Method and system for computer-aided evaluation of dynamic differential calorimetry measurement data
By combining self-learning algorithms and DSC, plastic samples are automatically identified and allocated, solving the problem of time-consuming and costly sorting in existing technologies and achieving faster and more economical sorting of plastic recyclables.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- NETZSCH GERATEBAU GMBH
- Filing Date
- 2025-11-06
- Publication Date
- 2026-06-30
AI Technical Summary
Existing technologies are time-consuming and costly in plastic waste sorting, making it difficult to quickly and accurately sort materials based on their properties, thus affecting the efficiency and quality of plastic recycling.
By employing a self-learning algorithm combined with differential scanning calorimetry (DSC) and a machine learning system, the material categories of different plastic samples are automatically identified and assigned through analysis of DSC measurement data, thus optimizing the sorting process.
It enables faster, more economical and accurate analysis and identification of plastic samples, improves the sorting efficiency and quality of plastic recyclables, and reduces sorting costs.
Smart Images
Figure CN122306871A_ABST
Abstract
Description
Invention Field
[0001] This invention relates to a method and system for computer-aided evaluation of dynamic differential calorimetry data, particularly for the analysis and identification of different plastic samples, for example, in the context of recycling sorting. Background Technology
[0002] Recycling is particularly important in the field of plastic waste because its near-complete non-biodegradability and subsequent accumulation in the environment are undesirable. For example, plastic reuse can be achieved through the melting and molding of plastic waste. To make this type of recycling effective, plastic waste is typically sorted according to its material properties (e.g., color and / or polymer type) before reprocessing. This sorting requires examining the composition of the plastic waste, and its implementation is often time-consuming and therefore costly due to the unpredictability of the introduced material flow.
[0003] When characterizing the high value of polymer recyclables, the degree of quality determination and assurance depends on the quality of the separation and identification methods used to process the raw materials. In particular, for effective reuse, the physical properties of polymer recyclables must be specified and guaranteed within tight tolerances. Modern computer-aided data processing of measurement data from efficient measuring equipment can accelerate and enhance the reliability of polymer recyclable quality analysis. Examples of systems and methods for classifying and sorting plastic materials using image processing systems and one or more sensor systems are illustrated in document US2022 / 0161298 A1.
[0004] The measurement principles of thermal analysis—particularly the physical measurement principles of differential scanning calorimetry (DSC)—can help experimentally characterize plastic waste of varying masses. DSC measuring equipment measures the heat capacity of a sample by recording the heat flow rate entering the sample and comparing it to a reference sample. The resulting graph of heat flow relative to sample temperature or its progression over time can determine material transition points, such as glass transition temperature or melting temperature, the crystallinity of thermoplastic matrices, or the curing behavior or residual heat of reaction in thermosetting materials.
[0005] Document WO 2022 / 170273 A1 discloses a system and method for classifying and sorting plastic materials of different colors using an image processing system or one or more sensor systems, wherein the acquired image data is processed in a machine learning system for sorting purposes to identify or classify each material. Document EP 4 209 781 A1 discloses a computer-implemented method for thermal analysis of material samples.
[0006] The article “Automated Differential Scanning Calorimetry Signal Analysis Based on Machine Learning” by Xin Lv, Shuyu Wang, Peng Shan, Yuliang Zhao, and Lei Zuo—in Measurement, Vol. 187, No. 110218, 2022—discloses a computer-aided evaluation method for DSC measurement data based on a semi-automated machine learning model. The article “Accurate Determination of Polyethylene (PE) and Polypropylene (PP) Content in Polyolefin Blends Using Machine Learning-Assisted Differential Scanning Calorimetry (DSC) Analysis” by Amir Bashirgonbadi, Yannick Ureel, Laurens Delva, Rudinei Fiorio, Kevin M. van Geem, and Kirn Ragaert—in Polymer Testing, Vol. 131, No. 108353, February 2024—discloses the application of artificial intelligence in evaluating DSC measurement curves for polymer recycling characterization. Summary of the Invention
[0007] One object of the present invention is to provide the possibility of analyzing DSC measurement data more economically, simply, and quickly. In particular, one object of the present invention is to accelerate and improve the analytical identification between different plastic samples, for example, in the classification and quantification of recyclables.
[0008] According to the invention, this objective is achieved in each case through the subject matter of the independent claims.
[0009] Advantageous implementations and improvements are derived from the dependent claims that reference the independent claim and the description with reference to the accompanying drawings.
[0010] The above-described embodiments and improvements can be combined arbitrarily with each other, as long as they are meaningful. Other possible embodiments, improvements, and implementations of the present invention include combinations of features of the invention not explicitly mentioned in the embodiments described above or below. In particular, those skilled in the art will add individual aspects as improvements or supplements to the corresponding basic forms of the present invention. Attached Figure Description
[0011] The invention will now be explained in more detail with reference to the accompanying drawings and embodiments. The drawings show: Figure 1 This is a flowchart of a computer-aided evaluation method for dynamic differential calorimetry measurement data according to an embodiment of the present invention; Figure 2 It is a system for computer-aided evaluation of dynamic differential calorimetry data—specifically for achieving evaluation based on… Figure 1The method of the embodiment—a schematic diagram of the embodiment; and Figure 3 This is a schematic diagram of a recyclable sorting apparatus according to an embodiment of the present invention, which has a system for computer-aided evaluation of dynamic differential calorimetry data.
[0012] In the accompanying drawings, unless otherwise specified, the same, functional, and operational elements, features, and components have the same reference numerals in every case. Detailed Implementation
[0013] Although specific embodiments and improvements are shown and described herein, those skilled in the art will preferably substitute several alternative and / or similar embodiments for the specific embodiments shown and described without departing from the scope of the invention. This application is intended to cover all modifications or variations of the specific embodiments described herein.
[0014] The accompanying drawings are intended to provide a further understanding of embodiments of the invention and, in conjunction with the description, to explain the principles and concepts of the invention. Other embodiments and many of the mentioned advantages will be apparent from the drawings. The drawings are to be understood as schematic only, and the elements in the drawings are not necessarily shown to scale relative to each other. Terms indicating direction, such as “up,” “down,” “left,” “right,” “above,” “below,” “horizontal,” “vertical,” “front,” “back,” etc., are used for illustrative purposes only and are not intended to limit the generality to the specific design shown in the figures.
[0015] The dashed lines in the attached diagram illustrate that the connections between components linked by the dashed lines do not necessarily have to be physically in contact with each other, but can also be wirelessly coupled to each other.
[0016] The following description references self-learning algorithms used in artificial intelligence (AI) systems. Generally, self-learning algorithms simulate cognitive functions that are correlated with human judgment and thinking abilities. In this context, by introducing new training information, the self-learning algorithm can dynamically adjust the knowledge gained so far from old training information according to changing circumstances in order to identify and infer patterns and regularities throughout the training information.
[0017] In the self-learning algorithm of this invention, all types of training that generate human knowledge gains can be used, such as supervised learning, partially supervised learning, autonomous learning (based on generative, non-generative, or deep adversarial networks (“Adversarial Networks”, ANs)), reinforcement learning, or active learning. Feature-based learning (“representation learning”) can be used in any of these cases. Specifically, the self-learning algorithm of this invention can iteratively adjust the parameters and features to be learned via feedback analysis.
[0018] The self-learning algorithm in the sense of this invention can be based on a regressor, a support vector network (SVN), a neural network (e.g., a convolutional neural network (CNN), Kohonen network, recurrent neural network, time-delayed neural network (TDNN), or oscillatory neural network (ONN)), a random forest classifier, a decision tree classifier, a Monte Carlo network, or a Bayesian classifier. In this case, the self-learning algorithm in the sense of this invention can use an attribute genetic algorithm, a k-means algorithm (e.g., the Lloyd algorithm or the MacQueen algorithm), or a TD learning algorithm (e.g., SARSA learning or Q learning).
[0019] Differential calorimetry data in the context of this invention can specifically include all datasets generated by a differential calorimetry device or differential calorimeter sensor. A differential calorimeter is a device used to measure the heat flow through various substances, particularly plastic samples. This device operates based on differential scanning calorimetry (DSC) technology, which aims to measure the heat released during enthalpy change between a sample and a reference material.
[0020] Here, a sample (e.g., a piece of plastic) is placed in a special chamber adjacent to an empty reference chamber. These two chambers are isolated and uniformly heated, for example, via a heating pad beneath them, which provides a continuous and predefined supply of heat. Due to the sample's heat capacity, endothermic or exothermic processes occur at the melting or sublimation point, as well as phase transitions, providing information about heat flux changes with temperature. Temperature changes during controlled heating and cooling in both chambers are measured at a specific time resolution by thermal sensors mounted in each chamber. Differential calorimetry provides information about characteristic values used to characterize the thermal properties of the plastic sample, such as glass transition temperature, melting point, enthalpy of reaction, crystallinity, and specific heat capacity.
[0021] Figure 1 A flowchart of an exemplary method M for computer-aided evaluation of dynamic differential calorimetry data is shown. This method M can be specifically used with a system for computer-aided evaluation of dynamic differential calorimetry data (e.g., in...). Figure 2 The system 30 shown in the exemplary diagram is used to achieve this. Method M can be used, for example, for analytical identification between different plastic samples—e.g., during recycling sorting in the recycling sorting device 100, such as... Figure 3 As shown in the example.
[0022] In the first step M1 of method M, the DSC measurement device 40 generates a DSC measurement curve for the unknown sample. In the second step M2, the AI system 10 determines the data patterns in the DSC measurement curve using a self-learning algorithm. Finally, based on the data patterns determined in step M3, the AI system 10 assigns the material categories contained in the DSC measurement curve to specific material categories or specific materials.
[0023] Figure 2 An exemplary illustration is shown of a system 30 for computer-aided evaluation of dynamic differential calorimetry (DSC) measurement data. This system 30 can be specifically used to achieve… Figure 1 Method M in the text.
[0024] System 30 includes AI system 10 and control system 20 for differential calorimetry (DSC measurement device) 40. Control system 20 has control processor 9, which is coupled to AI system 10 via input interface 7 and output interface 8, and to DSC measurement device 40 on the other hand.
[0025] The control system 20 is used to control the DSC measuring device 40 because the control processor 9 predetermines the control program for the DSC measuring device 40. For example, the control system 20 can preset protocol parameters for DSC measurement, such as the heating or cooling rate of the sample, the number of heating cycles to be performed, and the waiting time between the heating and cooling phases.
[0026] DSC measurement data can be input from the control processor 9 to the AI system 10 via the input interface 7. This DSC measurement data can be, for example, DSC measurement curves obtained from known or unknown samples, or data transmitted to the control processor 9 from other connected systems as computer-generated training data D.
[0027] AI system 10 includes a data analysis processor 1 and a machine learning system 5 (ML system). The ML system 5 also includes an AI processor 2, a rule set generator 3 based on a self-learning algorithm, and a reference rule set memory 4. AI system 10 communicates bidirectionally with control processor 9 via the data analysis processor 1. Control processor 9 can initially provide DSC measurement curves to AI system 10 as basic training data. This basic training data can be used as the basis for detecting patterns and regularities in the curve profile of the DSC measurement curves by rule set generator 3. Rule set generator 3 can include, for example, a regressor, support vector classifier, neural network, random forest classifier, decision tree classifier, Monte Carlo network, or Bayesian classifier.
[0028] Patterns and regularities detected in the curve profile of the DSC measurement curve are first iteratively stored in a dynamically and continuously updated training rule set. An operational reference rule set is formed from the training rule set, and the rule set generator 3 stores this operational reference rule set in the reference rule set memory 4. When the AI processor 2 receives a request Q from the data analysis processor 1 to determine data patterns in the DSC measurement curve of an unknown sample, the AI processor 2 uses the reference rule set stored in the reference rule set memory 4 as a reference. Compared to this reference, the AI processor 2 checks which allocation of a specific material category or a specific material best matches the material category included in the obtained DSC measurement curve. The rule set generator 3 can periodically update the reference rule set stored in the reference rule set memory 4 based on newly added DSC measurement data or based on new external presets.
[0029] The data pattern determination results are returned from AI processor 2 to data analysis processor 1. Then, data analysis processor 1 can output the analysis results to control processor 9 via output interface 8. The analysis results indicate which material category or material should be assigned to the unknown sample corresponding to the DSC measurement curve received by control processor 9.
[0030] AI system 10 may also have a control program database 6 coupled to data analysis processor 1. If data analysis processor 1 receives a DSC measurement curve from control processor 9, and this DSC measurement curve can only be assigned to a specific material category or a specific material by AI processor 2 with insufficient confidence or with only a confidence level below an adjustable confidence threshold, then data analysis processor 1 may suggest to control processor 9 a change to the control program that can be pre-given by control system 20. For example, if the possibility of classifying a material category or material is insufficient based on the DSC measurement curve as a query Q, AI processor 2 may provide indications regarding which parts of the DSC measurement curve make classification difficult. Based on the parts of the DSC measurement curve indicated by AI processor 2, data analysis processor 1 may select a control program from control program database 6 whose protocol parameters for DSC measurement are changed relative to the control program previously pre-given by control system 20, such that the problematic parts of the DSC measurement curve are substantially changed in the new DSC measurement.
[0031] Components of AI system 10 can be installed together with control system 20 in a local data processing facility. However, individual components or system parts can also be installed outside the local data processing facility. For example, the AI system can be operated in a cloud environment, allowing input interface 7 and output interface 8 to be accessed via a remote access network (e.g., the Internet).
[0032] In addition, the control processor 9 may have an input / output interface through which the user of system 30 can perform input and output 10.
[0033] Figure 3 An exemplary implementation of a system 30 for computer-aided evaluation of dynamic differential calorimetry (DC) measurement data in a recycling sorting apparatus 100 is shown. The recycling sorting apparatus 100 includes an automated sorting system 50, for example, for plastics. This automated sorting system 50 is coupled to a DSC measuring device 40, which is designed to record DSC measurement curves of unknown plastic samples processed in the automated sorting system 50. The recorded DSC measurement curves are transmitted to a control system 20, which in turn transmits them to an AI system 10 for automatically assigning specific materials or specific material categories to the material categories included in the recorded DSC measurement curves. The control system 20 and the AI system 10 can be combined, for example... Figure 2 Operate as explained.
[0034] The control system 20 is also used to control the DSC measuring device 40, and in particular to specify the protocol parameters for the DSC measurement of the DSC measuring device 40, such as the heating or cooling rate of the sample, the number of heating cycles to be performed, the waiting time between the heating and cooling phases, etc.
[0035] In the foregoing detailed description, various features have been summarized in one or more examples to enhance the rigor of the representation. However, it should be clear that the above description is merely illustrative and not restrictive. It is intended to cover all alternatives, modifications, and equivalents of the various features and embodiments. Given the foregoing description, many other examples will be readily apparent to those skilled in the art based on their expertise.
[0036] The embodiments were chosen and described in order to best represent the principles on which the invention is based and its potential applications in practice. Therefore, those skilled in the art can best modify and use the invention and its various embodiments with respect to their intended use. In the claims and description, the terms “comprising” and “having” are used as neutral linguistic terms corresponding to the term “including”. Furthermore, the use of the terms “a” and “an” should not, in principle, exclude multiple features and components described in this manner. Figure Labels
[0037] 1 Data Analysis Processor
[0038] 2 AI processors
[0039] 3. Rule set generator
[0040] 4. Reference rule set storage
[0041] 5 Machine Learning Systems
[0042] 6. Control Program Database
[0043] 7 Input Interface
[0044] 8 Output Interfaces
[0045] 9. Control Processor
[0046] 10 AI systems
[0047] 20 Control System
[0048] 30 System
[0049] 40 DSC measuring device
[0050] 50 Automatic Sorting System
[0051] 100 Recyclable Material Sorting Device
[0052] Q Request
[0053] 10 Inputs / Outputs
[0054] M. A computer-aided evaluation method for dynamic differential calorimetry data.
[0055] M1-M5 Method Steps
Claims
1. A system (30) for computer-aided evaluation of dynamic differential calorimetry (DSC) measurement data, said system (30) comprising: A control system (20) for the DSC measuring device (40) and having a control processor (9) designed to receive DSC measurement curves of unknown samples from the DSC measuring device (40); and AI system (10) having a data analysis processor (1) and a machine learning system (5), wherein the AI system (10) communicates bidirectionally with the control processor (9) via the data analysis processor (1) and is designed to identify data patterns in the DSC measurement curve of the unknown sample by using a self-learning algorithm, and assign the material category contained in the DSC measurement curve to a specific material category or a specific material based on the determined data pattern.
2. The system (30) according to claim 1, wherein, The machine learning system (5) has an AI processor (2), a rule set generator based on a self-learning algorithm (3), and a reference rule set memory (4).
3. The system (30) according to claim 2, wherein, The rule set generator (3) has a regressor, a support vector classifier, a neural network, a random forest classifier, a decision tree classifier, a Monte Carlo network, or a Bayesian classifier.
4. The system (30) according to any one of claims 1 to 3, wherein, The control processor (9) is also designed to pre-program the DSC measurement device (40) with a control program, wherein the DSC measurement device (40) can pre-program the values of the protocol parameters for the DSC measurement.
5. The system (30) according to claim 4, wherein, The AI system (10) has a control program database (6) coupled to the data analysis processor (1) and stores various control programs for the DSC measurement device (40).
6. The system (30) according to claim 5, wherein, The data analysis processor (1) is designed such that if the data analysis processor (1) receives a DSC measurement curve from the control processor (9), and the DSC measurement curve can only be assigned to a specific material category or a specific material by the AI processor (2) with insufficient confidence or with confidence below an adjustable confidence threshold, the data analysis processor (1) selects a control program from the control program database (6) whose protocol parameters regarding DSC measurement have changed relative to a control program previously given by the control system (20).
7. The system (30) according to any one of claims 1 to 6, wherein, The AI system (10) is implemented in a cloud environment.
8. A method (M) for computer-aided evaluation of dynamic differential calorimetry (DSC) measurement data, said method (M) comprising: One or more DSC measurement curves for an unknown sample (M1) are generated by the DSC measurement device (40); The AI system (10) determines the data pattern (M2) in the DSC measurement curve by using a self-learning algorithm; and Based on the determined data pattern, the AI system (10) assigns (M3) the material category contained in the DSC measurement curve to a specific material category or a specific material.
9. The method (M) according to claim 8, wherein, The AI system (10) has a machine learning system (5), which has an AI processor (2), a rule set generator based on a self-learning algorithm (3), and a reference rule set memory (4).
10. The method (M) according to claim 9, wherein, The rule set generator (3) has a regressor, a support vector classifier, a neural network, a random forest classifier, a decision tree classifier, a Monte Carlo network, or a Bayesian classifier.
11. The method (M) according to any one of claims 8 to 10, wherein, The DSC measuring device (40) generates the DSC measurement curve according to the control program, wherein the values of the protocol parameters for DSC measurement are given in advance.
12. The method (M) according to claim 11, wherein, The AI system (10) has a control program database (6) coupled to the data analysis processor (1) and stores various control programs for the DSC measurement device (40).
13. The method (M) according to claim 12, wherein, If the AI system (2) can only assign a particular material category or a particular material with insufficient confidence or with confidence below the adjustable confidence threshold (M3), then a control program is selected from the control program database (6) with different protocol parameters for DSC measurement relative to a control program previously given by the control system (20).
14. The system (30) according to any one of claims 1 to 7 is used for analyzing and identifying different plastic samples.
15. A recyclable sorting device (100) comprising a differential calorimeter, a DSC measuring device (40), and a system (30) for computer-aided evaluation of dynamic DSC measurement data according to any one of claims 1 to 7.