EP4332435A1 - Heating value determination - Google Patents

Abstract

Computer-implemented method for determining a predicted calorific value of a process medium in a thermal recycling plant, comprising providing the process medium by a supply unit, the process medium comprising a high-volume material flow, detecting at least one parameter of the process medium by means of a first sensor unit, and determining a predicted calorific value based on the recorded parameters of the provided process medium in a processing unit using a forecasting unit. Detecting the at least one parameter of the process medium includes detecting at least one of a grain size of the process medium, a particle quality of the process medium, a dust concentration of the process medium, a humidity of the process medium, a particle size distribution of the process medium, a critical object in the process medium or a biogenic fraction of the process medium.

Description

TECHNICAL FIELD OF THE INVENTION

This application claims priority over the German patent application DE 10 2022 122 433.3 filed on September 5, 2022 . The entire disclosure of the German patent application DE 10 2022 122 433.3 is hereby incorporated herein by reference.

The invention relates to a device and a computer-implemented method for determining the calorific value of high-volume material flows.

BACKGROUND OF THE INVENTION

In industrial waste disposal and recycling, different processes are used to recover secondary raw materials or to recover energy from waste. In many countries, for example, incineration processes are used to thermally utilize waste that is difficult or non-recyclable through incineration. In thermal recycling, the waste is continuously fed into a combustion chamber as a volume flow and burned with the addition of, for example, atmospheric oxygen. This thermal utilization generates thermal energy, which can be used, for example, to generate electricity and/or to heat buildings via a district heating network.

However, due to the different types of waste that are fed into the incineration process as a material stream, there may be fluctuations in this incineration process. For example, the material flow can contain objects or contaminants that influence the combustion process in an unpredictable way. Flame-retardant objects or substances can, for example, mean that more atmospheric oxygen has to be supplied or even fire-accelerating additives have to be added to the combustion chamber in order to ensure a continuous and low-emission combustion process. In contrast, you can For example, highly flammable or even explosive objects or substances lead to an unexpectedly strong combustion and additional waste has to be added as fuel. The combustion chamber can also be damaged during explosive combustion. These changes in the volume flow therefore lead to an unforeseen change in the combustion process and thereby reduce the efficiency of this combustion process. Furthermore, damage to the system and associated downtime of the thermal recycling plant can occur.

Different solutions for the targeted checking and sorting of material flows are known in the prior art. For example, the patent application describes WO 2020/082 176 A1 (Camirand et. al., Waste Robotics Inc.) a robotic impaler and a robotic method for performing interception and sorting of a selected item from a mass of items in a waste stream. The device comprises a robot manipulator, at one end of which a skewer end effector body is mounted, which can be extended and retracted. The device also includes an image processing system that is set up to capture images of the mass of objects, process the captured images to identify the selected object from the mass of objects and determine the position of the selected object. The identified object is removed from the material flow by means of the impaling end effector body.

The US patent also describes, for example US 10,799,915 B2 a system and method for sorting recyclable items and other materials. The system includes at least one image sensor, a controller and at least one slider device connected to the controller. The controller includes a processor and a memory and is set up to receive image data captured by the image sensor. The at least one slider device is set up to receive an actuation signal from the controller. The processor is set up to detect objects moving with the material flow and to detect at least one target object moving with the material flow by processing the image data and to determine an expected time at which the at least one target object will be within a deflection path of the slide device. The Controller generates the actuation signal selectively depending on whether an object detected in the image data, which includes at least one target object, and thereby causes the detected object to be removed from the material flow.

Furthermore, for example, the publication describes Pehlken, Alexandra, Patrick Eschemann, Henriette Garmatter, Fabian Cyris, and Astrid Nieße. 2021. "Use of artificial intelligence in the digitalization of waste incineration power plants". Society for computer science , Bonn a system and a process for the thermal utilization of waste in waste incineration power plants. The system records measurement and image data in a thermal recycling plant, which is processed and fed to pre-trained neural networks in order to classify different materials and thereby obtain information about waste composition and calorific value. From Pelken et al. However, it becomes clear that a forecast of the calorific value of waste is complex due to the heterogeneity of the waste in waste incineration power plants. Furthermore, it is clear from Pelken et.al that there are no solutions for analyzing high-volume material flows.

Various solutions are known in the prior art which enable targeted recognition of individual objects from a number of objects during recycling and thermal utilization of waste. However, solutions for predicting a composition or predicting a calorific value of an inhomogeneous and high-volume material flow for controlling a calorific value in a thermal recycling plant are not known.

SUMMARY OF THE INVENTION

Against this background, a device and a computer-implemented method for determining a predicted calorific value of a process medium as well as a device and a computer-implemented method for calorific value control in a thermal recycling plant using a predicted calorific value are provided.

The computer-implemented method for determining a predicted calorific value of a process medium in a thermal recycling plant includes providing the process medium by a supply unit, detecting at least one parameter of the process medium by means of a first sensor unit and determining a predicted calorific value based on the detected parameters of the provided process medium in a processing unit by means of a forecasting unit. The process medium includes, for example, a high-volume material stream. Detecting the at least one parameter of the process medium includes detecting at least one of a grain size of the process medium, a particle quality of the process medium, a dust concentration of the process medium, a humidity of the process medium, a particle size distribution of the process medium, a critical object in the process medium or a biogenic fraction of the process medium.

By determining the predicted calorific value, it is possible to qualitatively and quantitatively evaluate the high-volume material flow. This can increase process efficiency in downstream thermal recycling processes. Furthermore, this can prevent system malfunctions and failures in downstream thermal recycling processes. This transparency makes material flow changes and characteristics measurable. This measurability makes proactive, systematic actions possible to increase process efficiency. For example, the material flow that is fed into thermal recycling can be proactively controlled so that waste is burned in an optimal calorific value range that corresponds to the thermal recycling plant. This proactive control prevents, for example, thermal overstressing of system components due to excessive heating values in the material flow. In addition, increased emissions can be avoided due to suboptimal temperature development during thermal recycling. The forecast of the calorific value of the high-volume material flow makes it possible to contribute to reducing maintenance costs, damage in the system process, and more optimal thermal utilization of the material flow. This promotes a higher material flow throughput, which is in the commercial interest of a plant operator.

In a further aspect, a computer-implemented method for calorific value control in a thermal utilization plant using a predicted calorific value includes providing a process medium by a supply unit, detecting at least one parameter of the process medium by means of a first sensor unit, determining the predicted calorific value based on the detected parameters of the provided process medium in a processing unit by means of a forecasting unit and controlling a composition of the provided process medium by means of a composition control unit using the predicted calorific value of the provided process medium. The process medium includes, for example, a high-volume material stream. Detecting the at least one parameter of the process medium includes detecting at least one of a grain size of the process medium, a particle quality of the process medium, a dust concentration of the process medium, a moisture content of the process medium, a particle size distribution of the process medium, a critical object in the process medium or a biogenic fraction of the process medium.

By controlling the composition of the high-volume material stream, it is possible to increase process efficiency in thermal recycling processes.

In a further aspect, the computer-implemented method further comprises thermally utilizing the provided process medium in a combustion chamber to form a thermally utilized process medium.

By controlling the composition of the high-volume material stream, it is possible to increase process efficiency in thermal recycling processes.

In a further aspect, the computer-implemented method further comprises detecting an actual calorific value of the thermally utilized process medium by means of a second sensor unit.

By detecting the actual calorific value of the thermally utilized process medium using the second sensor unit, it is possible to improve the determination of the predicted calorific value.

In a further aspect, the computer-implemented method further comprises training a machine learning algorithm taking into account the actual calorific value of the thermally utilized process medium and the calorific value predicted based on the recorded parameters of the process medium, wherein the forecasting unit comprises the machine learning algorithm.

By training the machine learning algorithm, the determination of the predicted heating value is improved. This allows the effects of fluctuations in the composition of the process medium to be reduced.

In a further aspect, the training of the machine learning algorithm in the computer-implemented method takes place, taking into account the actual calorific value of the thermally utilized process medium and the predicted calorific value of the process medium, a continuous and iterative comparison of the actual calorific value and the predicted calorific value based on the recorded parameters of the process medium by the processing unit , as well as the iterative optimization of the model for forecasting the calorific value.

By training the machine learning algorithm, the determination of the predicted heating value can be improved. This allows the effects of fluctuations in the composition of the process medium to be reduced.

In a further aspect, the computer-implemented method further comprises detecting the at least one parameter of the process medium, comprising detecting at least one parameter of the high-volume material flow, wherein the high-volume material flow is a continuous and essentially heterogeneous material volume flow from household and/or industrial waste, and wherein the high-volume material flow is a mass flow with a volume rate of at least 1 t/min.

By recording a number of different parameters, different characteristics of the process medium are measured, which enable adequate conclusions to be drawn for a prediction of the calorific value. The characteristics can originate, for example, from color, thermal, depth (3D), or event, spectral, radar, grayscale images. By integrating the parameters in a uniform prediction model or by merging several prediction models, a continuous forecast of the calorific value is made possible.

In a further aspect, the at least one parameter of the process medium is detected using at least one of a color image camera, a mono camera, a high-speed camera, a thermal image camera, an infrared sensor, an X-ray device, a radar sensor, an event camera, and an induction sensor or a spectral sensor.

The sensor system used allows the multi-modal detection of the process medium. This multi-modal recording of the process medium makes it possible to measure various characteristics of the process medium and use them to predict the calorific value of the high-volume material flow.

In a further aspect, a system for determining a predicted calorific value of a process medium comprises a supply unit for providing the process medium, a first sensor unit set up to detect at least one parameter of the process medium, and a processing unit, electronically connected to the first sensor unit and set up to determine a predicted heating value based on the recorded parameters of the supplied process medium. The process medium includes, for example, a high-volume material stream.

The first sensor unit is connected to the processing unit. This makes it possible to specifically monitor and control the first sensor unit. In addition, the processing unit can change the data acquisition of the first sensor unit Adapt circumstances to optimize data collection. For example, data collection is actively controlled to prevent the collection of redundant sensor data.

In a further aspect, a system for a calorific value control in a thermal utilization plant using a predicted calorific value comprises a feed unit for providing the process medium, a first sensor unit set up to detect at least one parameter of the process medium, a processing unit, electronically connected to the first sensor unit and set up to determine a predicted calorific value of the supplied process medium, a combustion chamber for thermal utilization of the provided process medium into a thermally utilized process medium, and a composition control unit set up to control a composition of the supplied process medium. The process medium includes, for example, a high-volume material stream.

The composition of the aspects of the process medium from different process chain parts allows the changes in the process medium through the process of thermal recycling to be quantified or measurable. This is essential for adequate modeling and forecasting of the calorific value.

In a further aspect, the system comprises a second sensor unit, set up to determine the actual calorific value of the thermally utilized process medium.

By using the second sensor unit, parameters relating to the actual calorific value of the thermally utilized process medium can be recorded. This can improve the determination of the predicted calorific value. Furthermore, this can ensure consistent perception and corresponding data acquisition via a processing process of the process medium.

In a further aspect, the system includes the first sensor unit at least one of a color image camera, a mono camera, a high-speed camera, a thermal image camera, a 3D camera, an infrared sensor, an X-ray machine, a radar sensor, an event camera, an induction sensor or a spectral sensor.

The sensor system used allows multi-modal perception and recording of the process medium. Through this multi-modal recording of the process medium, computer-implemented methods can be used to enable a prediction of the corresponding calorific value.

In a further aspect, in the system the high-volume material flow is a continuous and essentially heterogeneous material volume flow from household and/or industrial waste.

A system designed in this way makes it possible to reliably detect a heterogeneous material volume flow and to reliably predict or forecast the calorific value of this mass flow.

In a further aspect, the high-volume material flow in the system is a mass flow with a volume rate of at least 1 t/min.

A system designed in this way makes it possible to reliably detect mass flows with a volume rate of at least 1 t/min and to reliably predict or forecast the calorific value of these mass flows.

In a further aspect, the computer-implemented method and device are used in a thermal waste treatment or recycling plant.

By integrating a multimodal sensor system in combination with the computer-implemented methods, an analysis and classification of the high-volume material flow at the feed unit is made possible. The actual characteristics of the 'actual' material flow can be measured for the calorific value forecast instead of relying on assumptions, such as those from delivery documentation of the material flow, which often have inaccuracy.

DESCRIPTION OF THE FIGURES

• Fig. 1 is a schematic representation of a system for determining a predicted calorific value of a process medium.
• Fig. 2a shows an overview of the synchronization module for synchronizing the recorded parameters.
• Fig. 2b shows an overview of the linking module for linking the recorded parameters.
• Fig. 2c shows an overview of the activity module for detecting active regions in the process medium.
• Fig. 2d shows an overview of the segmentation module for detecting segments in the process medium.
• Fig. 2e shows an overview of the module for detecting the grain size.
• Fig. 2f shows an overview of the module for detecting particle properties.
• Fig. 2g shows an overview of the module for recording the dust concentration.
• Fig. 2h shows an overview of the module for detecting moisture.
• Fig. 2i shows an overview of the module for recording the particle size distribution.
• Fig. 2j shows an overview of the module for detecting critical objects.
• Fig. 2k shows an overview of the module for recording the biogenic portion.
• Fig. 2l shows an overview of the module for collecting metadata.
• Fig. 3 is a schematic representation of a system for a calorific value control in a thermal recovery plant using a predicted calorific value.
• Fig. 4 is a schematic representation of the system for a calorific value control in a thermal utilization plant using the predicted calorific value comprising a composition control unit of a second embodiment.
• Fig. 5 is a process flow diagram depicting a computer-implemented method for determining a predicted calorific value of a process medium.
• Fig. 6 is a process flow diagram depicting a computer-implemented method for calorific value control in a thermal recovery plant using a predicted calorific value.

DETAILED DESCRIPTION OF THE INVENTION

The invention will now be described based on the drawings. It is understood that the embodiments and aspects of the invention described herein are merely examples and do not limit the scope of the claims in any way. The invention is defined by the claims and their equivalents. It is anticipated that features of one aspect or embodiment of the invention may be combined with a feature of another aspect or aspects and/or embodiments of the invention.

Fig. 1 is a schematic representation of a system 1 for determining a predicted calorific value H _i,p of a process medium 10. The system 1 comprises a feed unit 15, a first sensor unit 30 and a processing unit 40. The process medium 10 comprises a high-volume material stream 11. This high-volume Material stream 11 includes, for example, a material stream from waste such as that generated in an industrial raw material recycling plant.

The high-volume material stream 11 includes, for example, a large number of different household and/or industrial wastes, which are made from different materials. The high-volume material stream 11 is heterogeneous in its composition and contains, for example, different materials from the material groups metal (e.g. iron, aluminum), non-metal (e.g. graphite), organic materials (e.g. wood, plastic), inorganic Non-metals (e.g. ceramics, glass) and/or semiconductors (e.g. silicon). These can be present loosely in the high-volume material stream 11, be separably connected to one another or partially/completely inseparably connected to one another.

A high-volume material flow 11 is understood to be a material flow or mass flow with a volume of at least 1 t/min (metric tons per minute). To put it simply, the high-volume material flow 11 is a mass flow with a volume rate of at least 1 t/min. The high-volume material flow 11 is therefore a substantially continuous mass flow. The high-volume material flow 11 is therefore a substantially heterogeneous material volume flow from household and/or industrial waste. The high-volume material stream 11 consists of a large number of household and/or industrial waste, as is usually delivered to thermal recycling plants by waste collection vehicles.

The high-volume material flow 11 therefore refers to a material flow or mass flow which is characterized by a high volume rate of, for example, at least 1 t/min. A detection of parameters 35 of the high-volume material stream 11 differs significantly from the detection of individual objects in household and/or industrial waste, since due to the high volume rate and the inhomogeneous composition, detection of individual objects in the high-volume material flow 11 is not possible.

The high-volume material stream 11 has a calorific value Hi, the calorific value H _i depending on the composition of the high-volume material stream 11. The calorific value H _i des High-volume material flow 11 is usually 8-20 MJ/kg (megajoules per kilogram).

The process medium 10 in the form of the high-volume material stream 11 is provided via the feed unit 15. The first sensor unit 30 is provided on the feed unit 15. The first sensor unit 30 includes at least one of a color image camera, a mono camera, a high-speed camera, a thermal image camera, an infrared sensor, an X-ray machine, a radar sensor, an event camera, an induction sensor, or a spectral sensor. The first sensor unit 30 is arranged, for example, in a dust-protected housing and, for example, further comprises a lighting device in order to illuminate the process medium 10 in a suitable manner.

The first sensor unit 30 detects at least one parameter 35 of the process medium 10 comprising the high-volume material stream 11. The parameter 35 includes at least one of a grain size of the process medium 10, a particle quality of the process medium 10, a dust concentration of the process medium 10, a moisture content of the process medium 10, a Particle size distribution of the process medium 10, a critical object in the process medium 10 or a biogenic portion of the process medium 10. Furthermore, the parameter 35 can also include, for example, a material type of the process medium 10, a temperature distribution of the process medium 10 or a density quality of the process medium 10.

The processing unit 40 is electronically connected to the first sensor unit 30 and is set up to determine a predicted heating value H _i,p based on the recorded parameters 35 of the supplied process medium 10. The predicted calorific value H _i,p of the process medium 10 describes an expected calorific value of the process medium 10 during thermal utilization in a combustion chamber 50.

The processing unit 40 includes one or more computers, which computer(s) include a processor and a memory. If the processing unit 40 comprises multiple computers, these are multiple computers preferably connected to each other via a network connection. This network connection can, for example, also include a connection between the multiple computers via a secure connection over the Internet (virtual private network; also referred to as a “VPN Internet connection”). The Internet connection is established, for example, via a wired connection or a cellular connection. The processing unit 40 preferably comprises a data collection server 40a, a data processing server 40b and a persistence server 40c. The data collection server 40a, the data processing server 40b and the persistence server 40c may correspond to a physical computer. However, the data collection server 40a, the data processing server 40b and the persistence server 40c may also correspond to a plurality of distributed computers, for example located in different locations and connected via a VPN Internet connection.

The data collection server 40a is preferably provided on the first sensor unit 30 and electrically connected to the first sensor unit 30. The data collection server 40a includes functions to provide a user (not shown) with information about the recorded parameters 35 of the sensor unit 30. The information about the recorded parameters 35 can be provided, for example, via a dashboard (not shown). The data collection server 40a is preferably connected to the data processing server 40b.

The data processing server 40b preferably provides central computing capacity to carry out computationally intensive analyses. Through this centralization of computing capacity, parameters can be obtained from a large number of first sensor units 30 and second sensor units 55 (see Fig. 3 ) are collected, recorded and processed. The data processing server 40b can also collect additional information such as system information of the system 1 or weather information. The information stored on the data processing server 40b is preferably compressed.

The persistence server 40c is preferably connected to the data processing server 40b. The persistence server 40c is preferably used for storing historical information.

The predicted heating value H _i,p of the process medium 10 is determined by means of a forecast unit 60. The forecast unit 60 can be a physical unit or a software unit that is executed in the processing unit 40. The forecasting unit 60 is preferably a software unit. The forecasting unit 60 particularly preferably further comprises a trained machine learning algorithm 60a. The trained machine learning algorithm 60a includes, for example, a machine learning algorithm that was trained using unsupervised learning, supervised learning and/or reinforcement learning.

The forecasting unit 60 includes a plurality of modules 61. The modules 61 may be physical modules or software modules that are executed in the processing unit 40. The modules 61 include computer-implemented functions which are used by the forecasting unit 60 to determine the predicted calorific value H _i,p from the parameters 35 of the process medium 10 comprising the high-volume material flow 11 recorded by the first sensor unit 30.

The modules 61 process the parameters 35 of the process medium 10 recorded by the first sensor unit 30, comprising the high-volume material flow 11. For this purpose, the modules 61 use the detection results of the first sensor unit 30 provided in the form of a large number of individual images and/or individual information. This large number of individual images and/or individual pieces of information are preferably provided with a unique timestamp in order to be able to link the parameters 35 recorded by the various sensors of the first sensor unit 30 with one another.

The determination of the predicted heating value H _i,p of the process medium 10 initially includes pre-processing of the parameters 35 detected by the first sensor unit 30. The pre-processing takes place in a synchronization module 61a, a linking module 61b, an activity module 61c and a segmentation module 61d. For example, the process medium 10 can be used over a period of time t from the first Sensor unit 30 can be examined and, using the modules 61a-61k, for example an averaged predicted heating value Hi,p can be determined for this period t. Such an assessment of the process medium 10 over a period of time t makes it possible to make statements about the predicted calorific value Hi,p of a larger amount of substance or material.

Fig. 2a shows an overview of the synchronization module 61a for the synchronization of the recorded parameters 35. The synchronization module 61a includes an acquisition logic 61a-1 for the acquisition of the color images from the color camera, an acquisition logic 61a-2 for the acquisition of the grayscale images from the mono camera, an acquisition logic 61a- 3 for capturing the event images from the event camera, an acquisition logic 61a-4 for capturing the depth images from the stereo/3D camera, an acquisition logic 61a-5 for capturing the thermal images from the thermal camera, an acquisition logic 61a-6 for capturing the Signal images from the radar sensor, an acquisition logic 61a-7 for acquiring the spectral images from the spectral sensor and a fusion logic 61a-8. The fusion logic 61a-8 synchronizes the individual images and/or individual information acquired by the acquisition logics 61a-1 to 61a-7 based on the unique timestamp. Furthermore, the fusion logic 61a-8 synchronizes the individual images and/or individual information captured by the acquisition logics 61a-1 to 61a-7 with regard to a possible image shift or rotation of the captured individual images and/or individual information, which is caused, for example, by a different arrangement of the sensors of the first Sensor unit 30 can result.

Fig. 2b shows an overview of the linking module 61b for linking the recorded parameters 35 with image recordings of the process medium 10, the image recordings comprising, for example, color images or grayscale images of the process medium 10. The linking module 61b receives the individual images and/or individual information processed by the synchronization module 61a. The linking module 61b includes a detection logic 61b-1 for detecting the process medium 11. The linking module 61b further includes a fusion logic 61b-2 and a fusion logic 61b-3. The fusion logic 61b-2 essentially corresponds to the fusion logic 61a-8 and will therefore not be described again. The fusion logic 61b-3 synchronizes the recorded parameters 35 with the image recordings of the process medium 10.

Fig. 2c shows an overview of the activity module 61c for detecting active regions in the process medium 11. The activity module 61c receives the individual images and/or individual information processed by the linking module 61b. The activity module 61c includes an analysis logic 61c-1 for pre-processing the individual images and/or individual information from the parameters 35 recorded by the first sensor unit 35 for cleansing as well as normalization, a detection logic 61c-2 for detecting active image regions in color images, taking into account a temporal History, a detection logic 61c-3 for detecting active image regions in event images taking the time course into account, a detection logic 61c-4 for detecting active image regions in depth images taking the time course into account, a detection logic 61c-5 for detecting active image regions in thermal images taking into account the time course and a fusion logic 61c-6 for detecting adequate active regions by taking into account the different modalities from the analysis logic 61a-2 and the detection logics 61c-2 to 61c-5.

Fig. 2d shows an overview of the segmentation module 61d for detecting segments in the process medium 11. The segmentation module 61d receives the individual images and/or individual information processed by the fusion logic 61c-6 of the activity module 61c. The segmentation module 61d includes a preprocessing logic 61d-1 for preprocessing the image data for segmentation, a detection logic 61d-2 for detecting segments by grouping textures, edges and color gradients in the color image, a detection logic 61d-3 for detecting segments by grouping activity signals in event images, a detection logic 61d-4 for capturing segments by grouping spatial gradients in the depth image, a detection logic 61d-5 for capturing segments by grouping the temperature gradients in the thermal image, a detection logic 61d-6 for capturing segments using a neural network for image segmentation based on the modalities of color image, grayscale image, depth image, thermal image, event image, spectral image, radar image and a fusion logic 61d-7 to capture adequate segments through the Taking into account the results from the different modalities from the preprocessing logic 61d-1 and the detection logics 61d-2 to 61d-6.

The determination of the predicted calorific value H _i,p further includes further processing of the preprocessed parameters 35 detected by the first sensor unit 30. The determination of the predicted calorific value H _i,p includes a determination of at least one of the grain size of the process medium 10, a particle quality of the process medium 10, a dust concentration of the process medium 10, a moisture content of the process medium 10, a particle distribution of the process medium 10, a critical object in the process medium 10 or a biogenic portion of the process medium 10.

For this purpose, the modules 61 include at least one of a module 61e for detecting a grain size, a module 61f for detecting a particle texture, a module 61g for detecting a dust concentration, a module 61h for detecting a moisture, a module 61i for detecting a particle size distribution, a module 61j for detecting critical objects, a module 61k for detecting a biogenic fraction or a module 61l for detecting metainformation.

Fig. 2e shows an overview of module 61e for detecting the grain. The module 61e for detecting the grain size uses, for example, the color image captured by the color image camera of the first sensor unit 30 to determine the grain size of the high-volume material stream 11 based on the captured color image. To determine the grain size, methods from deep learning and machine learning are combined. The grain detection module 61e may be a separate physical module. However, the module 61e for detecting the grain size can also be executed by the processing unit 40 and/or be an integral part of the processing unit 40.

The module 61e receives the individual images and/or individual information processed by the fusion logic 61c-6 of the activity module 61c for detecting the grain. The module 61e for detecting the grain size includes analysis logic 61e-1 for extracting so-called quantification vectors that describe the active regions, an analysis logic 61e-2 for forming a prediction model based on quantification vectors with the aim of classifying the grain size and an analysis logic 61e-3 for preparing and providing the grain size results. In the analysis logic 61e-3, the grain in the active regions of the process medium 10 is classified using an absolute or relative scale. This classification of grain size can be represented, for example, in the form of a continuous relative range from 0.0 (fine) to 1.0 (coarse) or in a metric range, for example from 10cm to 100cm.

Fig. 2f shows an overview of the module 61f for detecting the particle properties. The particle condition includes a particle velocity and/or a particle activity. For example, the particle texture detection module 61f uses the event image captured by the event camera of the first sensor unit 30 to determine the distribution of the particle velocity and the orientation of the particle velocity based on the captured event image. Furthermore, the event image is used to determine the activity of the high-volume material stream 11 in the observation angle of the event camera. To determine the particle properties, methods from deep learning, such as neural networks, and methods from the field of machine learning are combined to determine the speed, orientation and activity of the particles. The module 61f for detecting the particle nature can be a separate physical module. However, the module 61e for detecting the particle properties can also be carried out by the processing unit 40 and/or be an integral part of the processing unit 40.

The module 61f for detecting the particle properties receives the individual images and/or individual information processed by the fusion logic 61d-7 of the segmentation module 61d for detecting the particle properties. The module 61f for detecting the particle properties includes an analysis logic 61f-1 for determining the particle movement (including speed and orientation) in the color image, an analysis logic 61f-2 for determining the particle movement (including speed and orientation) from event images, an analysis logic 61f-3 to determine the particle movement (including speed and orientation) in the depth image, an analysis logic 61f-4 for the Consideration of additional tracking methods to enable an assessment of particle speed and orientation, a fusion logic 61f-5 to capture adequate particles taking into account the particle movement to create a distribution regarding speed, orientation and size and an analysis logic 61f-6 for the Assessment and provision of results of particle properties. The analysis in the module 61f for detecting the particle properties is carried out, for example, with the aid of an optical flow method using the color images received from the event camera of the first sensor unit 30.

Fig. 2g shows an overview of the module 61g for detecting the dust concentration. The module 61g for detecting the dust concentration uses individual images and/or individual information captured by the first sensor unit 30 to determine an intensity of detected dust regions in the high-volume material stream 11 based on the captured individual images and/or individual information. To determine the dust concentration, methods from deep learning (neural networks) are combined with methods from machine learning to record the dust concentration. The dust concentration detection module 61g may be a separate physical module. However, the module 61g for detecting the dust concentration can also be carried out by the processing unit 40 and/or be an integral part of the processing unit 40.

The module 61g for detecting the dust concentration receives the individual images and/or individual information for detecting the dust concentration processed by the fusion logic 61d-7 of the segmentation module 61d. The dust concentration detection module 61g includes an analysis logic 61g-1 for identifying the degree of dust from color images, an analysis logic 61g-2 for identifying the degree of dust from thermal images, an analysis logic 61g-3 for identifying the degree of dust from grayscale images, an analysis logic 61g-4 for identifying the degree of dust from radar images, an analysis logic 61g-5 for identifying the degree of dust from spectral images, an analysis logic 61g-6 for identifying the degree of dust based on neural networks for detecting diffuse image regions, a fusion logic 61g-7 for detecting adequate dust segments with labeling the dust intensity and one Analysis logic 61g-8 for evaluating and providing dust concentration results.

Fig. 2h shows an overview of module 61h for detecting moisture. The module 61h for detecting the moisture uses, for example, individual images and/or individual information captured by the first sensor unit 30 in order to determine a distribution and intensity of detected moisture regions of the high-volume material flow 11 based on the captured individual images and/or individual information. To record moisture, methods from deep learning (neural networks) are combined with methods from machine learning. The moisture sensing module 61h may be a separate physical module. However, the module 61h for detecting the moisture can also be carried out by the processing unit 40 and/or be an integral part of the processing unit 40.

The module 61h for detecting the humidity or the degree of humidity receives the individual images and/or individual information for detecting the humidity processed by the fusion logic 61d-7 of the segmentation module 61d. Regions with a high degree of humidity are detected by analyzing the texture in the individual images and/or individual information captured by the color image camera. This texture quality can be determined, for example, based on a contrast, a brightness and/or a reflectance in partial areas of the color images. The moisture detection module 61h includes an analysis logic 61h-1 for identifying the moisture level from color images, an analysis logic 61h-2 for identifying the moisture level from thermal images, an analysis logic 61h-3 for identifying the moisture level from grayscale images, an analysis logic 61h-4 for identifying the degree of humidity from radar images, an analysis logic 61h-5 for identifying the degree of humidity from spectral images, an analysis logic 61h-6 for identifying the degree of humidity based on neural networks, a fusion logic 61h-7 for detecting adequate segments with identification of the degree of humidity and an analysis logic 61h-8 for assessing and providing moisture level results.

Fig. 2i shows an overview of the module 61i for recording the particle size distribution. The module 61i for recording the particle size distribution uses, for example, individual images and/or individual information captured by the first sensor unit 30 to determine a distribution of the particle size of the high-volume material stream 11 based on the captured individual images and/or individual information. To record the particle size distribution, methods from deep learning (neural networks) are combined with methods from machine learning. The module 61i for detecting the particle size distribution can be a separate physical module. However, the module 61i for detecting the particle size distribution can also be carried out by the processing unit 40 and/or be an integral part of the processing unit 40.

The module 61i for detecting the particle size distribution receives the individual images and/or individual information processed by the fusion logic 61d-7 of the segmentation module 61d for detecting the particle size distribution. The module 61i for detecting the particle size distribution includes an analysis logic 61i-1 for linking the segmentation results with 3D depth information for a detection of potential particle candidates with an associated estimate of the spatial dimensions of the particle candidates, a fusion logic 61i-2 for detecting adequate particles with identification of the Particle size and analysis logic 61i-3 for evaluating and providing the results of the particle size distribution.

Fig. 2j shows an overview of the module 61j for detecting critical objects. The module 61j for detecting critical objects uses individual images and/or individual information captured by the first sensor unit 30 to determine critical objects in the high-volume material stream 11 based on the captured individual images and/or individual information. To detect critical objects, methods from deep learning (neural networks) are combined with methods from machine learning. The critical object detection module 61j may be a separate physical module. However, the module 61j for detecting critical objects can also be executed by the processing unit 40 and/or be an integral part of the processing unit 40.

The critical object detection module 61j receives the individual images and/or processed by the fusion logic 61d-7 of the segmentation module 61d Individual information for detecting critical objects. The module 61j for detecting critical objects includes an analysis logic 61j-1 for linking the segmentation results with 3D depth information for an object candidate with an associated estimate of the spatial dimensions of the object candidates, a fusion logic 61j-2 for detecting adequate particles with identification of the particle size and an analysis logic 61j-3 for evaluating and providing the results of the particle size distribution.

Fig. 2k shows an overview of module 61k for recording the biogenic portion. To record the biogenic component, methods from deep learning (neural networks) are combined with methods from machine learning. The module 61k for detecting the biogenic portion can be its own physical module. However, the module 61k for detecting the biogenic portion can also be carried out by the processing unit 40 and/or be an integral part of the processing unit 40.

The module 61k for detecting the biogenic portion receives the individual images and/or individual information processed by the fusion logic 61d-7 of the segmentation module 61d for detecting the biogenic portion. Regions with a high biogenic content are detected by analyzing the texture in the color images captured by the color image camera. This texture quality can be determined, for example, based on a contrast, a brightness and/or a reflectance in partial areas of the color images. The module 61k for detecting the biogenic portion includes an analysis logic 61k-1 for identifying the biogenic portion from color images, an analysis logic 61k-2 for identifying the biogenic portion from thermal images, an analysis logic 61k-3 for identifying the biogenic portion from grayscale images, a Analysis logic 61k-4 for identifying the biogenic portion from radar images, an analysis logic 61k-5 for identifying the biogenic portion from spectral images, an analysis logic 61k-6 for identifying the biogenic portion based on neural networks, a fusion logic 61k-7 for detecting adequate segments with identification of the biogenic portion and an analysis logic 61k-8 for evaluating and providing the results of the biogenic portion.

Fig. 2l shows an overview of module 61l for collecting metadata. The module 61l for collecting the metadata can be a separate physical module. However, the module 61k for acquiring the metadata can also be executed by the processing unit 40 and/or be an integral part of the processing unit 40.

The module 61k for capturing the metadata receives the individual images and/or individual information processed by the fusion logic 61b-3 of the linking module 61b for capturing the metadata. The module 61k for capturing the metadata includes a capture logic 611-1 for information from external data sources (e.g. weather, season) taking into account the time and location of the sensor data recordings, a capture logic 61l-2 for system information from the thermal utilization plant (e.g. combustion data, emissions data) taking into account the time and location of the sensor data recordings and a fusion logic 61l-3 for a summary of the information to make the metadata available for further analysis.

The module 61k for acquiring the metadata uses, for example, system information of the system 1 acquired by the data processing server 40b or, for example, weather information, in order to obtain information linked to the acquired results of the modules 61a-61k. System information and external information are linked and synchronized with the results of the modules via locality and time history in order to make this meta-information available for further analysis.

By combining the raw data from the modules 61a-61k in one or more prediction models, it is possible to continuously predict the calorific value H _i of the high-volume material stream 11. The forecast of the predicted calorific value H _i,p is based on the parameters 35 of the process medium 10, which are determined by the first sensor unit 30, comprising the high-volume material flow 11. The forecast of the predicted calorific value H _i,p is carried out using the modules 61a-61k. The individual images and/or individual information processed in the modules 61a-61k to evaluate the parameters 35 of the material flow enable the Prognosis unit 60 to provide a reliable statement about a predicted calorific value H _i,p of the recorded process medium 10. This predicted heating value H _i,p can be determined, for example, for a specific section of the process medium 10.

By predicting the predicted calorific value H _i,p, the combustion of the process medium 10 in the combustion chamber 50 and its parameters can be anticipated before the process medium is actually burned. This enables proactive control of the combustion process in the combustion chamber 50. For example, fire-accelerating or fire-retarding additives can be added to the process medium 10 or, for example, parameters of the combustion process in the combustion chamber 50, such as oxygen supply, can be proactively adjusted. A proactive control therefore makes it possible for the composition of the process medium 10 for supply into the combustion chamber 50 to be actively changed in order to enable thermal utilization in the optimal combustion calorific value range of the combustion chamber 50 (see also the further configurations of the system which use this control the composition by means of a composition control unit 20 or a buffer memory 25).

The forecast of the predicted calorific value H _i,p makes it possible to reduce system wear and thus the maintenance costs for the operation of thermal recycling plants. In addition, by predicting the predicted calorific value H _i,p, the combustion chamber 50 can be operated in the optimal combustion calorific value band and thereby enable an optimal throughput (in tons) of the process medium 10 into the combustion chamber 50, which in turn enables the best possible commercial use of the process medium 10.

Fig. 3 shows a schematic representation of a system 2 for a calorific value control in a thermal utilization plant using a predicted calorific value H _i,p . System 2 essentially corresponds to system 1. Those components of system 2 that essentially correspond to those of system 1 will not be described again. The same reference numbers are used for those components of system 2 that essentially correspond to those of system 1 used. Compared to system 1, system 2 further comprises at least the combustion chamber 50 and a composition control unit 20.

The composition of the high-volume material stream 11 determines, for example, its suitability for thermal utilization. During thermal recycling, the high-volume material stream 11 is fed to the combustion chamber 50 and converted into a thermally recycled process medium 12. The thermal utilization of the high-volume material stream 11 includes combustion of the high-volume material stream 11 in the combustion chamber 50. The thermally utilized process medium consists of a solid phase 13 and a gaseous phase 14. The solid phase 13 is usually disposed of, whereas the gaseous phase 14 becomes energy - and heat generation is directed to a steam generator. After cooling on the steam generator, the gaseous phase 14 is fed to one or more cleaning steps and then released into the ambient air.

The combustion of high-volume material stream 11 can be carried out, for example, by adding atmospheric oxygen or fire-accelerating additives. Between a time when the at least one parameter 35 of the high-volume material stream 11 is recorded and a time when the high-volume process stream 11 is thermally utilized in the combustion chamber 50, 30 - 90 minutes usually pass (combustion time window). For this reason, the combustion time window of the process medium is taken into account for predicted heating values. In order to be able to proactively control the thermal utilization of the high-volume material stream 11, the system 2 has the composition control unit 20

The composition control unit 20 controls the composition of the high-volume material stream 11 before it is fed into the combustion chamber 50. The composition control unit 20 makes it possible to specifically control the calorific value H _i of the high-volume material stream 11. This optimizes the thermal utilization process and the addition of fire-accelerating additives can be reduced or particularly advantageously avoided. This control of the high-volume material flow 11 takes place taking into account the calorific value H _i,p predicted by the forecasting unit 60.

The composition control unit 20 can be designed, for example, as a separation device. In this embodiment, composition control unit 20 separates out components of the high-volume material stream 11 that have a detrimental effect on the predicted calorific value H _i,p of the high-volume material stream 11. For example, the composition control unit 20 is able to separate out components of the high-volume material stream 11 in which at least one parameter 35 was determined by at least one of the modules 61 to be disadvantageous for the thermal recycling process.

The system 2 further comprises the second sensor unit 55. The second sensor unit 55 is set up to determine the actual calorific value Hi,t of the thermally utilized process medium 12 on the basis of recorded parameters of the combustion chamber 50 that are related to the thermal utilization of the high-volume material stream 11. The second sensor unit 55 is preferably arranged on the combustion chamber 50 and records the parameters of the thermal utilization of the high-volume material stream 11 during combustion in the combustion chamber 50. The following combustion parameters, among others, are considered for the determination: primary air and secondary air of the volume flow, primary air pressure (e.g Rust zone), heating oil and/or gas consumption, steam to the primary air preheater, live steam quantity, primary air temperature, CO boiler outlet, CO2 boiler outlet, HCL boiler outlet, SO2 boiler outlet, oxygen content boiler outlet, moisture boiler outlet, NOx boiler outlet, volume flow, crane scale throughput, daily throughput. The combustion parameters are merged together with the recordings of the modules 61 in order to enable an overall uniform representation of combustion parameters, including sensory perception, for adequate predictive modeling of the calorific value forecast.

The combination of the recorded parameters 35, the evaluations from the module 61 and the combustion parameters recorded by the second sensor unit 55 enables the forecasting unit 60 to predict the predicted calorific value H _i,p . The forecasting unit 60 is preferably able to predict the predicted calorific value H _i,p in a future time window (60-90 minutes) for each incoming medium (e.g. individual deliveries of the process medium 11). This creates future ones Calorific values with overlapping time windows in the future. Thus, for a specific point in time in the future, the predicted calorific value H _i,p can be determined by fusion based on the overlapping calorific value forecasts.

The forecasting unit 60 includes the trained machine learning algorithm 60a. The trained machine learning algorithm 60a can be used to continuously adjust the calorific value forecast using the agents 61 and to improve the forecast accuracy. For this purpose, the trained prediction model for heating value prediction is continuously updated during the operation of the trained machine learning algorithm 60a ("incremental online learning") in order to take fluctuations in the composition of the process medium 10 over time into account in the model and to prevent drift or deviation of the model ( English "concept drift").

Fig. 4 shows a schematic representation of the system 2 for a calorific value control in a thermal utilization plant using a predicted calorific value H _i,p comprising a second embodiment of the composition control unit 20. The composition control unit 20 of the second embodiment can be designed, for example, as a buffer storage 25. This buffer storage 25 includes a large number of bunkers 25a-25d which are designed for buffering or intermediate storage of the high-volume material flow 11. In this second embodiment, the composition control unit 20 controls the high-volume process stream 11 in such a way that process media 10 with essentially homogeneous predicted heating values H _{i,p_a-d} are temporarily stored in the bunkers 25a-25d. The composition control unit 20 supplies the process media 10 to the combustion chamber 50 as needed in order to enable consistent thermal use of the process medium 10. The process media 10 is supplied as needed from the bunkers 25a-25d using the predicted heating values H _{i,p_a-d} of the process media in the bunkers 25a-25d.

Fig. 5 is a process flow diagram depicting a method 100 for determining a predicted heating value H _i,p of a process medium 10. The method 100 includes providing the process medium 10 in step S100 by a Feed unit 15. The method 100 further comprises detecting in step S110 at least one parameter 35 of the process medium 10 by means of the first sensor unit 35.

The method 100 also includes determining in step S120 the predicted calorific value H _i,p based on the detected parameters 35 of the provided process medium 10 in the processing unit 40 by means of the forecasting unit 60. Determining the predicted calorific value H _i,p based on the detected parameters 35 is carried out by the forecasting unit 60 using the modules 61.

Fig. 6 is a process flow diagram depicting a computer-implemented method 200 for calorific value control in a thermal utilization plant using a predicted calorific value H _i,p .

The computer-implemented method 200 includes providing in step S200 a process medium 10 by a supply unit 15. Providing in step S200 essentially corresponds to providing in step S100.

The computer-implemented method 200 further includes detecting in step S210 at least one parameter 35 of the process medium 10 by means of the first sensor unit 35. The detecting in step S210 essentially corresponds to the detecting in step S110.

The computer-implemented method 200 also includes determining in step S220 the predicted heating value H _i,p based on the recorded parameters 35 of the provided process medium 10 in the processing unit 40 by means of the forecasting unit 60. The determining in step S220 essentially corresponds to the determining in step S120 .

Compared to the method 100, the computer-implemented method 200 additionally includes controlling in step S225 a composition of the provided process medium 10 by means of a composition control unit 20 using the predicted calorific value H _i,p of the provided process medium 10.

The computer-implemented method 200 further includes thermal utilization in step S230 of the provided process medium 10 in the combustion chamber 50 to form a thermally utilized process medium 12.

The computer-implemented method 200 also includes detecting in step S240 an actual calorific value Hi,t of the thermally utilized process medium 12 by means of a second sensor unit 55.

The computer-implemented method 200 also includes training in step S250 of the machine learning algorithm 60a, taking into account the actual calorific value Hi,t of the thermally utilized process medium 12 and the predicted calorific value H _i,p of the process medium 10. Training S250 of the forecasting unit 60 taking into account the actual calorific value Hi,t of the thermally utilized process medium 10 and the predicted calorific value H _i,p of the thermally utilized process medium 12 includes an iterative comparison of the actual calorific value Hi,t and the predicted calorific value H _i,p by the processing unit 40.

Reference symbol list

1

system

2

system

10

Process medium

11

high-volume material flow

12

thermally recycled process medium

13

solid phase

14

gaseous phase

15

Feeding unit

20

Composition control unit

25

Buffer memory

30

first sensor unit

35

parameter

40

Processing unit

40a

Data collection server

40b

Data processing server

40c

Persistence server

50

combustion chamber

55

second sensor unit

60

Forecasting unit

60a

Machine learning algorithm

61

Modules

Hi

calorific value

Hip

predicted calorific value

Hit

actual calorific value

100

Computer-implemented method

200

Computer-implemented method

Claims (15)

Computer-implemented method (100) for determining a predicted calorific value (H _i,p ) of a process medium (10) in a thermal recycling plant, the method (100) comprising: Providing (S100) the process medium (10) by a supply unit (15), the process medium (10) comprising a high-volume material stream (11); Detecting (S110) at least one parameter (35) of the process medium (10) by means of a first sensor unit (30), wherein detecting (S110) the at least one parameter (35) of the process medium (10) involves detecting at least one of a grain size of the Process medium (10), a particle quality of the process medium (10), a dust concentration of the process medium (10), a moisture of the process medium (10), a particle size distribution of the process medium (10), a critical object in the process medium (10) or a biogenic proportion the process medium (10); and Determining (S120) a predicted calorific value (H _i,p ) based on the recorded parameters (35) of the provided process medium (10) in a processing unit (40) by means of a forecasting unit (60). Computer-implemented method (200) for calorific value control in a thermal utilization plant using a predicted calorific value (H _i,p ), the method (200) comprising: Providing (S200) a process medium (10) by a feed unit (15), the process medium (10) comprising a high-volume material stream (11); Detecting (S210) at least one parameter (35) of the process medium (10) by means of a first sensor unit (30), wherein detecting (S210) the at least one parameter (35) of the process medium (10) involves detecting at least one of a grain size of the Process medium (10), a particle quality of the process medium (10), a dust concentration of the process medium (10), a moisture of the process medium (10), a particle size distribution of the process medium (10), a critical object in the process medium (10) or a biogenic proportion the process medium (10); Determining (S220) the predicted calorific value (H _i,p ) based on the recorded parameters (35) of the provided process medium (10) in a processing unit (40) by means of a forecasting unit (60); and Controlling (S225) a composition of the provided process medium (10) by means of a composition control unit (20) using the predicted calorific value (H _i,p ) of the provided process medium (10). The computer-implemented method (100, 200) of claim 1 or 2, further comprising:
thermally utilizing (S230) the provided process medium (10) in a combustion chamber (50) to form a thermally utilized process medium (12). The computer-implemented method (100, 200) of claim 3, further comprising:
Detecting (S240) an actual heating value (H _i,t ) of the thermally utilized process medium (12) by means of a second sensor unit (55). Computer-implemented method (100, 200) according to any one of claims 1 to 4, further comprising:
Training (S250) of a machine learning algorithm (60a) taking into account the actual calorific value (H _i,t ) of the thermally utilized process medium (12) and the predicted calorific value (H _i,p ) based on the recorded parameters (35) of the process medium (10) , wherein the forecasting unit (60) comprises the machine learning algorithm (60a). The computer-implemented method (100, 200) of claim 5, wherein:
the training (S250) of the machine learning algorithm (60a), taking into account the actual calorific value (H _i,t ) of the thermally utilized process medium (12) and the predicted calorific value (H _i,p ) of the process medium (10), an iterative comparison of the actual calorific value ( H _i,t ) and the calorific value (H _i,p ) predicted by the processing unit (40) based on the recorded parameters (35) of the process medium (10). Computer-implemented method (100, 200) according to one of claims 1 to 6, in which
the detection (S110, S210) of the at least one parameter (35) of the process medium (10) includes detecting at least one parameter (35) of the high-volume material flow (11), wherein the high-volume material flow (11) is a continuous and essentially heterogeneous material volume flow from household and/or industrial waste, and wherein the high-volume material flow (11) is a mass flow with a volume rate of at least 1 t/min. Computer-implemented method (100, 200) according to any one of claims 1 to 7, wherein
detecting (S110, S210) the at least one parameter (35) of the process medium (10) with the help of at least one of a color image camera, a mono camera, a high-speed camera, a thermal image camera, an infrared sensor, an X-ray device, a radar sensor, an event ( Event) camera, an induction sensor or a spectral sensor. System (1) for determining a predicted calorific value (H _i,p ) of a process medium (10) in a thermal recycling plant, the system (5) comprising: a feed unit (15) for providing the process medium (10), the process medium (10) comprising a high-volume material stream (11); a first sensor unit (30) set up to detect at least one parameter (35) of the process medium (10); and a processing unit (40), electronically connected to the sensor unit (30) and set up to determine a predicted heating value (H _i,p ) based on the recorded parameters (35) of the supplied process medium (10). System (2) for a calorific value control in a thermal utilization plant using a predicted calorific value (H _i,p ), the system (2) comprising: a feed unit (15) for providing the process medium (10), the process medium (10) comprising a high-volume material stream (11); a first sensor unit (30) set up to detect at least one parameter (35) of the process medium (10); a processing unit (40), electronically connected to the first sensor unit (30) and set up to determine a predicted heating value (H _i,p ) based on the detected parameters (35) of the supplied process medium (10); a combustion chamber (50) for thermally utilizing the provided process medium (10) into a thermally utilized process medium (12); and a composition control unit (20) is set up to control a composition of the supplied process medium (10). System (2) according to claim 9 or 10, further comprising:
a second sensor unit (55), set up to determine the actual calorific value (Hi,t) of the thermally utilized process medium (12). System (1, 2) according to one of claims 9 to 11, wherein
the first sensor unit (30) comprises at least one of a color image camera, a mono camera, a high-speed camera, a thermal image camera, a 3D camera, an infrared sensor, an X-ray device, a radar sensor, an event (event) camera, an induction sensor or a spectral sensor. System (1, 2) according to one of claims 9 to 12, wherein
the high-volume material flow (11) is a continuous and essentially heterogeneous material volume flow from household and/or industrial waste. System (1, 2) according to one of claims 9 to 13, wherein
the high-volume material flow (11) is a mass flow with a volume rate of at least 1 t/min. Use of the computer-implemented methods according to one of claims 1 to 8 or the devices according to one of claims 9 to 14 in a thermal waste treatment or recycling plant.

Images