A method and system for sorting a metal-comprising object in a stream of metal-comprising objects

Abstract

The present disclosure relates to a method for sorting a metal- comprising object in a stream of metal-comprising objects, the method comprising measuring contamination levels over a surface of the metal-comprising object. The method further comprises locating at least one point on the surface where the contamination level is below a threshold level and directing a laser beam at the at least one point to locally form a plasma and measuring a spectrum of the plasma; or directing a laser beam at the at least one point to form an optical signal comprising a plurality of Raman wavelengths and measuring a spectrum of the Raman wavelengths. The method further comprises assigning a class to the metal-comprising object, based on the measured spectrum, and sorting the metal- comprising object into one of a first output fraction (A) or a second output fraction (B), based on the assigned class.

Description

[0001] A METHOD AND SYSTEM FOR SORTING A METAL-COMPRISING OBJECT IN A STREAM OF METAL-COMPRISING OBJECTS

[0002] Field of invention

[0003] The present invention generally relates to method for sorting a metalcomprising object in a stream of metal-comprising objects. The present invention also generally relates to a system for sorting a metal-comprising object in a stream of metal-comprising objects.

[0004] Technical Background

[0005] With the present-day focus on sustainability and green technology, recycling of materials is of ever-increasing importance. As the mining of virgin metal materials is both expensive, environmentally destructive, and often not according to criteria for sustainability, the recycling of metals contained in discarded objects is both necessary and highly desirable from an economic point of view. Once discarded metal-comprising objects have been roughly separated from non-metallic discarded objects, a further sorting should be performed, to classify and sort the metal-comprising objects and thereby facilitate a more efficient recycling process. Therefore, there exists a clear need to provide an improved method and system for sorting metal-comprising objects in a stream of metal-comprising objects.

[0006] Summary of the invention

[0007] In view of the above, it is an object of the present disclosure to provide an improved method for sorting a metal-comprising object in a stream of metal-comprising objects. The method comprises measuring contamination levels over a surface of the metal-comprising object, locating at least one point on the surface where the contamination level is below a threshold level, directing a laser beam at the located point to locally form a plasma and measuring a spectrum of the plasma; or directing a laser beam at the located point to form an optical signal comprising a plurality of Raman wavelengths and measuring a spectrum of the Raman wavelengths, the method further comprising, assigning a class to the metal-comprising object, based on the measured spectrum, and sorting the metal-comprising object into one of a first fraction or a second fraction, based on the assigned class.

[0008] According to an embodiment contamination levels are measured by a visible light sensor, a UV sensor, and / or a NIR sensor.

[0009] According to an embodiment measuring the spectrum of the plasma is performed by LIBS, or measuring the spectrum of the plurality of Raman wavelengths is performed by Raman spectroscopy, such as SERS.

[0010] According to an embodiment, locating at least one point further comprises object detection by laser-triangulation and I or image analysis.

[0011] According to an embodiment, object detection further includes height detection, based on time of flight techniques, structured light techniques, optical disparity techniques and I or output from height detectors.

[0012] According to an embodiment the method further comprises cleaning the object and optionally, drying the object, before measuring contamination levels.

[0013] According to an embodiment the method further comprises re-cleaning the object if no point can be located with a contamination level below the threshold level. According to an embodiment cleaning and I or re-cleaning comprises sanding, spraying with liquids, mechanical cleaning, chemical cleaning, high pressure water cleaning, laser-based cleaning, or any combination thereof.

[0014] According to an embodiment the method further comprises, if a plurality of points on the surface have a contamination level below the threshold level, selecting one point from the plurality of points and directing the laser beam to the selected point.

[0015] According to an embodiment directing the laser beam is performed by manipulating a movably arranged mirror reflecting the laser beam.

[0016] According to an embodiment assigning a class is performed by a machine learning algorithm .

[0017] According to an embodiment sorting of the metal comprising object into one of a first fraction or a second fraction is based on one of metal characteristics of the metal-comprising object, or alloy composition of the metal-comprising object.

[0018] It is also an object of the present disclosure to provide an improved system for sorting a metal-comprising object in a stream of metal-comprising objects. The system comprises a measuring unit for measuring contamination levels over a surface of the metal-comprising object, a laser unit comprising a laser direction unit and a spectrometer, and a processing unit comprising circuitry configured to receive measured contamination levels from the at least one measuring unit, locate at least one point on the surface where contamination levels are below a threshold level, and transmit the located point to the laser direction unit, wherein the laser direction unit is configured to form a plasma, or an optical signal comprising a plurality of Raman wavelengths, at the located point; wherein the spectrometer is configured to measure a spectrum of the plasma, or of the plurality of Raman wavelengths; wherein the circuitry is further configured to receive the measured spectrum from the spectrometer, and to assign a class to the metal-comprising object based on the received spectrum, and wherein the system further comprises at least one sorting unit for sorting the metal-comprising object into one a first fraction, or a second fraction, based on the assigned class.

[0019] According to a further embodiment, the measuring unit comprises a sensor, such as a visible light sensor, a UV sensor, and / or a NIR sensor.

[0020] According to a further embodiment, the laser unit is a LIBS unit, or a Raman spectroscopy unit, such as a SERS unit.

[0021] According to a further embodiment, the system comprises a cleaning unit and optionally, a drying unit.

[0022] According to a further embodiment, the cleaning unit is one of a sanding unit, a liquid spraying unit, a mechanical cleaning unit, a chemical cleaning unit, a high-pressure water cleaning unit, a laser-based cleaning unit, or any combination thereof.

[0023] According to a further embodiment, the circuitry of the processing unit is further configured, if a plurality of points on the surface have a contamination level below the threshold level, to select one point from the plurality of points and to direct the laser beam to the selected point.

[0024] According to an embodiment the laser direction unit comprises at least one movably arranged mirror. According to an embodiment the circuitry is further configured to assign a class by means of a machine learning algorithm.

[0025] According to a further embodiment the sorting unit is configured to sort the metal-comprising object based on one of metal characteristics of the metal-comprising object, or alloy composition of the metal-comprising object.

[0026] Brief description of the drawings

[0027] The aspects of the present inventive concept, including its particular features and advantages, will be readily understood from the following detailed description and the accompanying drawings. The figures are provided to illustrate the general structures of the present inventive concept. Like reference numerals refer to like elements throughout.

[0028] Fig. 1 A is a schematic diagram of the system according to the invention.

[0029] Fig. 1 B is a detailed schematic diagram of a part of the system according to the invention.

[0030] Fig. 2 is a flow chart for contamination detection and sorting of metalcomprising objects according to the inventive concept.

[0031] Fig. 3 is a flow chart for contamination detection and sorting of metalcomprising objects according to a further embodiment.

[0032] Detailed description

[0033] The inventive concept may be implemented in many different forms and should not be construed as limited to the variants set forth herein; rather, these variants are provided for thoroughness and completeness, and fully convey the scope of the present inventive concept to the skilled person.

[0034] With reference to fig. 1A, the system 1 according to the invention includes and input side T and a output side 1”. The system also includes a conveyor 2, for transporting a stream of metal-comprising objects from the input side T to the output side 1”. The general transport direction of the conveyor 2 is indicated by the solid arrow in fig. 1 . The conveyor 2 may include at least one conveyor belt. The conveyor 2 may be driven by a drive unit 2A. The conveyor 2 includes a transporting surface, on which the metalcomprising objects are transported. The conveyor 2 may have a speed of 0,2 - 20 m / s, preferably 0,4 - 15 m / s, most preferably 1 - 10 m / s.

[0035] The system 1 further includes a measuring unit 3, a laser unit 4B, and a sorting unit 5, arranged along the conveyor 2, from the input side T to the output side 1”. The system additionally includes a processing unit 6. The system may further include a positioning unit 4A, arranged between the measuring unit 3 and the laser unit 4B. The measuring unit 3, positioning unit 4A, laser unit 4, and sorting unit 5 are coupled to the processing unit 6. Preferably, the measuring unit 3, positioning unit 4A, laser unit 4B, and sorting unit 5 are placed over the transporting surface of the conveyor 2. In operation, the stream of metal-comprising objects is transported by the conveyor 2 from the input side T to the output side 1”, passing, respectively, the measuring unit 3, the positioning unit 4A, the laser unit 4B, and the sorting unit 5. The measuring unit 3, positioning unit 4A and laser unit 4B are detailed further hereinbelow, with reference to fig. 1 B.

[0036] The sorting unit 5 is arranged to sort metal-comprising objects into one of two or more output fractions, based on instructions from the processing unit 6. Thereto, the sorting unit 5 includes sorting means, such as one or more robotic arms, one or more pushers, or one or more nozzles, or any other device capable of moving objects. The sorting means are arranged to move a metal comprising object from the stream of metal-comprising objects into a desired output fraction. For each output fraction, the sorting unit 5 may be coupled to a bin, a container, or a chute, to collect the metal-comprising objects in the output fraction.

[0037] The one or more robotic arms may include means for applying suction to a metal-comprising object or mechanical gripping means. Thereby each robotic arm can grip a metal-comprising object and move the object to a desired output fraction. The one or more pushers may include a mechanical, pneumatic, spring-loaded and I or hydraulic pushers. Thereby each pusher can push and move a metal-comprising object to a desired output fraction. The one or more nozzles may be arranged to eject a pressurised gas to move a metal-comprising object into a desired fraction. Each nozzle includes a valve to open or close the nozzle and one or multiple nozzles may be activated simultaneously, depending, for instance, on the mass of the metalcomprising object. In operation, the metal-comprising object is moved by the sorting means into one of the two or more output fractions and thereby exits the system 1 .

[0038] The processing unit 6 includes circuitry 6A. The circuitry 6A includes at least one processor, such as a central processing unit (CPU), a microcontroller, microprocessor, graphic-card, or field programmable gate array (FPGA). The circuitry 6A further includes a working memory, an I / O module, a network connection module and, optionally, at least one screen. The processor is arranged to execute program code stored in the memory, in order to carry out the measuring, locating, object detection, assigning, cleaning, drying, transporting, processing, and I or sorting operations described herein.

[0039] The processing unit may optionally be coupled to the drive unit 2A. In operation, the speed of the conveyor 2 can be adapted by the processing unit 6 sending control signals to the drive unit 2A. The speed of the conveyor may be set, based on the through-put of metal-comprising objects, or other operational criteria. The circuitry 6A is configured to receive a measured contamination level for a metal-comprising object from the at least one measuring unit 3. The circuitry 6A is further configured to locate at least one point on the surface of the metal-comprising object where the contamination level is below a threshold level and to transmit the at least one located point to the laser unit 4B, further detailed below. The processing unit 6 may be a local processing unit, in proximity to the other units of the system. Alternatively, the processing unit 6 may be a remote processing unit. The processing unit 6 may be coupled to the other units of the system by means of cable connections, and / or by wireless connections. Further aspects of the processing unit 6 and circuitry 6A are described hereinbelow.

[0040] With reference to fig. 1 B, the measuring unit 3, the positioning unit 4A and the laser unit 4B are next described. The measuring unit 3 includes at least one light source 3A and a sensor 3B. The light source 3A may include one or more LED’s, halogen lights, or light bulbs. The light source 3A is arranged to illuminate the area on the conveyor 2 that is scanned by the sensor 3B. Advantageously, the sensor may thereby obtain a clear and bright image of the metal-comprising object, thereby improving accuracy of contamination level detection. The sensor 3B may be configured to detect optical radiation in the wavelength interval 100 nm - 1000 nm, in the wavelength interval 400 nm - 1100 nm, and I or in the wavelength interval 1100 nm - 1900 nm. The sensor 3B may be adapted for detecting visible light, NIR light, IR light, UV-light, or a combination thereof. Preferably, the sensor 3B may include a visible light sensor, such as a RGB (red, green blue) sensor, a UV sensor, an IR sensor, and / or a NIR (near infrared) sensor. According to one example, a first sensor may be adapted for detecting UV light, or mainly UV light. A second sensor may be adapted for detecting visible light, or mainly visible light. A third sensor may be adapted for detecting NIR light. The sensor 3B is configured to take one or more digital images of a metal-comprising object. The light source 3A and sensor 3B are coupled to the processing unit 6. The measuring unit 3 is configured to send the one or more images of the metal-comprising object to the processing unit 6.

[0041] The processing unit 6 is configured to process the received one or more images and locate one or more points on the metal-comprising object where contamination levels are below a threshold value. The threshold value may, for instance, be at 99% contamination, at 90% contamination, at 75% contamination, or at 50% contamination. In the image, a point may be represented by a single pixel in the image, or by multiple clustered pixels and I or neighbouring pixels. A point may thus refer to a point or to an area on the metal-comprising object. Optionally, the processing unit 6 may be configured to perform a foreground I background segmentation of the one or more images before locating one or more points. Thereby pixels can be located which are associated with the metal-comprising object and not with the image background. Further optionally, the processing unit 6 may be arranged to perform a connected component analysis. Thereby it is established which pixels are connected and picture the object in the one or more images.

[0042] The processing unit 6 may further be configured to locate one or more points on the metal-comprising object by color, color gradient, texture, shape and I or thermal signature, based on the processed images. Color may for instance, include hue, saturation, and I or lightness (HSL). Alternatively, color may include hue, saturation, and I or value (HVL). The processing unit 6 may be arranged to compare the color, color gradient, texture, shape and I or thermal signature of the metal-comprising object with one or more reference values, to determine if a point has a contamination level below the threshold level. The reference values may be stored by the processing unit 6. Alternatively, the circuitry 6A of the processing unit 6 may include a machine learning protocol to locate points with contamination levels below a threshold. The machine learning may be trained on sample data, such as images of metal-comprising objects, images of regions of metal-comprising object and I or pixel-level data of metal-comprising objects. The training may include supervised learning or unsupervised learning.

[0043] Located points are communicated by the processing unit 6 to the laser unit 4B and optionally, to the positioning unit 4A.

[0044] With continued reference to fig. 1 B, the positioning unit 4A is arranged to receive information on the one or more located points on a metalcomprising object from the processing unit 6. The positioning unit 4A is further arranged to perform object detection on the metal-comprising object. The object detection results in position information of the one or more located points on a metal comprising object. The position information may include two-dimensional position information. The two-dimensional position information is preferably provided in a two-dimensional coordinate system which is parallel to, and inert with respect to, the conveyor surface. The position information is stored by the processing unit 6. Advantageously, by providing accurate position information spectroscopy (detailed below) can be performed at desired and correct locations on the metal comprising objects.

[0045] Object detection may include laser triangulation on one or more located points. Alternatively, or additionally, object detection may include analysis of images of the metal-comprising object. Image analysis may include analysis of colour images, UV images, NIR images, X-ray images, and I or spectral images. Image analysis may be performed by a machine learning algorithm, or by other automated software. For instance, machine learning may include deep learning and I or supervised learning algorithms. Colour image analysis and / or spectral image analysis may include image segmentation. In each case, image analysis results in two-dimensional position information, as described herein above for laser triangulation. The processing unit 6A may be arranged to perform triangulation calculations and I or image analysis based on data received from the positioning unit 4A.

[0046] Object detection may further include height detection. Height detection preferably includes detection of height of the one or more located points with respect to the conveyor surface. Therefore, height detection is preferably in the direction normal to the conveyor surface. Height detection may be based on time-of-flight techniques, structured light techniques, optical disparity techniques and I or output from the one or more height detectors (detailed below). Height detection allows objects with substantially different heights, or with complicated three-dimensional geometries, to be processed by the system, whereas such objects may lead to detection errors for planar object recognition techniques. Consequently, pre-filtering or pre-ordering of objects before feeding the metal-comprising objects to the system is not necessary, leading to a less complicated system. Furthermore, height detection largely prevents issues with metal comprising objects having the same color as the conveyor surface. Such issues may occur for two-dimensional object recognition techniques based on RGB images, which cannot distinguish metal comprising objects from the conveyor surface when both have the same color. By using height detection it can be confirmed whether a point is located on the conveyor surface, or on an object carried by the conveyor surface.

[0047] The detected height information is stored by the processing unit 6. The height information may be combined with the two-dimensional position information. Thereby, three-dimensional position information of a located point may be provided. Advantageously, providing three-dimensional position information reduces issues with partially, or completely, overlapping objects, which may hamper two-dimensional object recognition techniques, such as edge detection or color distributions. Singulating, or otherwise filtering, the feed of metal comprising objects to avoid overlapping objects is therefore not necessary. The three-dimensional position information is utilized for laser induced breakdown spectroscopy (LIBS), detailed below. Optionally, the three-dimensional position information may be utilized for Raman spectroscopy, detailed below. Advantageously, the spectroscopy may thereby be focused precisely on a located point, resulting in accurate spectral data.

[0048] In a further embodiment, the processing unit 6 may be configured to compute geometrical data from the laser triangulation and I or image analysis. The geometrical data may include silhouettes, contour lines, surface areas, surface normal, surface curvatures, center-of-gravity, roughness and I or surface texture of the metal-comprising object. Advantageously, providing geometrical data may further improve distinguishing of overlapping, or partially overlapping objects, handling of objects with complex shapes, and handling of objects of varying sizes. The processing unit 6 may be configured to combine the geometrical data with the one or more processed images from the measuring unit 3. Combining geometrical data and the one or more processed images allows the processing unit 6 to identify which located points and I or which metal-comprising objects are best candidates for performing spectroscopy (detailed below). For instance, if the surface normals of a metal-comprising object point away from the field of view of the spectrometer (detailed below), the acquired spectral data may be of low quality. Advantageously, from a batch of metal-comprising objects, only the most suitable object may be selected for spectroscopy. Additionally and I or alternatively, if a plurality of points has been located, the processing unit 6 may be arranged to select only one point and send the selected point to the positioning unit 4A.

[0049] The positioning unit 4A may include a light source and a scanner, for object detection. The light source may include a laser 4A’, arranged to emit at least one laser beam 4A* Alternatively, the light source may include one or more LED’s, halogen lights, UV lights, or the like. The scanner may include at least one camera 4A”. Alternatively, or additionally, the positioning unit 4A may include a structured light scanner, an X-ray scanner, a spectral imaging scanner, or an ultra-sound scanner. Further alternatively, the positioning unit 4A may include at least two cameras, to enable stereo-vision images and I or optical disparity measurements. Optionally, the positioning unit 4A may include one or more height detection units, such as an array of height sensors.

[0050] The light source and the scanner are coupled to the processing unit 6. The laser beam 4A* emitted by the laser 4A’, may be a line laser. The laser beam 4A* may cover the width of the conveyor 2. The light source and the scanner are preferably aimed at the same line, or the same area, on the conveyor surface. The light source and the scanner may each be arranged in a fixed orientation. Alternatively, the orientation of the light source and I or the scanner may be adjustable. The positioning unit 4A may, for instance include an adjustable mirror, or adjustable mirror array, to direct the light emitted by the light source, such as the laser beam 4A* to a desired line, or area, of the conveyor surface. The scanner may be mounted on a moveable ball head or otherwise be adjustable. The angles between, for instance, the laser beam 4A*, the field of view of the camera 4A”, and the conveyor 2, may be recorded by the processing unit 6. Thereby, the processing unit 6 may perform triangulation calculations.

[0051] Additionally, the positioning unit 4A may be arranged to perform height detection of the one or more located points. Height detection may be based on time of flight techniques, structured light techniques, optical disparity techniques, and / or output from height detectors.

[0052] The processing unit 6 is arranged to calculate the position of a located point based on the laser triangulation and / or image analysis, together with optional height detection. Advantageously, an accurate three-dimensional position of a located point can thereby be derived, facilitating spectroscopy at the right location and thereby achieving accurate spectroscopy and a correct object classification.

[0053] With continued reference to fig. 1 B, the laser unit 4B is arranged to receive the two-dimensional position and I or the three-dimensional position of one or more located points from the processing unit 6. The laser unit 4B is further arranged to perform spectroscopy on the one or more located points. The laser unit 4B may be a LIBS unit, as shown in fig. 1 B. Alternatively, the laser unit 4B may be a Raman spectroscopy unit, such as a surface enhanced Raman spectroscopy (SERS) unit. The laser unit 4B may include a scanner 4B’, a laser focusing unit 4B”, a laser 4B’”, and a spectrometer 4C. The scanner 4B’, laser focusing unit 4B”, laser 4B’”, and spectrometer 4C are coupled to the processing unit 6.

[0054] The LIBS unit may be arranged to locally form a plasma on the located point. Thereto, the laser 4B’” is configured to direct a laser beam 4B* at the located point, thereby locally forming a plasma on the surface of the metalcomprising object. The Raman spectroscopy unit may be arranged to form an optical signal comprising a plurality of Raman wavelengths on the located point. Thereto, the laser 4B’” is configured to direct a laser beam 4B* at the located point to form the optical signal, on the surface of the metal-comprising object. The laser 4B’” may, for instance, be a Nd:YAG solid-state laser, or any other suitable solid state laser, or gas laser. The laser 4B’” may emit a laser beam 4B* with a wavelength in the optical domain, the near-infrared domain, or the near ultraviolet domain. The laser beam 4B* may be directed at the located point by the scanner 4B’ and the laser focusing unit 4B”, based on instructions from the processing unit 6. The scanner 4B’ may include one or more adjustable mirrors and / or a galvanometer. The laser focusing unit 4B” is arranged to focus the laser beam 4B* to a minimum radius, or to a line or area, at the located point. Thereby a plasma, or an optical signal comprising a plurality of Raman wavelengths, may be locally formed on the metal-comprising object.

[0055] The spectrometer 4C is arranged to measure spectral data of the plasma, or of the optical signal comprising a plurality of Raman wavelengths. Thereto, the spectrometer 4C may measure emitted light by the locally formed plasma. Alternatively, the spectrometer 4C may detect the plurality of Raman wavelengths comprised in the optical signal. The spectrometer 4C may be configured to analyze optical radiation in the wavelength interval 100 nm - 1000 nm, in the wavelength interval 400 nm - 1100 nm, and I or in the wavelength interval 1100 nm - 1900 nm. The spectrometer 4C may be adapted for analyzing visible light, NIR light, IR light, UV-light, or a combination thereof. The spectrometer 4C may be arranged to scan multi- spectral data or hyper-spectral data. Multi-spectral data may include non- adjacent spectral bands. Hyper-spectral data may include adjacent spectral bands. The circuitry 6A of the processing unit 6 is arranged to receive spectral data from the spectrometer 4C, and to assign a class to the metalcomprising object based on the received spectral data.

[0056] Optionally, the processing unit 6 may be arranged to perform normalization of the received spectral data, including temperature compensation and I or dark I white calibration. Further optionally, the processing unit 6 may be arranged for pre-processing of the spectral data. Pre-processing may include spectral binning and I or smoothing. Advantageously, pre-processing may improve the quality of the spectral data spectrally or pixel wise. The processing unit 6 may be further arranged to perform data reduction of the spectral data. Data reduction may be performed by applying principal component analysis, by expert-driven data reduction, and / or by applying partial least squares methods. Expert-driven data reduction may, for instance, include selecting only relevant spectral bands.

[0057] The processing unit 6 may be further arranged to compare the spectral data to previously measured spectral data of test objects for which the metal or alloy composition is known in detail. Alternatively, the processing unit 6 may be configured to compare the spectral data to a database with known metal or alloy spectral characteristics. In each case the processing unit 6 is arranged to assign a class based on the comparison.

[0058] The system 1 may be arranged to output metal-comprising objects comprising a desired metal, or alloy. The circuitry 6A may therefore be arranged to assign one of a first class A, or a second class B, to a metalcomprising object. In operation, the first class A is assigned when the desired metal or alloy is present, and the second class B is assigned when the desired metal or alloy is absent. In a further embodiment, the system 1 may be arranged to output several different metals and I or alloys. The first class may then be divided in two or more subclasses. The circuitry 6A may be arranged to assign one of the subclasses when a desired metal, or alloy, is present, or the second class B, when no desired metal or alloy is present. Each subclass is thereby associated with a different output fraction.

[0059] Referring to fig. 1 A, the system 1 may further include an optional return conveyor 7, for returning metal-comprising objects from the sorting unit 5 to the input side T of the system. The return conveyor 7 is indicated schematically by the dotted arrow in fig. 1 . The return conveyor 7 may include one or more conveyor bands. The sorting unit 5 may be configured to divert metal-comprising objects with contamination levels above the threshold level to the return conveyor 7. In operation, these contaminated metal-comprising objects may then be transported back to the input side T by the return conveyor 7. Next, contaminated metal-comprising objects may be cleaned in an optional cleaning unit, detailed below, before re-entering the measuring unit 3.

[0060] The system may further include an optional cleaning unit 8, for cleaning contaminated metal-comprising objects transported by the conveyor 2. The cleaning unit 8 is preferably arranged along the conveyor 2, at the input side T. The cleaning unit 8 is coupled to the processing unit 6. The cleaning unit 8 may include means for sanding, spraying with liquids, mechanical cleaning, chemical cleaning, high-pressure water cleaning, laserbased cleaning, or any combination thereof. Means for mechanical cleaning may, for instance, include one or more brushes, for brushing the metalcomprising objects. The cleaning unit 8 may further include a by-pass, allowing non-contaminated metal-comprising objects transported on the conveyor 2 to by-pass the cleaning unit 8. In operation, the cleaning unit 8 may receive contaminated metal-comprising objects from the stream of metalcomprising objects and I or from the return conveyor 7. The cleaning unit 8 may receive a signal from the processing unit 6 to clean, or to stop cleaning, metal-comprising objects in the stream of metal-comprising objects.

[0061] The system may also include an optional drying unit 9, for drying metal-comprising objects emerging form the cleaning unit 8. The drying unit 9 is preferably placed between the cleaning unit 8 and the measuring unit 3. The drying unit 9 is coupled to the processing unit 6. The drying unit 9 may include means for blow-drying, heat-drying, or combinations thereof. The drying unit 9 may further include a by-pass, allowing metal-comprising objects transported by the conveyor 2 to by-pass the drying unit 9. In operation, the drying unit 9 receive a signal from the processing unit 6 to dry, or to stop drying, metal-comprising objects in the stream of metal-comprising objects.

[0062] The system may further include an optional detection unit 10 arranged at the return conveyor 7. The detection unit 10 is coupled to the processing unit 6. The detection unit 10 is arranged to send a signal to the processing unit 6 if the number of metal-comprising objects on the return conveyor 7 reaches a predetermined threshold value. The processing unit 6 is configured, upon receiving the signal from the detection unit 10, to instruct the sorting unit 5 to output metal-comprising objects to the second fraction B instead of to the return conveyor 7. Thereby, an overflow of metal-comprising objects on the conveyor 2 can advantageously be prevented.

[0063] According to a further embodiment, the system may further comprise an auxiliary measuring unit 11 , schematically show in fig. 3. The auxiliary measuring unit 11 is arranged along the conveyor 2, at the input side T thereof. The auxiliary measuring unit 11 is coupled to the processing unit 6. In operation, contamination levels of metal-comprising objects may be measured in the auxiliary measuring unit 11 . Metal-comprising objects with high contamination levels may optionally be cleaned in the cleaning unit 8 and further optionally dried in the drying unit 9, before entering the measuring unit 3. Metal-comprising objects with low contamination levels may, instead, bypass the cleaning unit 8 and drying unit 9, and move directly to the measuring unit 3. The auxiliary measuring unit 11 may be coupled to the cleaning unit 8, via the processing unit 6. In operation, contamination information, such as the level of contamination or the location of the contamination on the metalcomprising object, as measured by the auxiliary measuring unit 11 may be communicated to the cleaning unit 8. Thereby, a more efficient cleaning may be achieved, where cleaning can be directed to the correct location on the metal-comprising object or can be performed with the required intensity based on the contamination level.

[0064] The return conveyor 7 may connect the sorting unit 5 with the optional cleaning unit 8. Alternatively, the return conveyor 7 may connect the sorting unit 5 with the auxiliary measuring unit 11 . The configuration and functioning of the auxiliary measuring unit 11 corresponds to the measuring unit 3 as described hereinabove, including at least one light source and a camera.

[0065] A schematic flow chart of the inventive method for sorting a metalcomprising object in a stream of metal-comprising objects is shown in fig. 2. According to the method, an input stream I of metal-comprising objects is transported from the input side T to the output side 1”. At the input side T an unsorted stream of metal-comprising objects is provided. At the output side 1” the stream is split into two or more fractions, as detailed hereinbelow. The stream of metal-comprising objects passes through the measuring unit 3, the laser unit 4 and the sorting unit 5, detailed hereinabove. The stream of metalcomprising objects may also pass through the optional cleaning unit 8, the optional drying unit 9 and I or the optional auxiliary measuring unit 11 , each detailed hereinabove.

[0066] The input stream I of metal-comprising objects may include a mixed- waste recycling stream, with both metal-comprising objects and non-metallic objects. Alternatively, the input stream I of metal-comprising objects may include a pre-sorted waste stream, including mainly, or exclusively, metalcomprising objects. Further alternatively, the input stream I may comprise a waste stream originating directly from metal industries.

[0067] Each metal-comprising object may have a metal fraction, by weight and I or by volume, of at least 5%, preferably at least 10%, more preferably at least 20%, most preferably at least 50%. Advantageously, a higher metal fraction results in a more efficient recycling process. Metal-comprising objects may, for example, include composite objects, such as objects comprising metal particles embedded in a matrix material.

[0068] The metal comprised in each metal-comprising object may include one or more pure metals and I or one or more metal alloys. Pure metals may include aluminium, copper, chromium, iron, magnesium, molybdenum, nickel, silver, or titanium. Alloys may include aluminium alloys, brass, bronze, gun metal, nickel alloys, solder, steel, stainless steel, or titanium alloys.

[0069] The metal-comprising objects in the input stream I may be contaminated. Contamination includes surface contamination, and may include foreign-particles, or foreign-layers, covering a part or the entirety of the surface of a metal-comprising object. Foreign particles may include dirt, debris, dust, rust, and the like. Foreign layers may include one or more layers of paint, coatings, grease, or other materials or compositions. The contamination may partly or wholly cover the surface of the metal-comprising object. Thereby, contamination may adversely affect the classification and sorting process.

[0070] For one or more metal-comprising objects in the stream I, surface contamination levels are measured. Measuring of contamination levels is performed by the measuring unit 3 and I or by the optional auxiliary measuring unit 11 , as detailed hereinabove. Measuring of contamination levels is performed by a taking one or more images of the metal-comprising object. The one or more images may, for instance, be visual light images, such as RGB images, UV images, or a NIR images. The one or more images are processed, to locate one or more points on the surface of each metalcomprising object where contamination levels are below a threshold level. Processing is performed by the processing unit 6, as detailed hereinabove.

[0071] Optionally, object detection may be performed by the positioning unit 4A, as described hereinabove. Object detection may include laser triangulation, on one or more located points. Alternatively, or additionally, object detection may include analysis of images of the metal-comprising object. Image analysis may include analysis of colour images, UV images, NIR images, X-ray images, and I or spectral images. Image analysis may be performed by automated software, including deep learning and / or supervised machine learning algorithms. Image analysis may include image segmentation, such as foreground / background segmentation.

[0072] Object detection results in position information. The position information is preferably provided in a two-dimensional coordinate system which is parallel to, and inert with respect to, the conveyor surface.

[0073] Object detection may further include height detection. Height detection results in height information above the conveyor surface in the surface normal direction, as detailed above. Thereby, objects with substantially different heights, or with complicated shapes, can be processed. Such objects may cause detection errors for purely two-dimensional object recognition techniques. Pre-filtering or pre-ordering of objects is therefore not necessary for the present method. Furthermore, height detection allows objects having the same color as the conveyor surface to be detected. For two-dimensional object recognition techniques, e.g. RGB image based techniques, distinguishing objects from the conveyor surface when both have the same color may not be possible with sufficient confidence levels.

[0074] The height detection may be based on time-of-flight techniques, structured light techniques, optical disparity techniques and I or output from the one or more height detectors. By combining height information with two- dimensional position information, three-dimensional position information may be derived. Three-dimensional position information is especially advantageous when applying LIBS. Thereby the LIBS laser can be focused precisely on a located point. Three-dimensional position information may also be utilized to apply Raman spectroscopy, directing the laser to a located point. Furthermore, three-dimensional position information may avoid issues with partially, or completely, overlapping objects. Such overlapping objects may be difficult to distinguish with two-dimensional techniques, such as edge detection or color distributions. Singulating, or otherwise filtering, the feed of metal comprising objects to avoid overlapping objects is therefore not necessary for the current method.

[0075] In a further embodiment, geometrical data may be computed from the data resulting from laser triangulation and I or image analysis. The geometrical data may include a silhouettes, contour lines, surface areas, surface normal, surface curvatures, center-of-gravity, roughness and I or surface texture of the metal-comprising object. Geometrical data of an object may be combined with the one or more processed images resulting from the contamination detection for the same object. The combination of geometrical data and processed image data may allow to determine suitability of the located point and I or metal-comprising object for spectroscopy. For instance, from a batch of metal-comprising objects, only the most suitable object may be selected for spectroscopy.

[0076] A laser beam is directed at the located point to locally form a plasma. Alternatively, the laser beam is directed at the located point to form an optical signal comprising a plurality of Raman wavelengths. The laser beam is emitted by the laser 4B’” of the laser unit 4B, as described hereinabove. The laser beam may have a wavelength in the optically visible domain, the NIR domain, or the near-UV domain. The light emitted by the plasma is detected by the spectrometer 4C. Alternatively, the optical signal comprising a plurality of Raman wavelengths is detected by the spectrometer 4C.

[0077] Next, a spectrum is measured of the plasma. Alternatively, a spectrum is measured of the Raman wavelengths comprised in the optical signal. The spectrum is measured by the spectrometer 4C. The measured spectrum may comprise multi-spectral data, or hyper-spectral data. Multi-spectral data may include non-adjacent spectral bands. Hyper-spectral data may include adjacent spectral bands. The spectrum is communicated to the processing unit 6.

[0078] In a further embodiment, a spectrum may be measured on several located points. Consequently, the one or more located points with the most suitable spectral data may be selected. The most suitable spectral data, may, e.g., include the spectral data with least noise, with clearest spectral metal or alloy signature, or the like. Thereby, the present method may be optimized and a more reliable sorting may be achieved.

[0079] Next, a class is assigned to the metal-comprising object, based on the measured spectrum. The class is assigned by the processing unit 6, as described hereinabove. The spectral data of the measured spectrum may be normalized, pre-processed, and I or reduced, before a class is assigned. Normalization may include temperature compensation and I or dark I white calibration. Pre-processing may include spectral binning and I or smoothing. Advantageously, pre-processing may improve the quality of the spectral data spectrally, or pixel wise. Reduction of spectral data may include applying principal component analysis, applying expert-driven data reduction, and I or applying a partial least squares methods. Expert-driven data reduction may, for instance, include selecting only relevant spectral bands.

[0080] Spectral data may be compared to previously measured spectral data for test objects of known metal or alloy composition. Alternatively, spectral data may be compared to a database comprising known spectral characteristics of different metals and I or alloys. Based on the comparison, a class may be assigned to the metal-comprising object. A first class may be assigned when a desired metal, or alloy, is present. A second class may be assigned when the desired metal, or alloy, is absent. Classification may be performed by a machine learning algorithm. The machine learning algorithm may be trained to classify metal-comprising objects by means of supervised learning, unsupervised learning, or a hybrid learning method.

[0081] Alternatively, the first class may be divided in two or more subclasses. Thereby, objects may be classified into two or more classes with different metals and I or alloys. One of the subclasses may then be assigned when a first desired metal, or alloy, is present, and the second class may be assigned when a second desired metal, or alloy, is present, where the second desired metal or alloy is different from the first desired metal or alloy. Each subclass is thereby associated with a different output metal or alloy.

[0082] The method further includes sorting metal-comprising objects into one of two or more output fractions, based on the assigned class. Sorting is performed by the sorting unit 5. As shown in fig. 2, the two or more output fractions may, for instance, include a first output fraction A, and a second output fraction B. The first output fraction A may comprise metal-comprising objects that can be recycled and, e.g., comprise a sufficient alloy content. The second output fraction B may comprise metal-comprising objects that cannot be recycled and, e.g., comprise an insufficient alloy content. Metal-comprising objects in output fraction B may, for instance, be discarded. Metal-comprising objects in output fraction A may, for instance, be used as a raw material for further industrial processing.

[0083] Alternatively, one or more metal-comprising objects may be added to a return fraction C. Metal-comprising objects in the return fraction C may be too contaminated to be sorted into an output fraction. The return fraction C may be returned to the input stream I. Thereby, the return fraction C may undergo a new cycle of optional cleaning, drying and contamination detection. Advantageously, it is thereby ensured that the output fractions comprise metal-comprising objects that are clean enough to be classified and sorted. With reference to fig. 3, according to a further embodiment, contamination levels of the metal-comprising objects may be detected before entering a cleaning and I or drying cycle. Thereto, the stream of metalcomprising objects may pass through an optional auxiliary measuring unit 11 , as described hereinabove.

[0084] Advantageous to the present invention, the cleaning process can be applied selectively on contaminated parts, leading to energy efficiency and cost effectiveness. Drying is optional, leading to further energy efficiency. The laser may be selectively routed, providing increased reliability.

[0085] Adaptive laser power control may be based on contamination sensor data, e.g., contamination or pre-classification data.

[0086] The inventive concept may further include automatic transfer learning. Such learning could be performed by an Al model. Models could be shared across many locations as long as the sorted metal-comprising objects look similar.

[0087] It is to be noted that the steps in the steps or acts of the above method may be conducted in any suitable order and hence not just in the order given above. Further, one or more of the steps or acts may be conducted in parallel. It is also to be noted that the steps or acts may be conducted by different equipment, at different times and / or at different sites. In other words, as an example, the method may be performed in a distributed manner at a plurality sites where different steps or acts are conducted at different points in time. However, the method may to advantage be conducted in the sequence described above at a single site.

[0088] Additionally, variations to the disclosed variants can be understood and effected by the skilled person in practicing the claimed invention, from a study of the drawings, the disclosure, and the appended claims. In the claims, the word “comprising” does not exclude other elements, and the indefinite article “a” or “an” does not exclude a plurality. The mere fact that certain measures are recited in mutually different dependent claims does not indicate that a combination of these measured cannot be used to advantage. Reference numbers

[0089] 1 system

[0090] T input side

[0091] 1” output side

[0092] 2 conveyor

[0093] 2A drive unit

[0094] 3 measuring unit

[0095] 3A light source

[0096] 3B sensor

[0097] 4A positioning unit

[0098] 4A laser

[0099] 4A* laser beam

[0100] 4A” camera

[0101] 4B laser unit

[0102] 4B’ scanner

[0103] 4B* laser beam

[0104] 4B” focus unit

[0105] 4B’” laser

[0106] 4C spectrometer

[0107] 5 sorting unit

[0108] 6 processing unit

[0109] 6A circuitry

[0110] 7 return conveyor

[0111] 8 cleaning unit

[0112] 9 drying unit

[0113] 10 detection unit 11 auxiliary measuring unit

[0114] A first output fraction

[0115] B second output fraction

[0116] C return fraction

[0117] I Input

Claims

CLAIMS1 . A method for sorting a metal-comprising object in a stream of metalcomprising objects, the method comprising:- measuring contamination levels over a surface of the metalcomprising object;- locating at least one point on the surface where the contamination level is below a threshold level; o directing a laser beam at the located point to locally form a plasma and measuring a spectrum of the plasma; or o directing a laser beam at the located point to form an optical signal comprising a plurality of Raman wavelengths and measuring a spectrum of the Raman wavelengths; the method further comprising:- assigning a class to the metal-comprising object, based on the measured spectrum; and- sorting the metal-comprising object into one of a first fraction (A), or a second fraction (B), based on the assigned class.

2. The method of any of claim 1 , wherein contamination levels are measured by a visible light sensor, a UV sensor, and / or a NIR sensor.

3. The method of claim 1 or 2, wherein measuring the spectrum of the plasma is performed by LIBS, or where measuring the spectrum of the plurality of the Raman wavelengths is performed by Raman spectroscopy, such as SERS.

4. The method of any of claims 1 - 3, wherein locating at least one point further comprises object detection by laser-triangulation and I or image analysis.

5. The method of claim 4, wherein object detection further includes height detection, based on time of flight techniques, structured light techniques, optical disparity techniques and I or output from height detectors.

6. The method of any one of claims 1 - 6, further comprising cleaning the object and optionally, drying the object, before measuring contamination levels.

7. The method of claim 6, further comprising re-cleaning the object if no point can be located with a contamination level below the threshold level.

8. The method of claim 6 or 7, wherein cleaning and I or re-cleaning comprises sanding, spraying with liquids, mechanical cleaning, chemical cleaning, high-pressure water cleaning, laser-based cleaning, or any combination thereof.

9. The method of any of claims 1 - 8, further comprising, if a plurality points on the surface have a contamination level below the threshold level, selecting one point from the plurality of points and directing the laser beam to the selected point.

10. The method of any of clais 1 - 9, wherein directing the laser beam is performed by manipulating a movably arranged mirror reflecting the laser beam.11 . The method of any of claims 1 - 10, wherein assigning a class is performed by a machine learning algorithm.

12. The method of any of claims 1 - 11 , wherein sorting of the metalcomprising object into one of a first fraction (A) or a second fraction (B) is based on one of metal characteristics of the metal-comprising object, or alloy composition of the metal-comprising object.

13. A system (1 ) for sorting a metal-comprising object in a stream of metalcomprising objects, the system comprising: a measuring unit (3) for measuring contamination levels over a surface of the metal-comprising object; a laser unit (4B) comprising a laser direction unit (4B’) and a spectrometer (C); and a processing unit (6) comprising circuitry (6A) configured to- receive measured contamination levels from the at least one measuring unit (3);- locate at least one point on the surface where contamination levels are below a threshold level; and- transmit the located point to the laser direction unit (4B’); wherein the laser direction unit (4A) is configured to form a plasma, or an optical signal comprising a plurality of Raman wavelengths, at the located point; wherein the spectrometer (4C) is configured to measure a spectrum of the plasma, or of the plurality of Raman wavelengths; wherein the circuitry (6A) is further configured to- receive the measured spectrum from the spectrometer (4C); and to- assign a class to the metal-comprising object based on the received spectrum;and wherein the system (1 ) further comprises at least one sorting unit (5) for sorting the metal-comprising object into one a first fraction (A), or a second fraction (B), based on the assigned class.

14. The system of claim 13, wherein the measuring unit (3) comprises a sensor (3B), such as a visible light sensor, a UV sensor, and / or a NIR sensor.

15. The system of claim 13 or 14, wherein the laser unit (4B) is a LIBS unit, or a Raman spectroscopy unit, such as a SERS unit.

16. The system of any one of claims 13 - 15, further comprising a cleaning unit (8) and optionally, a drying unit (9).

17. The system of claim 16, wherein the cleaning unit (8) is one of a sanding unit, a liquid spraying unit, a mechanical cleaning unit, a chemical cleaning unit, a high-pressure water cleaning unit, a laserbased cleaning unit, or any combination thereof.

18. The system of any of claims 13 - 17, wherein the circuitry (6A) is further configured, if a plurality of points on the surface have a contamination level below the threshold level, to select one point from the plurality of points and to direct the laser beam to the selected point.

19. The system of any of claims 13 - 18, wherein the laser direction unit (4A) comprises at least one movably arranged mirror.

20. The system of any of claims 13 - 19, wherein the circuitry (6A) is further configured to assign a class by means of a machine learningalgorithm.21 . The system of any of claims 13 - 20, wherein the sorting unit (5) is configured to sort the metal-comprising object based on one of metal characteristics of the metal-comprising object, or alloy composition of the metal-comprising object.

Images