Edge-cloud collaboration platform for intelligent coking monitoring of cracking furnace tubes and working method thereof

Abstract

The edge-cloud collaboration platform for intelligent coking monitoring of cracking furnace tubes includes an edge layer and a cloud layer, which can store and analyze big data, propose suggestions on optimization and improvement, and feed the suggestions back to the edge layer. The edge layer includes an intelligent temperature measuring device for an outer surface of a cracking furnace tube and/or an ethylene DCS/data acquisition device; the cloud layer includes a cracking furnace safety warning device, an intelligent coking diagnosis and prediction device for a cracking furnace tube, a hybrid job scheduling device, a multi-workflow scheduling device, a virtualized resource scheduling device, and a virtual resource optimization device; and the intelligent temperature measuring device for an outer surface of a cracking furnace tube includes an identification device for furnace tube and overlapped tube, and an abnormal data detection device.

Claims

1. An edge-cloud collaboration platform for intelligent coking monitoring of cracking furnace tubes, comprising: an edge layer; and a cloud layer, wherein the edge layer includes an intelligent temperature measuring device for an outer surface of the cracking furnace tubes and/or an ethylene Distributed Control System (DCS)/data acquisition device; the intelligent temperature measuring device includes a furnace tube temperature metering device, a target temperature distance measuring device, and a position angle optimization device; the ethylene DCS/data acquisition device includes one or more from a furnace tube outlet temperature device, a venturi pressure device, and a crossover section pressure device; and the intelligent temperature measuring device for the outer surface of the cracking furnace tubes further includes an overlapped tube identification device for the furnace tubes, and an abnormal data detection device; the cloud layer includes a cracking furnace safety warning device, an intelligent coking diagnosis and prediction device for the cracking furnace tubes, a hybrid job scheduling device, a multi-workflow scheduling device, a virtualized resource scheduling device, and a virtual resource optimization device; and the overlapped tube identification device for the furnace tubes includes an intelligent temperature measuring instrument, a PC terminal, and a cloud server; wherein the intelligent temperature measuring instrument includes an internal integrated infrared temperature measuring module, a laser ranging module, and an embedded processor; and data obtained by the internal integrated infrared temperature measuring module and the laser ranging module are transmitted to the PC terminal and/or the cloud server after being processed by the embedded processor; the abnormal data detection device is configured to detect outliers, outliers are referred to data that deviates significantly from other data and does not follow general conduct or pattern of data tend to appear in a measured data set, which caused by data noise or occurrence of special abnormal behaviors and states in the data source.

2. A working method of the edge-cloud collaboration platform according to claim 1, comprising the steps of: S1. the edge layer acquiring data and performing preliminary processing and analysis; S2. the cloud layer receiving data and performing big data statistical analysis; and S3. the cloud layer proposing suggestions of optimizing and improving the edge layer; wherein the step S1 includes an overlapped tube identification method for the furnace tubes, the overlapped tube identification method including: S111. acquiring data; S112. constructing a Dynamic Convolution Neural Network (DCNN)-based overlapped tube identification model for furnace tubes; S113. reconstructing a DCNN-based overlapped tube identification model for the furnace tubes; S114. metering an outer surface temperature of the furnace tubes; and S115. uploading outer surface temperature of the furnace tubes to the cloud server, and wherein the step S1 includes an abnormal data detection method, and the abnormal data are outliers, the abnormal data detection method including: S121. collecting an outer surface temperature of the cracking furnace tubes as a target data set by the intelligent temperature measuring instrument; S122. training the target data set by using a Gaussian mixture model; S123. associating sample points in the training data with constituent points of the Gaussian mixture model according to training results; S124. screening the constituent points in a predetermined standard deviation calculation manner; and S125. acquiring the outliers in the sample points according to screening results.

3. The working method of the edge-cloud collaboration platform according to claim 2, wherein in step S111, the internal integrated infrared temperature measuring module and the laser ranging module of the intelligent temperature measuring instrument collects raw one-dimensional data of the outer surface temperature and a distance of the furnace tubes and an inner wall temperature and a distance of a furnace hearth; in step S112, the raw one-dimensional data acquired by the intelligent temperature measuring instrument is transmitted to the PC terminal, the PC terminal transforms the acquired raw one-dimensional data into a two-dimensional histogram; feature extraction is performed on the furnace tube distance data in the two-dimensional histogram to obtain a distance feature map of overlapped tube and non-overlapped tube of the furnace tubes so as to form a data set for a DCNN network model to train; and the data set is input into the DCNN network model of the PC terminal so as to train the data set and build a DCNN overlapped tube recognition model at the PC terminal; in step S113, the DCNN-based overlapped tube identification model for the furnace tubes after being trained by the PC terminal is transplanted to the embedded processor inside the intelligent temperature measuring instrument so as to reconstruct the DCNN overlapped tube recognition model within the embedded processor; in step S114, the intelligent temperature measuring instrument collects raw one-dimensional data of the outer surface temperature and distance of the furnace tube and the inner wall temperature and distance of the furnace hearth in real time, extracts furnace tube distance data in the raw one-dimensional data, and returns initial position coordinates of the furnace tube distance data in the raw one-dimensional data, determines whether the furnace tube corresponding to the extracted furnace tube distance data is an overlapped tube by using the DCNN overlapped tube identification model in the embedded processor, returns an identification type label for each furnace tube, and meters the outer surface temperature of the furnace tube of the non-overlapped tube and overlapped tube at corresponding starting position coordinates according to the starting position coordinates of the returned furnace tube identification type label and the furnace tube distance data; and in step S115, the intelligent temperature measuring instrument uploads the calculated outer surface temperature of the furnace tube to the cloud server.

4. The working method of the edge-cloud collaboration platform according to claim 3, wherein in step S112, during feature extraction of the furnace tube distance data in the two-dimensional histogram, the temperature data of the outer surface of the furnace tube and the inner wall of the furnace hearth in the two-dimensional histogram are removed first, the distance data of the inner wall of the furnace hearth in the two-dimensional histogram is removed, and then invalid threshold data with low relevance in the furnace tube distance data is removed to extract the distance feature map of overlapped tube and non-overlapped tube of the furnace tube.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

(1) FIG. 1 is an architecture diagram of an edge-cloud collaboration platform for intelligent coking monitoring of cracking furnace tubes according to an embodiment.

(2) FIG. 2 is a frame diagram showing a collaboration ability of the edge-cloud collaboration platform according to an embodiment.

DETAILED DESCRIPTION OF EMBODIMENTS

(3) The drawings of the present invention are for illustrative purpose only and are not to be construed as limiting the invention. Some components in the drawings may be omitted, enlarged, or reduced for better illustrating the following embodiments, and sizes of these components do not represent that of actual products. For those skilled in the art, it will be understood that some known structures and descriptions thereof in the drawings may be omitted.

(4) As shown in FIGS. 1 and 2, an edge-cloud collaboration platform for intelligent coking monitoring of the cracking furnace tubes according to an embodiment includes an edge layer and a cloud layer. The edge layer includes an intelligent temperature measuring device for an outer surface of a cracking furnace tube and/or an ethylene DCS/data acquisition device. The intelligent temperature measuring device for an outer surface of a cracking furnace tube includes a furnace tube temperature metering device, a target temperature distance measuring device, and a position angle optimization device. The ethylene DCS/data acquisition device includes one or more of a furnace tube outlet temperature device, a venturi pressure device, and a crossover section pressure device. The cloud layer includes a cracking furnace safety warning device, an intelligent coking diagnosis and prediction device for the cracking furnace tube, a hybrid job scheduling device, a multi-workflow scheduling device, a virtualized resource scheduling device, and a virtual resource optimization device. The intelligent temperature measuring device for the outer surface of the cracking furnace tube includes an overlapped tube identification device for the furnace tube, and an abnormal data detection device.

(5) As shown in a frame diagram in FIG. 2 showing a collaboration ability of the edge-cloud collaboration platform, the edge-cloud collaboration platform includes an edge layer and a cloud layer. The edge layer includes an intelligent temperature measuring device for an outer surface of a cracking furnace tube and/or an ethylene DCS/data acquisition device. The edge layer thus can perform cracking furnace data acquisition with a wide range and a deep level, perform protocol conversion and edge processing of heterogeneous data, and achieve data integration to a cloud platform. The intelligent temperature measuring device for the outer surface of the cracking furnace tube includes a furnace tube temperature metering device, a target temperature distance measuring device, a position angle optimization device; the ethylene DCS/data acquisition device includes one or more of a furnace tube outlet temperature device, a venturi pressure device, and a crossover section pressure device.

(6) The cloud layer uses enterprise private cloud as the infrastructure, and achieves the storage and intelligent analysis of big data in the ethylene industry combined with virtualized resource management and cloud job scheduling. The cloud layer includes a cracking furnace safety warning device, an intelligent coking diagnosis and prediction device for the cracking furnace tube, a hybrid job scheduling device, a multi-workflow scheduling device, a virtualized resource scheduling device, and a virtual resource optimization device. The intelligent temperature measuring device for the outer surface of the cracking furnace tube includes an overlapped tube identification device for the furnace tube, and an abnormal data detection device.

(7) The edge layer and the cloud layer achieve whole process optimization for lifecycle of a cracking furnace tube in a collaborative manner; the edge-cloud collaboration ability and connotation primarily include data collaboration and intelligence collaboration.

(8) Data collaboration: edge nodes acquire ethylene cracking furnace operating parameters, perform preliminary processing and analysis on the data in accordance with rules and data model, upload the processed data to the cloud, in which the data uploaded from the edge layer to the cloud platform in the present collaboration platform includes the outer surface temperature of the cracking furnace tube obtained by the intelligent temperature measuring device independently developed by a research group, and the cracking furnace tube outlet temperature, the venturi pressure, and the crossover section pressure that are acquired by the ethylene DC S/data acquisition device mounted by a petrochemical enterprise. The cloud constantly receives the data of the edge nodes, and performs big data statistical analysis based on that the mass operation state data includes abnormal state data, including data-driven furnace tube intelligent coking monitoring and scalable value-added services such as furnace tube coking, furnace tube maintenance, furnace tube replacement, separate optimization for furnace hearth and thermal field, and cracking furnace safety warning.

(9) Intelligence collaboration: On the one hand, an edge node provides data input for cloud cracking furnace intelligent coking prediction model training; on the other hand, the edge node is responsible for edge inference execution, i.e., executing a CNN-based detection algorithm for overlapped tube of the cracking furnace to realize accurate measurement of outer surface temperature of each furnace tube in a cracking furnace (in the present collaboration platform, an edge node, i.e., a non-contact intelligent temperature measuring instrument, which realizes efficient intelligent collaboration by deploying AI chips). The cloud refines relevant critical operating parameter models through large data analysis and abstraction, including outlier detection based on Gaussian mixture models, continuous optimization training and compression transplantation of CNN network parameters, and providing suggestions of optimizing and adjusting field devices and ensuring that the devices are in normal operation and production efficiency is improved.

(10) The overlapped tube identification device for furnace tube includes an intelligent temperature measuring instrument, a PC terminal, and a cloud server. The intelligent temperature measuring instrument includes an internal integrated infrared temperature measuring module, a laser ranging module, and an embedded processor. Data obtained by the internal integrated infrared temperature measuring module and the laser ranging module are transmitted to the PC terminal and/or the cloud server after being processed by the embedded processor. The internal integrated infrared temperature measuring module of the intelligent temperature measuring instrument is used for temperature measurement, the laser ranging module is used for ranging, and data obtained by the internal integrated infrared temperature measuring module and the laser ranging module are transmitted to the PC terminal and/or the cloud server after being processed by the embedded processor.

(11) The working method of the edge-cloud collaboration platform for intelligent coking monitoring of the cracking furnace tube, including the following steps: S1. the edge layer acquiring data and performing preliminary processing and analysis; S2. the cloud layer receiving data and performing big data statistical analysis; and S3. the cloud layer proposing suggestions of optimizing and improving the edge layer.

(12) The edge layer includes an intelligent temperature measuring device for the outer surface of the cracking furnace tube and the ethylene DCS/data acquisition device, the edge layer thus can perform cracking furnace data acquisition with a wide range and a deep level, perform protocol conversion and edge processing of heterogeneous data, and achieve data integration to a cloud platform. The cloud constantly receives the data of the edge nodes, and performs big data statistical analysis based on that the mass operation state data includes abnormal state data, including data-driven furnace tube intelligent coking monitoring and scalable value-added services such as furnace tube coking, furnace tube maintenance, furnace tube replacement, separate optimization for furnace hearth and thermal field, and cracking furnace safety warning. The cloud layer provides suggestions of optimizing and adjusting the field devices, and ensures that the devices are in normal operation and the production efficiency is improved.

(13) The step S1 includes an overlapped tube identification method for furnace tube, including

(14) S111. acquiring data;

(15) S112. constructing a DCNN-based overlapped tube identification model for the furnace tubes;

(16) S113. reconstructing a DCNN-based overlapped tube identification model for the furnace tubes;

(17) S114. metering the outer surface temperature of the furnace tubes; and

(18) S115. uploading the outer surface temperature of the furnace tubes to the cloud server.

(19) The step S1 includes an abnormal data detection method, in which the abnormal data are outliers, the method includes:

(20) S121. collecting outer wall temperature of the ethylene cracking furnace tube as a target data set by an intelligent temperature measuring instrument;

(21) S122. training the target data set by a Gaussian mixture model;

(22) S123. associating sample points in the training data with constituent points of the Gaussian mixture model according to training results;

(23) S124. screening the constituent points in a predetermined standard deviation calculation manner; and

(24) S125. acquiring the outliers in the sample points according to the screening results.

(25) Specifically, in step S111, the internal integrated infrared temperature measuring module and the laser ranging module of the intelligent temperature measuring instrument collects raw one-dimensional data of the outer surface temperature and distance of the furnace tube and the inner wall temperature and distance of the furnace hearth;

(26) in step S112, a DCNN-based overlapped tube identification model for furnace tube is constructed. The raw one-dimensional data acquired by the intelligent temperature measuring instrument is transmitted to the PC terminal, the PC terminal transforms the acquired raw one-dimensional data into a two-dimensional histogram; feature extraction is performed on the furnace tube distance data in the two-dimensional histogram to obtain a distance feature map of overlapped tube and non-overlapped tube of the furnace tube so as to form a data set for a DCNN network model to train; and the data set is input into the DCNN network model of the PC terminal so as to train the data set and build a DCNN overlapped tube recognition model at the PC terminal;

(27) in step S113, a DCNN-based overlapped tube identification model for the furnace tube is reconstructed. The DCNN-based overlapped tube identification model for the furnace tube after being trained by the PC terminal is transplanted to the embedded processor inside the intelligent temperature measuring instrument so as to reconstruct the DCNN overlapped tube recognition model within the embedded processor;

(28) in step S114, the outer surface temperature of the furnace tube is metered. The intelligent temperature measuring instrument collects raw one-dimensional data of the outer surface temperature and distance of the furnace tube and the inner wall temperature and distance of the furnace hearth in real time, extracts furnace tube distance data in the raw one-dimensional data, and returns initial position coordinates of the furnace tube distance data in the raw one-dimensional data, determines whether the furnace tube corresponding to the extracted furnace tube distance data is an overlapped tube by using the DCNN overlapped tube identification model in the embedded processor, returns an identification type label for each furnace tube, and meters the outer surface temperature of the furnace tube of the non-overlapped tube and overlapped tube at corresponding starting position coordinates according to the starting position coordinates of the returned furnace tube identification type label and the furnace tube distance data; and

(29) in step S115, the intelligent temperature measuring instrument uploads the calculated outer surface temperature of the furnace tube to the cloud server.

(30) It is preferable to use STM32F767VET6 having CortexM7 as the kernel as the embedded processor of the intelligent temperature measuring instrument, and deploy an optimized software kernel of a neural network on the embedded processor as CMSISNN, in which the CMSISNN kernel library includes functions such as deep convolution, pooling, activation, and full connection that are required for implementation of the DCNN network, and thus it is possible for us to reconstruct the DCNN overlapped tube recognition model in the embedded processor.

(31) Firstly, according to the embodiment, the difference in distance features of overlapped tube and non-overlapped tube is used to train and generate the DCNN overlapped tube recognition model. Secondly, according to the characteristics of the embedded processor, the DCNN overlapped tube recognition model is transplanted to the embedded processor in the intelligent temperature measuring instrument. Then, the DCNN overlapped tube recognition model combines the data processing algorithm and temperature value measurement method inside the intelligent temperature measuring instrument so as to calculate and obtain the temperature of the overlapped tube and the non-overlapped tube. The DCNN overlapped tube recognition model trained by the present embodiment can distinguish overlapped tube and non-overlapped tube with high precision, which increases measurement accuracy of outer surface temperature of the crack furnace tube, and realizes edge computing function of the intelligent temperature measuring instrument.

(32) In the step S112, during feature extraction of the furnace tube distance data in the two-dimensional histogram, the temperature data of the outer surface of the furnace tube and the inner wall of the furnace hearth in the two-dimensional histogram are removed first, the distance data of the inner wall of the furnace hearth in the two-dimensional histogram is removed, and then invalid threshold data with low relevance in the furnace tube distance data is removed to extract the distance feature map of overlapped tube and non-overlapped tube of the furnace tube.

(33) The step S122 of training the target data set by using a Gaussian mixture model, including:

(34) initializing a first parameter {(α.sub.i, μ.sub.i, Σ.sub.i)|1≤i≤k} of the Gaussian mixture model; and

(35) computing the sample points to generate a posterior probability of the constituent points by utilizing a first formula:

(36) $p_M\left(z_j=i|x_j\right)=\frac{P\left(z_j=i\right).Math.p_M\left(x_j|z_j=i\right)}{p_M\left(x_j\right)}=\frac{{\alpha }_i.Math.p\left(x_j|{\mathrm{.Math.}}_i,{\Sigma }_i\right)}{{.Math.}_{l=1}^k{\alpha }_l.Math.p\left(x_j|{\mathrm{.Math.}}_i,{\Sigma }_i\right)},$

(37) in which P.sub.M is a posterior probability of Z.sub.j, the posterior probability of Z.sub.j is a posterior probability, generated by an i-th constituent point, of the sample point X.sub.j.

(38) The step S122 of training the target data set by using a Gaussian mixture model, including:

(39) calculating a second parameter {(α′.sub.i, μ′.sub.i, Σ′.sub.i)|i=1, 2, . . . k} of the Gaussian mixture model by using a second formula

(40) $\hspace{1em}\{\begin{array}{c}{\mu }_i=\frac{\underset{j=1}{\overset{m}{.Math.}}{\gamma }_{\mathrm{ji}}{x}_j}{\underset{j=1}{\overset{m}{.Math.}}{\gamma }_{\mathrm{ji}}}\\ {\Sigma }_i=\frac{\underset{j=1}{\overset{m}{.Math.}}{\gamma }_{\mathrm{ji}}\left(x_j-{\mu }_i\right){\left(x_j-{\mu }_i\right)}^T}{\underset{j=1}{\overset{m}{.Math.}}{\gamma }_{\mathrm{ji}}}\\ {\alpha }_i=\frac{1}{m}\underset{j=1}{\overset{m}{.Math.}}{\gamma }_{\mathrm{ji}}\end{array};$

(41) converting the first parameter {(α.sub.i, μ.sub.i, Σ.sub.i)|1≤i≤k} to the second parameter {(α′.sub.i, μ′.sub.i, Σ′.sub.i)|i=1, 2, . . . k};

(42) calculating and generating the posterior probability γ.sub.ji(1≤i≤k) of the constituent points according to the first formula based on the second parameter; and

(43) recalculating the second parameter according to the posteriori probability by using the second formula.

(44) Further, the step S123 of associating sample points in the training data with constituent points of the Gaussian mixture model according to training results, including:

(45) associating each of the sample points in the target data set onto a corresponding constituent point; and

(46) calculating a mean value and/or variance of parameters of the constituent points.

(47) The step S124 of screening the constituent points in a predetermined standard deviation calculation manner, including: performing deviation calculation on the mean value and variance of parameters by using a triple standard deviation method to generate a corresponding range of difference values; screening the constituent points according to the range of difference values, and determining that the constituent points falling outside of the range of difference values are the outliers.

(48) Obviously, the above embodiments of the present invention are merely examples for clear illustrating the technical solution of the present invention, and are not intended to limit the implementations of the present invention. Any modification, equivalent substitution, improvement, or the like within the spirit and principle of the claims of the present invention should be included in the scope of the claims of the invention.