Online detection method for chlorine content of industrial solid waste alternative fuel

Abstract

The invention discloses an on-line chlorine content detection method for industrial solid waste alternative fuel, and relates to the technical field of solid waste recovery, and the method comprises the following steps: arranging a near infrared spectrometer, and scanning an incoming material by using the near infrared spectrometer to obtain a near infrared spectrum; preprocessing the near infrared spectrum to obtain spectral characteristics; carrying out dimensionality reduction on the spectral features to obtain dimensionality-reduced spectral features; and constructing a deep residual neural network based on the dimensionality-reduced spectral features, and training the deep residual neural network by using structural parameters to obtain an output discrimination result of the chlorine content. According to the invention, a plurality of spot scanning probes are used for scanning incoming materials in a mutually parallel reciprocating motion manner, so that the sampling representativeness is improved, the small change of a sample is captured, and the chlorine content detection accuracy is improved; a two-layer or three-layer convolutional neural network structure is adopted, and a main path and a short cut path are superposed, so that the learning ability and performance of the deep residual neural network are improved.

Description

Technical Field

[0001] The invention relates to the technical field of solid waste recycling, in particular to an online detection method for the chlorine content of industrial solid waste alternative fuels. Background Art

[0002] As an emerging waste-to-energy technology, RDF has shown great potential in alleviating global warming and solving waste disposal problems. However, the high chlorine content in RDF has always been a key factor restricting its widespread application, mainly in terms of boiler corrosion and harmful gas emissions. At present, there are still many technical difficulties and limitations in the identification and control of chlorine concentration in RDF. Traditional chlorine content detection methods, such as chemical titration and ion chromatography, often require complex sample pretreatment processes, which are not only time-consuming and labor-intensive, but also difficult to achieve online real-time monitoring. X-ray Although fluorescence spectroscopy can be used for rapid detection, the equipment cost is high and there are radiation safety risks. Although thermogravimetric analysis has high accuracy, the analysis cycle is long and it is difficult to meet the rapid detection needs in industrial production. In addition, most of these methods are destructive tests and cannot guarantee the integrity of the samples. In terms of spectroscopy applications, although near-infrared spectroscopy technology has been proven to have the advantages of non-destructive, rapid and economical in many fields, its application research in the detection of chlorine content in solid waste alternative fuel RDF is still insufficient. Existing near-infrared spectroscopy analysis mostly focuses on the use of line spectroscopy technology, which has problems such as large signal transmission loss and poor stability.

[0003] At the same time, the current common solutions have many shortcomings. On the one hand, traditional spectral data processing methods, such as partial least squares (PLS) and principal component analysis (PCA), often find it difficult to effectively extract and utilize spectral features when faced with complex solid waste alternative fuel (RDF) samples, resulting in low model prediction accuracy. On the other hand, the existing solid waste alternative fuel (RDF) chlorine content identification models are mostly based on simple machine learning algorithms, which perform poorly when processing high-dimensional and nonlinear spectral data, especially when the sample size is limited. The generalization ability and robustness of these models often cannot meet the needs of actual applications. Therefore, the existing technology still has significant deficiencies in the rapid, accurate and non-destructive detection of the chlorine content of solid waste alternative fuel (RDF). Summary of the invention

[0004] In view of the fact that the prior art still has significant deficiencies in the rapid, accurate and non-destructive detection of the chlorine content of solid waste alternative fuel (RDF), and the problems of large signal transmission loss and poor stability in line spectrum technology, the present invention is proposed.

[0005] Therefore, the problem to be solved by the present invention is how to provide a method for efficiently identifying the chlorine content in waste plastics and classifying them, collecting spectral characteristic information of solid waste using point light sources instead of line light sources, and realizing complete signal transmission.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] In the first aspect, an embodiment of the present invention provides an online detection method for the chlorine content of industrial solid waste alternative fuel, which includes arranging a near-infrared spectrometer and using the near-infrared spectrometer to scan the incoming material to obtain a near-infrared spectrum; preprocessing the near-infrared spectrum to obtain spectral characteristics; reducing the dimension of the spectral characteristics to obtain reduced-dimensional spectral characteristics; constructing a deep residual neural network based on the reduced-dimensional spectral characteristics, and training the deep residual neural network using structural parameters to obtain an output discrimination result of the chlorine content.

[0008] As a preferred solution of the online detection method of chlorine content of industrial solid waste alternative fuel described in the present invention, the near-infrared spectrometer is connected to the point-shaped infrared signal receiver at the end with an optical fiber, the point-shaped infrared signal receiver is fixed on a slide, and a lighting lamp is also fixed on the slide to provide a background light source. The lighting lamp is a high-power halogen lamp, the position is parallel to the receiver, the direction of the lighting lamp is in the same direction as the signal collection point of the receiver, the slide is controlled by a motor, and can move back and forth on the guide rail, and the entire guide rail is horizontally mounted above the conveyor belt in the form of a gantry; the scanning refers to installing A near-infrared spectrometers in parallel and adjacently above the conveyor belt, and scanning the incoming materials by using point scanning probes to reciprocate on their respective tracks in the form of the same frequency and equidistant phase to obtain a near-infrared spectrum.

[0009] As a preferred solution of the online detection method for chlorine content of industrial solid waste alternative fuel described in the present invention, the preprocessing includes band selection, abnormal data screening, SG filtering smoothing and SNV normalization transformation; the band selection means that after the near-infrared spectrometer receives the near-infrared spectrum data, a number of data points are sampled equidistantly in the A1-A2 band to obtain a near-infrared spectrum of a specific band; the abnormal data screening includes: if the intensity of a certain point of the near-infrared spectrum is ≥ a preset maximum value, the corresponding near-infrared spectrum is determined to be invalid data, and the corresponding near-infrared spectrum is excluded from the data; if the intensity of all points of the near-infrared spectrum is < a preset maximum value and the intensity of a certain point is ≥ a preset minimum value, the corresponding near-infrared spectrum is determined to be valid data; if the intensity of all points of the near-infrared spectrum is < a preset minimum value, the corresponding near-infrared spectrum is determined to be invalid data, and the corresponding near-infrared spectrum is excluded from the data.

[0010] As a preferred solution of the online detection method of chlorine content of industrial solid waste alternative fuels described in the present invention, the SG filtering smoothing refers to using the python function signal.savgol_filter to output the original data after multiple polynomial smoothing of several data point windows, wherein the near infrared spectrum x i The value y to be fitted i Take point B on each side and calculate to get the best parameters of the quadratic fitting polynomial, then move the window to the next near-infrared spectrum x i+1 , the specific formula is as follows:

[0011]

[0012] Among them, x i is the near infrared spectrum; i is the near infrared spectrum x i The value to be fitted; a i is the optimal parameter of the quadratic fitting polynomial; b i is the error term for different data points; the optimal parameter a of the quadratic fitting polynomial is calculated by using the least squares method i , and then obtain the near-infrared spectrum x i The value y to be fitted i , the specific formula is as follows:

[0013]

[0014] in, is the fitting value of the predicted near-infrared spectrum Y; X is the design matrix; X T is the transposed matrix of the design matrix X; -1 is the inverse matrix of the design matrix X; Y is the actual near-infrared spectrum x i The value y to be fitted i A vector of observations.

[0015] As a preferred solution of the online detection method of chlorine content of industrial solid waste alternative fuels described in the present invention, the SNV standardization transformation refers to calculating the mean and variance of a spectrum and performing standardized shaping, and scaling the spectrum signal using the function package preprocessing.scale. The specific formula is as follows:

[0016]

[0017] Among them, x i is the near-infrared spectrum; x is the standardized spectral data; σ x is the standard deviation; n is the number of spectra; x j is the jth near-infrared spectrum.

[0018] As a preferred solution of the online detection method for chlorine content of industrial solid waste alternative fuels described in the present invention, the structural parameters of the deep residual neural network are as follows in a serial order: the reduced-dimensional spectral feature is used as input for sigmoidization and passes through Q1 one-dimensional convolutional neural network layers, wherein the convolution kernel size is P1 units, the stride parameter is P2, the padding parameter is P2, the input is S2 channels, and the output is S3 channels; after Q2 residual neural network layers of the same size, the input is S3 channels, and the output is S3 channels. channel; passes through Q1 residual neural network layers of different sizes, where the input is S3 channels and the output is S4 channels; passes through Q1 residual neural network layers of different sizes, where the input is S4 channels and the output is S5 channels; passes through Q1 residual neural network layers of different sizes, where the input is S5 channels and the output is S6 channels; after sampling through Q1 average pooling layers, the output is reduced in dimension to obtain one-dimensional data; passes through Q1 linear fully connected layers, where the input is S6 dimensions and the output is S7 dimensions, and the output is judged, and the largest output is taken as the judgment result.

[0019] As a preferred embodiment of the online detection method for chlorine content of industrial solid waste alternative fuels described in the present invention, the structure of the deep residual neural network is composed of a two-layer convolutional neural network or a three-layer convolutional neural network; for deep residual neural networks of the same size, the structure of the deep residual neural network is a main path and a shortcut path in parallel, and in the last step, the main path and the shortcut path are superimposed to obtain a superimposed main path and a superimposed shortcut path; the order of the superimposed main path is as follows: Q1 convolutional neural network layers, wherein the convolution kernel size is P1 units, the stride parameter is P3, and the padding The parameter is P2, the input is n channels, and the output is m channels; Q1 batchNorm layers; Q1 sigmoid mapping layers; Q1 convolutional neural network layers, where the convolution kernel size is P1 units, the stride parameter is P2, the padding parameter is P2, the input is m channels, and the output is m channels; Q1 batchNorm layers; the order of the superimposed shortcut path is as follows: Q1 convolutional neural network layers, where the convolution kernel size is P1 units, the stride parameter is P3, the padding parameter is P2, the input is n channels, and the output is m channels; Q1 batchNorm layers.

[0020] In the second aspect, in order to further solve the safety problems existing in the recycling of solid waste, the embodiment of the present invention provides an online detection system for the chlorine content of industrial solid waste alternative fuel, which includes: a spectral acquisition module for installing a near-infrared spectrometer, and using a point scanning probe to scan the incoming material in a parallel and reciprocating form to obtain a near-infrared spectrum; a spectral processing module for performing band selection, abnormal data screening, SG filtering smoothing and SNV normalization transformation on the near-infrared spectrum to obtain spectral characteristics; a spectral dimension reduction module for reducing the dimension of the spectral characteristics using the principal component analysis PCA algorithm to obtain reduced spectral characteristics; a discrimination result module for inputting the reduced spectral characteristics into a deep residual neural network, and training the deep residual neural network using structural parameters to obtain an output discrimination result of the chlorine content.

[0021] In a third aspect, an embodiment of the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, any step of the online detection method for the chlorine content of an industrial solid waste alternative fuel as described in the first aspect of the present invention is implemented.

[0022] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, any step of the online detection method for the chlorine content of an industrial solid waste alternative fuel as described in the first aspect of the present invention is implemented.

[0023] The beneficial effects of the present invention are as follows: the present invention installs a plurality of infrared spectrum detectors above the conveyor belt, fixes a point infrared signal receiver and an illuminator on a slide, controls the reciprocating translation movement on the guide rail by a motor, and uses a plurality of point scanning probes to move back and forth in parallel with each other to scan the incoming materials and obtain spectral data, thereby realizing a comprehensive coverage scan of the incoming materials on the conveyor belt, improving the representativeness of the sampling, and being able to capture the slight changes of the sample, thereby improving the accuracy and reliability of the chlorine content detection; by performing band selection, abnormal data screening, SG filtering smoothing and SNV standardization transformation processing on the acquired near-infrared spectrum, the extraction of key information is ensured, invalid data is effectively removed, and the interference caused by environmental factors and instrument differences is reduced; by performing dimension reduction on the pre-processed spectral data through a principal component analysis PCA algorithm, and using a deep residual neural network for training and discrimination, it is helpful to extract spectral features and perform efficient data analysis; by adopting a two-layer or three-layer convolutional neural network structure, including a main path and a shortcut path, and superimposing them in the last step, the learning ability and performance of the deep residual neural network are improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. Among them:

[0025] Figure 1 This is a schematic diagram of the overall structure of the device for judging the quality of the recycled plastic colored spots in Example 1.

[0026] Figure 2 This is a schematic flow chart of an online detection method for chlorine content in industrial solid waste alternative fuels based on near infrared spectroscopy in Example 1.

[0027] Figure 3 Schematic diagram of the structure of the deep residual neural network in Example 1.

[0028] Figure 4 This is a schematic diagram of a sample sample of an online detection method for chlorine content in industrial solid waste alternative fuel based on near infrared spectroscopy in Example 2.

[0029] Figure 5 This is the original spectrum data diagram measured for the sample in Example 2.

[0030] Figure 6 This is a normalized spectral data diagram of the sample measured in Example 2.

[0031] Figure 7 This is a schematic diagram of the evolution curve of the industrial solid waste alternative fuel object based on near-infrared spectroscopy in Example 2 after repeated training for 6 times. DETAILED DESCRIPTION

[0032] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.

[0033] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0034] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0035] Example 1

[0036] Reference Figure 1 to Figure 3 , which is the first embodiment of the present invention, provides an online detection method for the chlorine content of industrial solid waste alternative fuel, comprising the following steps:

[0037] S1: Arrange a near infrared spectrometer, and use the near infrared spectrometer to scan the incoming material to obtain a near infrared spectrum.

[0038] Preferably, Figure 1 The figure shows the overall structure diagram of the device for judging the quality of recycled plastic miscellaneous spots. The near-infrared spectrometer is connected to the point-shaped infrared signal receiver at the end by an optical fiber. By adopting the NIR1700 infrared spectrometer and configuring the integration time of 50mm, the point-shaped infrared signal receiver is fixed on the slide. The model is Delta GRH5-GY-041, and the stroke is optional from 50 to 800mm. The current stroke is 1240mm. The slide is also fixed with a lighting lamp to provide a background light source. The lighting lamp is a high-power halogen lamp, which is parallel to the receiver. The direction of the lighting lamp is in the same direction as the signal collection point of the receiver. The slide is controlled by a motor and can move back and forth on the guide rail. The entire guide rail is horizontally mounted above the conveyor belt in the form of a gantry. In actual applications, 4 to 5 line scanning points of infrared spectra can be used to increase the scanning rate. The device includes a movable infrared camera 1①, a halogen lamp②, a movable infrared camera 2③, a robotic arm④, a conveyor belt⑤ and a collection frame⑥.

[0039] Furthermore, scanning refers to installing multiple near-infrared spectrometers in parallel and adjacently above the conveyor belt, and scanning the incoming materials by reciprocating the point scanning probes on their respective tracks with the same frequency and equidistant phase to obtain the near-infrared spectrum.

[0040] Preferably, compared with the prior art which usually uses a single spectrometer or a fixed multi-spectrometer, the present invention significantly improves the scanning rate and coverage by utilizing multiple near-infrared spectrometers installed in parallel and adjacent to each other, and by reciprocating the point scanning probes on their respective tracks with the same frequency and equidistant phase. It can not only comprehensively collect incoming material information, but also adapt to conveyor belts of different widths and wastes of different shapes, thereby improving the applicability and efficiency of the system and obtaining more comprehensive near-infrared spectral data.

[0041] S2: Preprocess the near-infrared spectrum to obtain spectral features.

[0042] Preferably, Figure 2 The figure shows the implementation process of the online detection method of chlorine content in industrial solid waste alternative fuels based on near-infrared spectroscopy. The preprocessing includes band selection, abnormal data screening, SG filtering and smoothing, and SNV standardization transformation.

[0043] Specifically, band selection means that after the near-infrared spectrometer receives the near-infrared spectrum data, it samples a number of data points equidistantly in the 1082.62-1578.09nm band to obtain a near-infrared spectrum in a specific band.

[0044] Furthermore, when obtaining near-infrared spectra in a specific band, the conveyor belt width is set to 120 cm, the probe speed is 5 m / s, and the conveyor belt running speed is 1 m / s. When the material passes through the point signal receiver via the conveyor belt, the molecules of the material will absorb near-infrared photons and produce characteristic vibrations and combined vibrations. These characteristic absorption peaks reflect the chemical bonds and structures of the molecules and will be captured by the infrared spectrometer. The characteristic vibrations include stretching vibrations and bending vibrations.

[0045] Furthermore, in actual experiments, since the reflectance spectrum of light-colored plastics is clearer, the peaks of the obtained spectral graph are more characteristic, while black plastics require more data sets to support algorithm judgment.

[0046] Specifically, for near-infrared spectra in specific bands, experiments were conducted by collecting a large amount of near-infrared spectral data and analyzing the experimental data. It was found that in the 1082.62-1578.09nm band, the characteristic absorption peak of RDF plastic is more significant, and the 1082.62-1578.09nm band covers the absorption range of almost all samples, which can provide richer chemical and structural information while ensuring the integrity of the spectral data.

[0047] Specifically, the abnormal data screening includes: if the intensity of a certain point of the near-infrared spectrum is ≥ a preset maximum value, the corresponding near-infrared spectrum is determined to be invalid data, and the corresponding near-infrared spectrum is excluded from the data.

[0048] If the intensity of all points of the near-infrared spectrum is less than the preset maximum value and the intensity of a certain point is greater than or equal to the preset minimum value, the corresponding near-infrared spectrum is determined to be valid data.

[0049] If the intensity of all points of the near-infrared spectrum is less than the preset minimum value, the corresponding near-infrared spectrum is determined to be invalid data and the corresponding near-infrared spectrum is excluded from the data.

[0050] Specifically, SG filtering smoothing refers to the use of the python function signal.savgol_filter to output the original data after multiple polynomial smoothing of several data point windows, where the near-infrared spectrum x i The value y to be fitted i Take 3 points on each side and calculate to get the best parameters a0, a1 and a2 of the quadratic fitting polynomial, and then move the window to the next near-infrared spectrum x i+1, where i = 0, the specific formula is as follows:

[0051]

[0052] Among them, x i is the near infrared spectrum; i is the near infrared spectrum x i The value to be fitted; a i is the optimal parameter of the quadratic fitting polynomial; b i is the error term for different data points.

[0053] Furthermore, the optimal parameter a of the quadratic fitting polynomial is calculated by using the least squares method. i , and then obtain the near-infrared spectrum x i The value y to be fitted i , the specific formula is as follows:

[0054]

[0055] in, is the fitting value of the predicted near-infrared spectrum Y; X is the design matrix, which contains the polynomial basis functions used for fitting, that is, all near-infrared spectra x i The value of and the corresponding polynomial expansion; X T is the transposed matrix of the design matrix X; -1 is the inverse matrix of the design matrix X; Y is the actual near-infrared spectrum x i The value y to be fitted i A vector of observations.

[0056] Specifically, SNV standardization transformation refers to calculating the mean and variance of a spectrum and performing standardized shaping, and scaling the spectrum signal to [0,1] using the function package preprocessing.scale. The specific formula is as follows:

[0057]

[0058] Among them, x i is the near infrared spectrum; is the standardized spectral data; σ x is the standard deviation; n is the number of spectra; x j is the jth near-infrared spectrum.

[0059] Preferably, by analyzing the characteristic absorption peaks of the incoming materials based on a large amount of experimental data, the absorption range that can cover the incoming materials is selected to provide rich chemical and structural information, and the near-infrared spectral data is further processed in combination with abnormal data screening, SG filtering smoothing and SNV normalization transformation operations, which effectively improves the data quality.

[0060] S3: Reduce the dimension of the spectral features to obtain reduced-dimensional spectral features.

[0061] Preferably, the dimension reduction of spectral features includes the following steps: based on the principal component analysis PCA algorithm, select the PCA class of sklearn.decomposition in Python to instantiate pca=PCA(n_components=num_of_compo,whiten=0), specify the number of principal components to be retained and whether to perform whitening, wherein the selected number of principal components n_components is an integer selected from [16,30] according to the actual debugging effect, and the whitening parameter whiten=0.

[0062] The PCA class is instantiated to reduce the dimension of the spectral features and obtain the reduced-dimensional spectral features.

[0063] Preferably, since near-infrared spectral data usually contains hundreds or even thousands of wavelength points, direct use will lead to high computational complexity and time-consuming problems. By using the principal component analysis (PCA) algorithm to reduce the dimension of the spectral features, the training time and computing resource requirements of the subsequent deep residual neural network are significantly reduced, the efficiency of the entire system is improved, a large amount of redundant information and highly correlated features are eliminated, and the data structure is simplified.

[0064] S4: A deep residual neural network is constructed based on the reduced-dimensional spectral features, and the deep residual neural network is trained using the structural parameters to obtain the output discrimination result of the chlorine content.

[0065] Preferably, Figure 3 Shown is a schematic diagram of the structure of a deep residual neural network, and the deep residual neural network is trained using structural parameters, where the structural parameters of the deep residual neural network are as follows in a serial order: the reduced spectral features are used as input for sigmoidization and pass through a one-dimensional convolutional neural network layer, where the convolution kernel size is 3 units, the stride parameter is 1, the padding parameter is 1, the input is 1 channel, and the output is 64 channels.

[0066] After 2 residual neural network layers of the same size, the input is 64 channels and the output is 64 channels.

[0067] After 1 residual neural network layer of different sizes, the input is 64 channels and the output is 128 channels.

[0068] After 1 residual neural network layer of different sizes, the input is 128 channels and the output is 256 channels.

[0069] After 1 residual neural network layer of different sizes, the input is 256 channels and the output is 512 channels.

[0070] After sampling through one average pooling layer, the output is reduced in dimension to obtain one-dimensional data.

[0071] After 1 linear fully connected layer, the input is 512 dimensions and the output is 2 dimensions. The output is judged and the largest output is taken as the judgment result.

[0072] Preferably, the structure of the deep residual neural network consists of a two-layer convolutional neural network or a three-layer convolutional neural network.

[0073] Furthermore, for a deep residual neural network of the same size, the structure of the deep residual neural network is a main path and a shortcut path in parallel, and in the last step the main path and the shortcut path are superimposed to obtain a superimposed main path and a superimposed shortcut path.

[0074] Specifically, the order of the main path is as follows: 1 convolutional neural network layer, where the convolution kernel size is 3 units, the stride parameter is 1, the padding parameter is 1, and the input and output are both n channels.

[0075] 1 batchNorm layer.

[0076] 1 sigmoid mapping layer.

[0077] 1 convolutional neural network layer with a kernel size of 3 units, a stride parameter of 1, a padding parameter of 1, and n channels for both input and output.

[0078] 1 batchNorm layer.

[0079] Specifically, the shortcut path does not perform any processing and outputs the input as it is.

[0080] Furthermore, the order of the main path after superposition is as follows: 1 convolutional neural network layer, where the convolution kernel size is 3 units, the stride parameter is 2, the padding parameter is 1, the input is n channels, and the output is m channels.

[0081] 1 batchNorm layer.

[0082] 1 sigmoid mapping layer.

[0083] 1 convolutional neural network layer with a kernel size of 3 units, a stride parameter of 1, a padding parameter of 1, an input of m channels, and an output of m channels.

[0084] 1 batchNorm layer.

[0085] Furthermore, the order of the superimposed shortcut pathway is as follows: 1 convolutional neural network layer, where the convolution kernel size is 3 units, the stride parameter is 2, the padding parameter is 1, the input is n channels, and the output is m channels;

[0086] 1 batchNorm layer.

[0087] Specifically, the output judgment result of the chlorine content includes:

[0088] All outputs of the deep residual neural network are judged. If the output P is the maximum value of all outputs, the chlorine content of the corresponding industrial solid waste alternative fuel is determined to be P.

[0089] Otherwise, the maximum value of other outputs is determined.

[0090] The chlorine content P of the industrial solid waste alternative fuel is judged. If the chlorine content P> the first threshold value, the corresponding industrial solid waste alternative fuel is determined to be waste with a high chlorine content. The data is transmitted to the sorting equipment to separate the waste with a high chlorine content, thereby achieving maximum resource utilization of solid waste.

[0091] If the chlorine content P is less than or equal to the first threshold value, the corresponding industrial solid waste alternative fuel is determined to be waste with a low chlorine content, and waste separation is not performed.

[0092] Specifically, with regard to the chlorine content of industrial solid waste alternative fuels, a large amount of chlorine content data of industrial solid waste alternative fuels is collected, and classification thresholds are set in combination with the resource utilization requirements of solid waste, so as to maximize the separation of corresponding high-chlorine content wastes. In subsequent operations, the classification thresholds are adjusted in combination with changes in actual needs and the accuracy of separation results, thereby obtaining the optimal first threshold.

[0093] Preferably, by designing residual layers of different sizes for solid waste chlorine content detection, the deep residual neural network can better adapt to the characteristics of spectral data, adaptively learn features of different scales, and improve the model's ability to process complex data; based on the output of the deep residual neural network, high-chlorine content waste is accurately identified, precise separation is achieved, resource utilization is maximized, and by transmitting data to sorting equipment, high-chlorine content waste is automatically separated, improving processing efficiency and reducing labor costs and errors.

[0094] The present embodiment also provides an online detection system for the chlorine content of industrial solid waste alternative fuels, including: a spectral acquisition module, used to install a near-infrared spectrometer, and use a point scanning probe to scan the incoming material in a parallel and reciprocating manner to obtain a near-infrared spectrum; a spectral processing module, used to perform band selection, abnormal data screening, SG filtering smoothing and SNV standardization transformation on the near-infrared spectrum to obtain spectral features; a spectral dimension reduction module, used to use a principal component analysis (PCA) algorithm to reduce the dimension of the spectral features to obtain reduced-dimensional spectral features; a discrimination result module, used to input the reduced-dimensional spectral features into a deep residual neural network, and use structural parameters to train the deep residual neural network to obtain an output discrimination result of the chlorine content.

[0095] This embodiment also provides a computer device, which is applicable to an online detection method for the chlorine content of an industrial solid waste alternative fuel, and includes: a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute computer executable instructions to implement an online detection method for the chlorine content of an industrial solid waste alternative fuel as proposed in the above embodiment.

[0096] The computer device may be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device may be a touch layer covering the display screen, or a key, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse, etc.

[0097] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, an online detection method for the chlorine content of an industrial solid waste alternative fuel as proposed in the above embodiment is implemented; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as a static random access memory (Static Random Access Memory, referred to as SRAM), an electrically erasable programmable read-only memory (Electrically Erasable Programmable Read-Only Memory, referred to as EEPROM), an erasable programmable read-only memory (Erasable Programmable Read Only Memory, referred to as EPROM), a programmable read-only memory (Programmable Red-Only Memory, referred to as PROM), a read-only memory (Read-Only Memory, referred to as ROM), a magnetic storage device, a flash memory, a disk or an optical disk.

[0098] In summary, the present invention installs multiple infrared spectrum detectors above the conveyor belt, fixes the point infrared signal receiver and the lighting lamp on the slide, controls the reciprocating translation movement on the guide rail by a motor, and uses multiple point scanning probes to move back and forth in parallel with each other to scan the incoming materials to obtain spectral data, thereby realizing a comprehensive coverage scan of the incoming materials on the conveyor belt, improving the representativeness of the sampling, and being able to capture the slight changes of the sample, thereby improving the accuracy and reliability of the chlorine content detection; by performing band selection, abnormal data screening, SG filtering smoothing and SNV standardization transformation processing on the acquired near-infrared spectrum, the extraction of key information is ensured, invalid data is effectively removed, and the interference caused by environmental factors and instrument differences is reduced; by subjecting the pre-processed spectral data to principal component analysis PCA algorithm for dimensionality reduction, and using deep residual neural network for training and discrimination, it is helpful to extract spectral features and perform efficient data analysis; by adopting a two-layer or three-layer convolutional neural network structure, including a main path and a shortcut path, and superimposing them in the last step, the learning ability and performance of the deep residual neural network are improved.

[0099] Example 2

[0100] Reference Figures 4 to 7 , which is the second embodiment of the present invention. This embodiment is different from the first embodiment in that, in order to verify its beneficial effects, operating data and related instructions of the present invention in an actual environment are provided.

[0101] In this example, a conveyor belt equipped with multiple point scanning probes is configured. Each probe is fixed on a slide that can move back and forth along the guide rail and arranged in parallel. A high-power halogen lamp is installed on each slide to provide a stable background light source to ensure the quality of spectral data collection. During the operation of the production line, the probe scans the solid waste passing by. The sample of solid waste is as follows: Figure 4 As shown in the figure, the spectral data of the object is collected by a near-infrared spectrometer, where the original spectral data is as follows Figure 5 shown.

[0102] against Figure 5 The original spectral data collected is subjected to band selection to collect spectral data of specific bands of incoming materials. The collected band range is 1082.62-1578.09nm to ensure that key spectral information related to chlorine content is covered. The spectral data of specific bands is then subjected to abnormal data screening, SG filtering smoothing and SNV standardization transformation to homogenize the data and reduce the impact of equipment variation. The normalized spectral data obtained is as follows: Figure 6 shown.

[0103] Based on the normalized spectral data, the principal component analysis (PCA) algorithm is used to reduce the dimension and extract the most important features. The reduced-dimensional data is then input into the pre-trained deep residual neural network (ResNet) model for further feature learning and classification. The input features are used to determine whether the incoming material contains a high chlorine content. The computer configuration parameters during the experiment are shown in Table 1. The data is transmitted to the sorting equipment based on the test results to separate the waste with a high chlorine content, thereby achieving the maximum resource utilization of solid waste. Figure 7 The figure shows a schematic diagram of the evolution curve of the industrial solid waste alternative fuel object repeatedly trained 6 times based on near-infrared spectroscopy, wherein the upper curve is the fitting curve of train, and the lower curve is the fitting curve of value. It can be seen that the present invention collects the spectral feature information of solid waste by using point light source instead of line light source, and uses principal component analysis PCA algorithm for dimensionality reduction, and deep residual neural network for training and discrimination, so as to efficiently identify and classify the chlorine content in the waste plastic, thereby improving the accuracy and reliability of chlorine content detection.

[0104] Table 1 Experimental computer configuration parameters

[0105] Configuration properties parameter system Windows 10 64 CPU Intel(R)Celeron(R)CPU G1840@2.80GHz Memory 8.00GB GPU NVIDIA RTX 1060Ti

[0106] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. An online detection method for chlorine content of industrial solid waste alternative fuels, characterized in that: include: Arranging a near infrared spectrometer, and using the near infrared spectrometer to scan the incoming material to obtain a near infrared spectrum; Preprocessing the near infrared spectrum to obtain spectral features; Performing dimension reduction on the spectral feature to obtain a reduced-dimensional spectral feature; A deep residual neural network is constructed based on the reduced-dimensional spectral features, and the deep residual neural network is trained using structural parameters to obtain an output discrimination result of the chlorine content.

2. The method for online detection of chlorine content in industrial solid waste alternative fuels according to claim 1, characterized in that: The near-infrared spectrometer is connected to a point-shaped infrared signal receiver at the end by an optical fiber, and the point-shaped infrared signal receiver is fixed on a slide, and a lighting lamp is also fixed on the slide. The lighting lamp is a high-power halogen lamp, and its position is parallel to the receiver. The direction of the lighting lamp is in the same direction as the signal collection point of the receiver. The slide is controlled by a motor and can move back and forth on the guide rail. The entire guide rail is horizontally mounted above the conveyor belt in the form of a gantry; The scanning refers to installing A near-infrared spectrometers in parallel and adjacently above the conveyor belt, and scanning the incoming materials by reciprocating the point scanning probes on their respective tracks in the form of the same frequency and equidistant phase to obtain a near-infrared spectrum.

3. The method for online detection of chlorine content in industrial solid waste alternative fuels according to claim 2, characterized in that: The preprocessing includes band selection, abnormal data screening, SG filtering and smoothing, and SNV standardization transformation; The band selection means that after the near-infrared spectrometer receives the near-infrared spectrum data, it samples a number of data points equidistantly in the A1-A2 band to obtain the near-infrared spectrum of a specific band; The abnormal data screening includes: If the intensity of a certain point of the near-infrared spectrum is ≥ the preset maximum value, the corresponding near-infrared spectrum is determined to be invalid data, and the corresponding near-infrared spectrum is excluded from the data; If the intensity of all points of the near-infrared spectrum is less than the preset maximum value and the intensity of a certain point is greater than or equal to the preset minimum value, the corresponding near-infrared spectrum is determined to be valid data; If the intensity of all points of the near-infrared spectrum is less than the preset minimum value, the corresponding near-infrared spectrum is determined to be invalid data and the corresponding near-infrared spectrum is excluded from the data.

4. The method for online detection of chlorine content in industrial solid waste alternative fuels according to claim 3, characterized in that: The SG filtering smoothing refers to using the python function signal.savgol_filter to output the original data after multiple polynomial smoothing of several data point windows, where the near infrared spectrum x i The value y to be fitted i Take point B on each side and calculate to get the best parameters of the quadratic fitting polynomial, then move the window to the next near-infrared spectrum x i+1 , the specific formula is as follows: Among them, x i is the near infrared spectrum; i is the near infrared spectrum x i The value to be fitted; a i is the optimal parameter of the quadratic fitting polynomial; b i is the error term for different data points; By using the least squares method to calculate the optimal parameters a of the quadratic fitting polynomial i , and then obtain the near-infrared spectrum x i The value y to be fitted i , the specific formula is as follows: in, is the fitting value of the predicted near-infrared spectrum Y; X is the design matrix; X T is the transposed matrix of the design matrix X; -1 is the inverse matrix of the design matrix X; Y is the actual near-infrared spectrum x i The value y to be fitted i A vector of observations.

5. The method for online detection of chlorine content in industrial solid waste alternative fuels according to claim 4, characterized in that: The SNV standardization transformation refers to calculating the mean and variance of a spectrum and performing standardized shaping, and scaling the spectral signal using the function package preprocessing.scale. The specific formula is as follows: Among them, x i is the near infrared spectrum; is the standardized spectral data; σ x is the standard deviation; n is the number of spectra; x j is the jth near-infrared spectrum.

6. The method for online detection of chlorine content in industrial solid waste alternative fuels according to claim 5, characterized in that: The structural parameters of the deep residual neural network are as follows in series order: The reduced spectral features are used as input for sigmoidization and passed through Q1 one-dimensional convolutional neural network layers, where the convolution kernel size is P1 units, the stride parameter is P2, the padding parameter is P2, the input is S2 channels, and the output is S3 channels; After Q2 residual neural network layers of the same size, the input is S3 channels and the output is S3 channels; After Q1 residual neural network layers of different sizes, the input is S3 channels and the output is S4 channels; After Q1 residual neural network layers of different sizes, the input is S4 channels and the output is S5 channels; After Q1 residual neural network layers of different sizes, the input is S5 channels and the output is S6 channels; After sampling through Q1 average pooling layers, the output is reduced in dimension to obtain one-dimensional data; After Q1 linear fully connected layers, the input is S6 dimension and the output is S7 dimension. The output is judged and the largest output is taken as the judgment result.

7. The method for online detection of chlorine content in industrial solid waste alternative fuels according to claim 6, characterized in that: The structure of the deep residual neural network is composed of a two-layer convolutional neural network or a three-layer convolutional neural network; For the deep residual neural network of the same size, the structure of the deep residual neural network is a main path and a shortcut path in parallel, and in the last step, the main path and the shortcut path are superimposed to obtain the superimposed main path and the superimposed shortcut path; The order of the main paths after superposition is as follows: Q1 convolutional neural network layers, where the convolution kernel size is P1 units and the stride parameter is P3, The padding parameter is P2, the input is n channels, and the output is m channels; Q1 batchNorm layers; Q1 sigmoid mapping layer; Q1 convolutional neural network layers, where the convolution kernel size is P1 units and the stride parameter is P2, The padding parameter is P2, the input is m channels, and the output is m channels; Q1 batchNorm layers; The order of the shortcut paths after superposition is as follows: Q1 convolutional neural network layers, where the convolution kernel size is P1 units and the stride parameter is P3, The padding parameter is P2, the input is n channels, and the output is m channels; Q1 batchNorm layers.

8. An online detection system for chlorine content of industrial solid waste alternative fuel, based on an online detection method for chlorine content of industrial solid waste alternative fuel according to any one of claims 1 to 7, characterized in that: include, The spectrum acquisition module is used to install a near-infrared spectrometer and use a point scanning probe to scan the incoming material in a parallel and reciprocating manner to obtain a near-infrared spectrum; The spectrum processing module is used to select the bands of the near-infrared spectrum, filter out abnormal data, perform SG filtering and smoothing, and perform SNV standardization transformation to obtain the spectrum characteristics; The spectral dimension reduction module is used to reduce the dimension of spectral features using the principal component analysis (PCA) algorithm to obtain reduced-dimensional spectral features; The discrimination result module is used to input the reduced-dimensional spectral features into the deep residual neural network, and use the structural parameters to train the deep residual neural network to obtain the output discrimination result of the chlorine content.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method for online detection of chlorine content in an industrial solid waste alternative fuel as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method for online detection of chlorine content in an industrial solid waste alternative fuel as claimed in any one of claims 1 to 7 are implemented.

Images