured for controlling a sorting process being part of the polyurethane foam product recycling process such that different polyurethane foam products and / or parts of a polyurethane foam product are sorted based on the SAN content of the respective polyurethane foam product and / or parts of the polyurethane foam product.

3. The apparatus according to any of claims 1 and 2, wherein the control data is configured for controlling process parameters of a chemical lysis reaction being part of the recycling process based on the SAN content.

4. The apparatus according to any of the preceding claims, wherein the SAN model comprises a vector support machine.

5. The apparatus according to any of the preceding claims, wherein the part of the polyurethane foam product is subjected to near infrared light with wavelengths between 950 and 1700 nm.

6. The apparatus according to any of the preceding claims, wherein the apparatus further comprises: a classification model providing unit for providing a foam classification model for classifying the at least a part of the polyurethane foam product into a class of one or more foam classes, wherein the foam classes depend on the polyurethane foam types and / or further additives in a polyurethane foam product, and wherein the classification model is a machine learning based model trained utilizing historical data to classify a polyurethane foam product into a class of the one or more classes based on the measurement data, a classification unit for classifying the at least a part of the polyurethane foam product based on the classification model and the measurement data, wherein the control data generation unit is configured for generating control data for controlling at least a part of the polyurethane foam product recycling process further based on the classification of the part of the polyurethane foam product.

7. The apparatus according to claim 6, wherein the one or more classes comprise at least one of the following classes: HR-MDI, HR-TDI, HR-TDI CM, HyperSoft-MDI, Hyper- Soft-TDI, Standard, Standard-CME, Standard-HR-mix-TDI, VE-MDI, more preferably, TDI- Standard and no-Standard.

8. The apparatus according to any of claims 6 and 7, wherein the classification model comprises a support vector machine.

9. A system for generating control data for controlling at least a part of a polyurethane foam product recycling process, wherein the system comprises: a measurement device configured for illuminating at least a part of a polyurethane foam product with light in a predetermined wavelength range and measuring a spectrumreceived from the part of the polyurethane foam product illuminated with the infrared light for generating and providing measurement data indicative of the measurement, and an apparatus according to any of the preceding claims.

10. An apparatus for training a machine learning SAN model utilizable in the apparatus according to any of claims 1 to 8, wherein the apparatus comprises: a historical data providing unit for providing historical data fortraining the SAN model, wherein the historical data comprises a) measurement data of a plurality of polyurethane foam products with different SAN contents, and b) a respective SAN content for each of the plurality of polyurethane foam products, wherein the measurement data is indicative of a spectrum received from at least a part of a respective polyurethane foam product when the part of the respective polyurethane foam product is subjected to light in a predetermined wavelength range, a model providing unit for providing a trainable machine learning based SAN model, a training unit for training the trainable SAN model based on the historical data such that the trained SAN model is configured to determine the SAN content of a polyurethane foam product based on the respective measurement data of the polyurethane foam product, and a trained model providing unit for providing the trained SAN model.

11. A computer-implemented method for generating control data for controlling at least a part of a polyurethane foam recycling process for recycling polyurethane foam products, wherein the polyurethane foam product is provided to a recycling process, wherein the provided polyurethane foam product optionally comprises styrene-acrylonitrile (SAN) particles and optionally comprises different polyurethane foam types, wherein the method comprises: providing measurement data associated with at least a part of the polyurethane foam product, wherein the measurement data is indicative of a spectrum received from the part of the polyurethane foam product when the part of the polyurethane foam product is subjected to light in a predetermined wavelength range,providing a machine learning based SAN model for determining an SAN content in the part of the polyurethane foam product, wherein the SAN model has been trained utilizing historical training data comprising measurement data of different polyurethane foam products and a respective SAN content of the polyurethane foam products to determine the SAN content in a polyurethane foam product based on the measurement data, determining the SAN content in the at least a part of the polyurethane foam product based on the SAN model and the measurement data, generating control data for controlling at least a part of the polyurethane foam product recycling process based on the determined SAN content of the part of the polyurethane foam product.

12. A computer-implemented method for training a machine learning SAN model utilizable in the apparatus according to any of claims 1 to 8, wherein the method comprises: providing historical data for training the SAN model, wherein the historical data comprises a) measurement data of a plurality of polyurethane foam products with different SAN contents, and b) a respective SAN content for each of the plurality of polyurethane foam products, wherein the measurement data is indicative of a spectrum received from at least a part of a respective polyurethane foam product when the part of the respective polyurethane foam is subjected to light in a predetermined wavelength range, providing a trainable machine learning based SAN model, training the trainable SAN model based on the historical data such that the trained SAN model is configured to determine the SAN content of a polyurethane foam product based on the respective measurement data of the polyurethane foam product, and providing the trained SAN model.

13. A polyurethane recycling process for recycling a polyurethane foam product, wherein the polyurethane recycling process is controlled utilizing control data generated by the apparatus according to any of claims 1 to 8 and / or the method according to claim 11 .

14. A computer program product for generating control data for controlling at least a part of a polyurethane foam product recycling process for recycling a polyurethane foam product, wherein the computer program product comprises program code means for causing an apparatus according to any of claims 1 to 9 to carry out the method according to claim 11.

15. A computer program product for generating control data fortraining a machine learning SAN model utilizable in the apparatus according to any of claims 1 to 9, wherein the computer program product comprises program code means for causing an apparatus according to claim 10 to carry out the method according to claim 12.

Images