CN115758915A - Heat exchange tube overhauling and intelligent operation and maintenance control method of energy pile active bridge deck deicing and snow melting system - Google Patents

Abstract

The invention belongs to the technical field of geothermal energy development and utilization, bridge engineering and deicing and snow melting, and provides a heat exchange tube overhauling and intelligent operation and maintenance control method of an energy pile active bridge deck deicing and snow melting system, which comprises an energy pile active bridge deck deicing and snow melting federal learning operation and maintenance control model and a heat exchange tube overhauling device based on an electromagnetic induction principle; the energy pile active bridge deck deicing and snow melting federal learning operation and maintenance control model preliminarily judges and gives an alarm for positioning, and a heat exchange tube overhauling device based on the electromagnetic induction principle carries out fine repairing operation; and the damaged point is repaired by an automatic mending agent or a screwed inner-layer pipe. The heat exchange tube overhauling device based on the electromagnetic induction principle comprises a double-layer intelligent heat exchange tube structure, a miniature magnetoelectric sensor, a cable storage disc, an automatic data acquisition instrument and a control end. The invention realizes the fine control management and the advanced early warning of the energy pile active bridge deck deicing and snow melting system, accurately positions and repairs the damaged points, globally and accurately regulates and controls the snow melting system and improves the deicing and snow melting efficiency.

Description

Heat exchange tube overhauling and intelligent operation and maintenance control method of energy pile active bridge deck deicing and snow melting system

Technical Field

The invention relates to the technical field of geothermal energy development and utilization, bridge engineering and deicing and snow melting, in particular to a heat exchange tube overhauling and intelligent operation and maintenance control method of an energy pile active bridge deck deicing and snow melting system.

Background

The phenomenon of icing of accumulated snow on the bridge deck seriously affects the traffic transportation capacity of highways and municipal roads in cities in northern China, and researches show that the icing of the accumulated snow can cause the road surface adhesion coefficient to be reduced by 60-70 percent, thus causing the brake failure and the skidding of vehicles and causing serious traffic accidents. In order to solve the problems of bridge deck corrosion, environmental pollution and high cost in the traditional snow melting technology, researchers provide various active bridge pavement deicing and snow melting technologies, which mainly comprise phase change material deicing, solar energy deicing and snow melting, electric heating of the pavement, geothermal energy deicing and the like. The phase-change material deicing technology has a good deicing effect, but the phase-change material is too high in cost and is mostly used for exemplary test engineering. The solar snow melting technology has low energy consumption, but the deicing stability depends on the illumination condition to a great extent, and the stability is poor. The electric heating snow and ice melting technology has good thermal stability, but has large electric energy consumption and high operation cost. In contrast, shallow geothermal heat has a wide prospect in deicing and snow melting as a widely distributed renewable clean energy source.

The active bridge deck deicing and snow melting technology of the energy pile is characterized in that heat exchange tubes are embedded in a bridge pile foundation and a bridge deck, shallow geothermal energy can be extracted by the heat exchange tubes embedded in the energy pile, and then the shallow geothermal energy is lifted by a heat pump unit and then is conveyed into a heat exchange pipeline of the bridge deck, so that the aim of deicing and snow melting of the bridge deck is finally fulfilled. The technology has the advantages of controllable construction cost, simple construction, high stability and low operation cost.

The invention discloses a Chinese patent application No. CN202010660161.9, which is named as an electric heating ice and snow preventing and melting system for roads and a laying method thereof, and discloses the electric heating ice and snow preventing and melting system for the roads, wherein the electric heating ice and snow preventing and melting system comprises a plurality of heating systems, and each heating system comprises a road section, a road surface slippery state sensor, a temperature sensor and a plurality of heating cables. The heating cable replaces the traditional spraying mode, and the road surface condition can be monitored in real time to automatically heat and deice.

The invention discloses a Chinese patent application No. CN 202110608504.1, which is named as an intelligent snow and ice melting system for urban pedestrian roads and a construction method thereof, and discloses a snow and ice melting system for urban pedestrian roads, which comprises an intelligent control module, a data acquisition module, a pedestrian road module and a terminal system; the intelligent control module comprises a server and an intelligent controller, and the data acquisition module comprises a high-definition camera, an icing detector and a temperature and humidity sensor.

The invention discloses a road ice and snow melting system and method for a ground heat pump, which is named as a road ice and snow melting system and method for a ground heat pump with a buried pipe, and discloses the road ice and snow melting system for the ground heat pump with a ground heat pump unit, a shallow geothermal heat exchange part structure and a heating pipeline part structure.

The invention discloses a Chinese patent application No. CN 202210017291.X, which is named as an active ice and snow melting system for a road surface and a control method thereof, and discloses a control method for comparing and controlling the start and stop of the system by acquiring a road surface temperature and a road ice and snow picture and a database.

The invention discloses a ground source heat pump control method and system, which is named as CN201610200872.1, and realizes advanced control by using a predicted value of greenhouse temperature instead of a real-time temperature value as a part of input of a heat pump unit controller.

The invention discloses a method and a device for detecting and optimally controlling the performance of a ground source heat pump system, which are named as a method and a device for detecting and optimally controlling the performance of the ground source heat pump system and have the application number of CN202010244228.0, and further determine the real-time operation state of the system by calculating the energy efficiency ratio of a real-time unit operated by the system according to collected real-time operation data of the ground source heat pump system. And determining a system adjusting mode by comparing the historical data of the system operation.

The research on the energy pile active bridge deck deicing and snow melting technology mainly focuses on system composition, equipment and construction methods, and operation and maintenance control research aiming at the energy pile active bridge deck deicing and snow melting system is lacked. The few control methods mainly aim at controlling the ground source heat pump unit according to the accumulated snow condition of the bridge floor or the requirement change of the tail end of the air conditioner. In actual engineering, the operation of the active energy pile bridge deck deicing and snow melting system is dynamic coordination among a bridge deck buried pipe end, a unit equipment end and an energy pile buried pipe end, only a heat pump unit is controlled, the stable operation of the system is not facilitated, overload alarm of a host is easily caused, and the service life is shortened. Therefore, a scientific and reasonable method for fine operation and maintenance regulation and cooperative management is needed to improve the correctness of the control scheme and the stability of system operation.

Machine learning is a scientific technology that enables a computer to automatically analyze and obtain rules from a class of data by establishing a suitable algorithm and predict unknown data using the rules. A large number of data samples are trained through machine learning, control schemes under different operation conditions are obtained, and the method can be used for intelligent control of an energy pile active bridge deck deicing and snow melting system and initial positioning and alarming of faults in time.

The utility model discloses a chinese utility model patent application number is CN201721077937.4, and the name is a can survey polyethylene composite pipe forever, discloses a can survey polyethylene composite pipe forever, including the bilayer structure tubular product that polyethylene layer and modified PP coating constitute, whether the electric current through the universal meter detection trace strap between inlayer and the skin judges the body and has destroyed.

The invention has the Chinese patent application number of CN202111660338.6, is named as an anti-blocking dredging system for municipal road engineering and a working method thereof, and discloses the anti-blocking dredging system for the municipal road engineering, which consists of a dredging trolley, a trolley driving mechanism, a visual detection mechanism, an industrial and mining detection mechanism and a display mechanism.

The metal strips in the composite pipeline technology need to be assembled and connected, short circuit or open circuit is easy to occur in the construction stage, and the municipal pipeline dredging system is only limited to a large-diameter municipal pipe network. The existing technology can not be applied to detection and maintenance of deep-buried energy piles with small calibers and heat exchange pipes in the pavement layer of the bridge deck slab.

Disclosure of Invention

In order to overcome the following defects and problems in the prior art: (1) lack of control methods for deck pipe laying structures; (2) An active multi-module cooperative dynamic control method for deicing and snow melting of a bridge floor of an energy pile is lacked. (3) Various heat exchange tubes are designed permanently or semi-permanently, and cannot be maintained after faults occur. Therefore, the invention provides a heat exchange tube overhauling and intelligent operation and maintenance control method of an energy pile active bridge deck deicing and snow melting system.

The technical scheme of the invention is as follows: a heat exchange tube overhauling and intelligent operation and maintenance control method of an energy pile active bridge deck deicing and snow melting system comprises an energy pile active bridge deck deicing and snow melting federal learning operation and maintenance control model and a heat exchange tube overhauling device based on an electromagnetic induction principle, wherein the energy pile active bridge deck deicing and snow melting federal learning operation and maintenance control model is formed by a central server and a local client; performing preliminary judgment and preliminary alarm positioning according to an energy pile active bridge deck deicing and snow melting federal learning operation and maintenance control model formed by a central server and a local client, and arranging a heat exchange tube overhauling device based on an electromagnetic induction principle to perform fine repairing operation on the basis;

the active bridge deck deicing and snow melting federal learning operation and maintenance control model of the energy pile formed by the central server and the local clients comprises 3 local client modules: the device comprises a bridge deck buried pipe end control module, a unit equipment control module and an energy pile buried pipe end control module; the local models of the local modules of the 3 local clients are subjected to learning training on the local clients, then calculation results are encrypted and uploaded to a central processing unit to pass through a global federal learning model, and results obtained by the global federal learning model are issued to the local models of the local clients again; through repeated iteration between the client and the central server until the global federated learning model is stable, finally different module regulation and control schemes matched with each other are made;

the method comprises the following specific steps:

d1: acquiring bridge deck pipe burying end operation parameters, bridge deck pipe burying end control parameters, unit equipment operation parameters, unit equipment control parameters, energy pile pipe burying end operation parameters and energy pile pipe burying end control parameters, performing sample data quality analysis and cleaning treatment, extracting a data distribution rule for the cleaned sample data, analyzing sample data characteristics, and distributing the sample data characteristics to corresponding calculation module client sides;

the sample data may be obtained by at least one of: numerical simulation calculation, an on-site monitoring technology, a questionnaire survey technology, an image recognition technology and a thermal infrared technology.

The data-washed object comprises invalid values, abnormal values, missing values and repeated values. The content of data cleansing includes identifying invalid values, outliers, and missing values; outliers, invalid values, and missing values are processed.

Alternatively, the identification method of invalid, abnormal, and missing values may be implemented based on one of: 1. by drawing a scatter diagram which can reflect the correlation between the two groups of data, the outlier is intuitively identified as an abnormal value or an invalid value; 2. abnormal values can be identified by the 3 σ principle when the data obeys normal distribution, in which case, a value having a deviation of more than 3 times of the standard deviation from the average among a group of measurement values is defined as an abnormal value; 3. and (4) counting the maximum value, the minimum value, the median and the upper quartile and the lower quartile of the data set and drawing a box diagram. Outliers or invalid values may be identified based on the upper and lower quartiles of the boxplot.

Alternatively, the processing method of invalid values, abnormal values and missing values may be implemented based on one of the following: 1. deleting directly; 2. fitting invalid values, abnormal values and missing values according to a regression model or maximum likelihood estimation; 3. and filling invalid values, abnormal values and missing values according to the mean, median and mode of the statistical data characteristics.

D2: update the local model of 3 local clients: a bridge deck buried pipe end control support vector machine local model, a unit equipment control random forest local model and an energy pile buried pipe end control decision tree local model;

d3: encrypting the local model results of the 3 local clients into desensitization parameters in a public key mode, and uploading the desensitization parameters to a central server; the relevant data of the client i is expressed as

Can be encrypted into

D4: the central server decrypts the encryption desensitization parameters uploaded by the 3 local clients by using a private key, performs decoding operation on the encryption desensitization parameters, performs security aggregation, and then updates a global federated learning model; the central server can pass

A decoding operation of the encryption desensitization parameter is performed. Updating a global federal learning model by federal learning at the central server side according to one or more synchronous modes of gradient averaging, federal averaging and knowledge distillation; after the weight of the global federal learning model is updated in each round, the central server calculates the error and the accuracy of the global federal learning model;

preferably, all client models are aggregated for federal learning based on knowledge distillation, according to formula

Global model weight update for 3 clients, where W _n+1 As the global model parameters for the nth round,

and uploading the client sub-model weight of the server for the nth round of clients i. After each round of model weight updating, the central server calculates the error and accuracy of the global model.

D5: the central server generates a public key for encrypting transmission data from the global federal learning model and sends the public key to each client; according to the global federal learning model, each local client updates the iteration results of other clients as new sample data attribute parameters; for example, the random forest local model of the unit equipment control module client updates the operation parameters and the control parameters of the bridge deck buried pipe control client and the operation parameters and the control parameters of the energy pile buried pipe control client according to the global shared model.

D6: repeating the steps D2-D5 for continuous iteration until the global federal learning model is stable, and finally calculating by the client according to the global federal learning model to obtain a corresponding result; and calculating the control states matched with each other in the active bridge deck deicing and snow melting system of the energy piles, wherein the control states comprise the opening degrees of various regulating valves of a frozen water circulation loop at the end of the buried pipe of the bridge deck, the snow melting time consumption of the end of the buried pipe of the bridge deck, unit control parameters and energy pile buried pipe end control parameters.

The operation parameters of the bridge deck pipe burying end comprise: ambient temperature, ambient humidity, wind speed, amount of snow falling, surface snow free rate, water level of water collector, differential pressure difference, flow rate of collector, pressure of collector, temperature of collector the system comprises a bridge floor buried pipe water supply temperature, a bridge floor buried pipe backwater temperature, a bridge floor buried pipe water supply flow, a bridge floor buried pipe backwater flow, a bridge floor buried pipe water supply pressure and a bridge floor buried pipe backwater pressure; the unit equipment operating parameters comprise: working liquid level of a condenser, inlet temperature of the condenser, working liquid level of an evaporator, inlet temperature of the evaporator and pressure of a compressor; the operating parameters of the energy pile pipe burying end respectively comprise: the soil temperature, the water level of the water collector at the buried pipe end of the energy pile, the pressure difference, the flow rate of a collecting pipe, the pressure, the temperature, the water supply temperature of the energy pile, the return water temperature, the flow rate and the water pressure; the control parameters of the buried pipe end of the bridge deck comprise: the opening of a bridge floor buried pipe backwater control valve, the opening of a bridge floor buried pipe water supply control valve, the opening of a collector pipe control valve, the opening of a collector branch pipe control valve, the opening of a water distributor collector pipe control valve, the opening of a water distributor branch pipe control valve and the opening of a water replenishing pump control valve; the unit control parameters include: the opening degree of a water replenishing pump control valve, the opening degree of a cooling circulating pump control valve, the opening degree of a freezing circulating pump control valve, the opening degree of an expansion valve and a host machine; the energy pile pipe burying end control parameters comprise: the device comprises an energy pile water collector control valve opening, a water collector branch pipe control valve opening, an energy pile water distributor collector control valve opening, a water distributor branch pipe control valve opening, an energy pile backwater control valve opening and a water supply control valve opening.

The energy pile pipe burying control module is used for establishing a decision tree local model of energy pile pipe burying control parameters according to the running parameters of the energy pile pipe burying end, the unit equipment running parameters updated by the central server and the bridge deck pipe burying end running parameters as sample attribute data; determining the heat taking quantity of the energy pile and control parameters of the energy pile pipe burying end, encrypting and uploading the parameters to a central processing unit; the decision tree model is formed by one or more of the following ID3 algorithm, C4.5 algorithm and CART algorithm; the method comprises the following specific steps:

d2.1.1: decrypting the public key of the encrypted transmission data by using a private key, updating a bridge floor buried pipe end control module, a unit equipment control module and an energy pile buried pipe end control module of 3 local client sides according to a global federal learning model of a central server, and taking an iteration result as a new sample data attribute parameter; calculating all attribute values A corresponding to any attribute A _i The coefficient of kini of (a);

given the kini coefficient for sample set D:

then the set D has a kini coefficient of

Wherein, C _k For the subset of samples in D that belong to class k, k being the number of types, D ₁ 、D ₂ For the sample set D, according to whether the attribute A is equal to A _i A subset of the segmented samples.

D2.1.2: selecting the attribute with the minimum Gini coefficient and the corresponding attribute value as the optimal attribute and the optimal segmentation point from all the attributes A and all the attribute values corresponding to the attributes A; generating two child nodes by using the optimal attribute and the optimal segmentation point, and distributing corresponding samples to root nodes;

d2.1.3: taking each root node as a complete data set, iteratively calling the step D2.1.1-the step D2.1.2, dividing samples according to a suboptimal attribute as a basis, taking the samples with the same suboptimal attribute value as the same sample set to form leaf nodes, and sequentially iterating by adopting a REP method, a PEP method or an MEP method until a decision tree pruning condition is met and a stop condition that the number of samples in the leaf nodes or the kiney coefficient is less than a threshold value is reached to form a decision tree;

d2.1.4: comparing the decided control parameters of the energy pile pipe burying end with an equipment control threshold, and when the calculation result meets the equipment control threshold, converting the decision calculation result into encryption parameters by using a public key as advanced prediction control information and uploading the encryption parameters to a central server for global federal learning model iteration; when the calculation result exceeds the device control threshold, executing a step D2.1.5;

d2.1.5: outputting the decided control parameters of the energy pile buried pipe end and giving an alarm, taking a control threshold value as advanced prediction control information, converting the advanced prediction control information into encryption parameters by using a public key, and uploading the encryption parameters to a central server for iteration of a global federal learning model; arranging a heat exchange tube overhauling device based on the electromagnetic induction principle to perform refined detection and positioning;

the bridge floor buried pipe end control module takes the running parameters of the bridge floor buried pipe end, the unit equipment running parameters updated by the central server and the running parameters of the energy pile buried pipe end as sample attribute data of a bridge floor buried pipe end control local model; establishing a bridge deck buried pipe end control support vector machine local model, predicting a calculation control scheme, encrypting and uploading to a central server; the method comprises the following specific steps:

d2.2.1: decrypting the public key of the encrypted transmission data by using a private key, updating the energy pile embedded pipe end control module, the unit equipment control module and the bridge deck embedded pipe end control module client according to a global federal learning model of a central server, and taking an iteration result as a new sample data attribute parameter; taking the environmental temperature, the environmental humidity, the wind speed, the snow fall amount, the snow-free rate of the surface, the water level, the pressure difference and the temperature difference of a water collector, the temperature, the flow and the pressure of the supply and return water of the buried pipe of the bridge deck as a sample attribute space set, and taking control parameters of the buried pipe end of the bridge deck as a learning target;

d2.2.2: introducing relaxation variables xi _i Constructing a nonlinear segmentation support vector classifier considering soft intervals with a penalty coefficient C, and representing the running parameters of the end of the bridge deck buried pipe and the bridge deck buried pipeThe relationship between the end control parameters; the prediction accuracy and the self stability of the local model of the bridge deck buried pipe end control support vector machine are expressed by a loss function L = max (0, | z | - ∈); converting the loss function into a condition maximum function;

wherein, w is a normal vector of the hyperplane, C is a penalty coefficient, xi and xi ^* Is a relaxation factor, and belongs to a hyper-parameter for determining the width of the boundary; y is _i The actual measurement result of the training sample is obtained; f (X) _i ) = w.phi (X) + b is a classification hyperplane of a bridge deck buried pipe end control support vector machine local model; i is a training sample number; n is the number of training set samples; phi (X) is a nonlinear mapping function;

d2.2.3: converting the condition most-valued function into a multivariate function through a Lagrange function to solve, making the partial derivative of the Lagrange function to the optimization target w, b and xi be 0 to obtain a Lagrange multiplier, and converting the condition most-valued function into a dual function so as to find the minimum value of the prediction boundary;

d2.2.4: nonlinear mapping function phi (X) contained in classification hyperplane in bridge floor buried pipe end control support vector machine local model, and inner product phi (X) of nonlinear mapping function phi (X) _i ) ^T φ(X _j ) Selecting one or more than one kernel functions of a Gaussian kernel, a linear kernel, a polynomial kernel and a Sigmoid kernel for combined processing;

d2.2.5: optimizing model parameters in a local model of a bridge deck buried pipe end control support vector machine: an insensitive loss function E, a penalty coefficient C, and hyper-parameters gamma, lambda, alpha, C and d in a kernel function; selecting one or more optimization methods of a simulated annealing method, a grid search method, a particle swarm optimization method, a PSO algorithm and a genetic algorithm;

d2.2.6: inputting target operation parameters of the bridge deck buried pipe, and predicting and calculating the bridge deck buried pipe end by using a trained bridge deck buried pipe end control support vector machine local model;

d2.2.7: comparing the calculation result with a control threshold of the chilled water circulation loop control valve, and when the calculation result meets the control threshold of the chilled water circulation loop control valve, taking the prediction calculation result as advanced prediction control information, converting the advanced prediction control information into encryption parameters by using a public key, and uploading the encryption parameters to a central server for global federal learning model iteration; when the calculation result exceeds the control threshold value of the chilled water circulation loop control valve, executing a step D2.2.8;

d2.2.8: outputting the calculated control result of the buried pipe end of the bridge deck and giving an alarm, taking an equipment control threshold value as advanced prediction control information, converting the advanced prediction control information into encryption parameters by using a public key, and uploading the encryption parameters to a central server for iteration of a global federal learning model; meanwhile, arranging a heat exchange tube overhauling device based on the electromagnetic induction principle to carry out fine detection and positioning;

the unit equipment control module is used for establishing a random forest local model of unit equipment control parameters according to various operation parameters of the unit equipment, energy pile operation parameters updated by a central server and bridge floor buried pipe end operation parameters as sample attribute data, determining the unit equipment control parameters and encrypting and uploading the unit equipment control parameters to the central processor; the method comprises the following specific steps:

d2.3.1: decrypting the public key of the encrypted transmission data by using a private key, and using the energy pile running parameter updated according to a global federal learning model of a central server and the running parameter of the buried pipe end of the bridge deck as a new sample data attribute parameter; according to local sample data, taking each state parameter of each unit device as a sample attribute space set, and taking a unit device control parameter as a learning target; randomly generating a sample subset from a unit equipment control training sample by using a Bootstrap sampling method, taking the sample subset as a training sample of one decision tree model, and repeatedly sampling for k times to form k decision tree training samples;

d2.3.2: performing decision tree training according to the attribute subsets in the k training samples to form k mutually independent random decision trees;

d2.3.3: voting the unit equipment control schemes predicted by the k decision trees by a random selection classifier of a random forest, and taking a voting result as an optimal control scheme of the unit equipment;

d2.3.4: comparing the decided unit equipment control parameters with the unit equipment control threshold, and when the calculation result meets the unit equipment control threshold, taking the decision calculation result as advanced prediction control information, converting the advance prediction control information into encryption parameters by using a public key, and uploading the encryption parameters to a central server for global federal learning model iteration; when the calculation result exceeds the control threshold value of the unit equipment, executing a step D2.3.5;

d2.3.5: and outputting the decided control result of the unit equipment and giving an alarm, taking a control threshold value of the unit equipment as advanced prediction control information, converting the advanced prediction control information into encryption parameters by using a public key, and uploading the encryption parameters to a central server for iteration of a global federated learning model.

When the simulated annealing method is adopted for optimization in the step D2.2.5, the method comprises the following steps;

d2.2.5.1: randomly generating an initial parameter set for interactive verification, and recording an error value EEP as a current annealing system state E ₀ Initial temperature T ₀ The annealing end temperature is T ₁ ；

D2.2.5.2: m 'according to a perturbation algorithm' _i ＝m _i +s·(μ-0.5)(B _i -A _i ) Disturbing the parameters to form a new parameter set, and obtaining the state E of the current annealing system through interactive verification _n Calculate Δ E = E _n -E _n-1 ；

Wherein: m' _i For perturbed variables, m _i Is the current variable, s is the perturbation ratio, μ is [0,1 ]]Random number of (A), B _i 、A _i Is the current variable m _i A range of (a);

d2.2.5.3: when the delta E is less than 0, a new parameter set is accepted, and the step is jumped to the step D2.2.5.5; otherwise, receiving the corresponding parameter set according to Metropolis criterion exp (delta E/KT) -mu > 0, and jumping to the step D2.2.5.4; if the above conditions are not met, refusing to accept the critical state, returning to the step D2.2.5.2, generating a new parameter set in a disturbing way again, and carrying out interactive verification until the parameter set acceptance condition in the step D2.2.5.3 is met;

d2.2.5.4: setting the end temperature to T ₁ Setting the EEP times of the global maximum calculation error value as N; when T is reached ₁ Or stopping annealing when N is generated, otherwise, returning to the step D2.2.5.2;critical state cross validation error E accepted at stop anneal _n At the lowest, the corresponding parameter is the best prediction parameter.

The heat exchange tube overhauling device based on the electromagnetic induction principle comprises a double-layer intelligent heat exchange tube structure 1, a miniature magnetoelectric sensor 3 containing a dredging module, a cable storage disc 4, an automatic data acquisition instrument 5 and a control end 6; the double-layer intelligent heat exchange tube structure 1 is U-shaped or snakelike, the U-shaped double-layer intelligent heat exchange tube structure 1 is arranged in the energy pile, and the snakelike double-layer intelligent heat exchange tube structure 1 is arranged in a bridge deck; two ends of the double-layer intelligent heat exchange tube structure 1 are respectively connected with an automatic data acquisition instrument 5 through a cable storage disc 4, and the automatic data acquisition instrument 5 is connected with a control end 6; the double-layer intelligent heat exchange tube structure 1 comprises inner polyethylene tubes 1-A, outer polyethylene tubes 1-B and magnetic induction rings 2; the inner polyethylene pipe 1-A is positioned in the inner wall of the outer polyethylene pipe 1-B, and lubricating oil is filled between the inner polyethylene pipe 1-A and the outer polyethylene pipe 1-B and is in threaded connection; the magnetic induction rings 2 are embedded in the pipe walls of the outer polyethylene pipes 1-B and are arranged at equal intervals along the axial direction of the double-layer intelligent heat exchange pipe structure 1; the miniature magnetoelectric sensor 3 containing the dredging module moves in the inner polyethylene pipe 1-A and comprises a dredging part, a series motor 3-5, a driving wheel 3-7 and a cable 3-8; one end of the series excited motor 3-5 is connected with a cable 3-8, and the other end is fixedly connected with the dredging piece through a connecting component 3-2 by mounting a gear 3-3; the side surface of the series excited motor 3-5 is connected with a plurality of driving wheels 3-7 to slide on the inner wall of the inner polyethylene pipe 1-A; the magnetoelectric sensors 3-6, the surrounding cameras 3-4 and the illuminating lamps 3-9 are all arranged on the surface of the series excitation motor 3-5; the magnetoelectric sensors 3-6 are fed with electric signals, when the magnetoelectric sensors pass through the magnetic induction ring 2, the electromagnetic frequency changes, and the electric signals are transmitted to the automatic data acquisition instrument 5 through the connected cables 3-8; the data automatic acquisition instrument 5 automatically records the time when the magnetoelectric sensors 3 to 6 pass through each magnetic induction ring 2 in sequence, transmits the information to the control end 6, and judges that the two magnetic induction rings 2 are blocked when the electrical signal transmission of the next magnetic induction ring 2 still does not exist after the magnetic induction rings 2 pass through a set time.

The magnetic induction ring 2 is 5-8 mm wide and is integrally formed with the outer polyethylene pipe 1-B; the distance between the magnetic induction rings 2 is 0.3-0.5 m; the dredging piece is a spiral metal wire 3-1 and/or a hot melting rod.

The heat exchange tube overhauling device based on the electromagnetic induction principle is used for detecting blockage and damage of a heat exchange pipeline and comprises the following steps:

s1: after the energy pile bridge surface deicing and snow melting federal learning operation control gives out a failure preliminary alarm, a switch of the automatic data acquisition instrument 5 is turned on, a magnetic induction ring is used for covering the magnetoelectric sensors 3-6 to move, when the magnetic induction ring meets induction points of the magnetoelectric sensors 3-6, an audible and visual alarm is given out, meanwhile, the automatic data acquisition instrument 5 gives an indication, and the automatic data acquisition instrument 5 works normally to carry out the subsequent steps;

s2: during testing, pipelines are arranged in the energy piles in the vertical direction, and the magnetoelectric sensors 3-6 are rotationally placed in the cable storage disc 4 by utilizing the gravity of the magnetoelectric sensors; a pipeline is arranged in the bridge deck pavement layer in the horizontal direction, and the magnetoelectric sensor 3-6 moves the miniature magnetoelectric sensor 3 containing the dredging module in the pipeline through the driving wheel 3-7; when the center of the magnetoelectric sensor 3-6 is intersected with the magnetic induction ring 2, the automatic data acquisition instrument 5 sends out buzzing sound accompanied with light indication; starting the surrounding cameras 3-4 to perform image real-time monitoring and sound wave monitoring in the pipeline, and automatically recording the number of the measuring points passing through each magnetic induction ring 2 and the interval time recording, real-time image and sound wave waveform information of the measuring points passing through two adjacent magnetic induction rings 2 by the automatic data acquisition instrument 5;

s3: the automatic data acquisition instrument 5 transmits the automatically recorded data to the control end 6, and one or more of the following methods are adopted to analyze the acquired data and accurately position the damage point: isolated forest algorithm, X-ray digital imaging technology, ultrasonic detection, ultrasonic guided wave, ultrasonic C scanning, ultrasonic phased array and high-temperature thickness measurement method; and the control end 6 accurately positions the position of the plugging point according to the analyzed abnormal time and position of the data.

The ultrasonic phased array method in the step S3 is to change the phase relationship when the array element receives the sound wave from a certain point in the object by controlling the excitation of each array element in the array transducer and the time delay of receiving the pulse, and change the focus point and the orientation of the sound beam to synthesize the phased array beam, thereby scanning the information.

In the ultrasonic guided wave detection in the step S3, a plurality of sensor probes are placed at a plurality of positions in the pipeline, so that the sensor probes form a sensor to sense the condition of the pipeline wall, the damage condition of the pipeline wall is detected along with the information sensed by the sensor in the pipeline through the impact and reflection of the guided wave, and the problem of the pipeline wall is found out through the analysis and calculation of related data in the later period. The proper wave mode is generated under the normal pipe wall thickness, the frequency which is much lower than that of the normal ultrasonic flaw detection is needed, the frequency which is usually used by the guided wave is 60-100 kHz, therefore, the detection sensitivity of the guided wave to a single defect is lower than that of the normal ultrasonic flaw detection which uses the frequency in the MHz level, but the guided wave detection has the advantages that the guided wave can propagate for a long distance of 20-30 m and has small attenuation, therefore, the pulse echo array can be fixed at one position for detection in a wide range, and the method is particularly suitable for detecting the inner and outer wall corrosion of the in-service pipeline and the dangerous defects of a welding seam. The low-frequency guided wavelength distance ultrasonic detection method is used for quickly detecting the in-service state of a pipeline, the corrosion of the inner wall and the outer wall can be detected at one time, and the plane defects of the section of the pipeline can also be detected.

The ultrasonic C-scan detection in the step S3 is to perform depth scan on the pipe, and the basic working principle is that ultrasonic waves are generated inside the detected component through a reflection probe, and the received information about the defect part is transmitted back through the probe by using the ultrasonic waves, and various index conditions of the reflected waves are displayed on software, so as to determine the specific information about the position and size of the pipe defect.

When the data collected in the step S3 is analyzed by adopting an isolated forest algorithm, an isolated tree is established for isolating attributes in the attribute space in the sample set, abnormal values are separated through model training in the same attribute space, and normal values are located in deep subnodes of the isolated tree; the method comprises the following specific steps:

s3.1: the current signal, the sound wave signal and the electromagnetic signal transmitted to the control end 6 on the magnetoelectric sensors 3-6 are used as sample attributes to form a sample set, and the sample set is represented as X = (X) ₁ ,x ₂ ,x ₃ ,…,x _n ) Wherein x is _n ＝(x _n1 ,x _n2 ,x _n3 ,…,x _nm )，x _nm Is the nth type attribute signal at the mth moment;

s3.2: by usingThe attribute dimension q serves as an attribute cutting point to cut attribute types to form different isolated trees, the attribute value cutting point p serves as an abnormal value cutting point for each isolated tree, and the attribute values in the attribute space are cut; q is a random value falling within the range (0, n), and p is a random value falling within min (x) _n )＜p＜max(x _n ) A random value of the interval;

s3.3: forming a hyperplane in a sample space according to the segmentation of the attribute cutting point q and the abnormal value cutting point p, classifying samples smaller than the abnormal value cutting point p into one class to form a left node of the isolated tree, and otherwise forming a right node;

s3.4: repeating the step S3.2 to the step S3.3 until the leaf node of each isolated tree only contains one sample, and finishing the training of the isolated tree model;

s3.5: sample data x at each moment _i Putting the sample into an isolated tree model for traversal calculation, and calculating to obtain the average height h (x) of each sample data _i ) Computing an anomaly score for the sample data according to:

wherein h (x) is the height of x in each isolated tree, c (n) is the average value of the path length at a given number of samples n, and E (h (x)) is the expected path length value of x in a plurality of isolated trees;

when the score S (x, psi) > 0.75, the signal corresponding to the sample at the moment is regarded as an abnormal signal;

when the score S (x, psi) is 0.4-0.75, the signal corresponding to the sample at the moment is regarded as a suspected abnormal signal;

when the score S (x, ψ) < 0.4, the sample at that time is regarded as a normal signal.

The heat exchange tube overhauling device based on the electromagnetic induction principle is used for dredging and repairing a heat exchange pipeline and comprises the following steps:

t1: starting the surrounding cameras 3-4 and the illuminating lamps 3-9, and transmitting the surrounding cameras and the illuminating lamps to the control end 6 in real time to enable the surrounding cameras 3-4 to be aligned to the blockage;

t2: starting dredging work, starting a series excitation motor 3-5 for controlling the spiral metal wire 3-1 when a blockage except the blockage caused by hot melting of the pipeline is found, and rotating the spiral metal wire 3-1 to a working rotating speed; slowly lowering a sensor cable 3-8 in a vertical pipeline by using the self gravity of the magnetoelectric sensor 3-6; controlling a magnetoelectric sensor 3-6 to move by using a driving wheel 3-7 in the pipeline in the horizontal direction, and slowly lowering a sensor cable 3-8; the spiral metal wire 3-1 stirs the blocked foreign matters until the foreign matters are removed; when the blockage is caused by pipeline hot melting, the spiral metal wire 3-1 and the connecting component 3-2 of the dredging module are replaced by hot melting rods, the magnetoelectric sensor 3-6 is placed at the blockage again, the heating rod is started to reach the hot melting temperature of the polyethylene material, and the sensor cable 3-8 is slowly lowered by utilizing the self gravity of the magnetoelectric sensor 3-6 until the blockage is dredged;

t3: taking out the miniature magnetoelectric sensor 3 containing the dredging module from the pipeline, carrying out pipeline pressure test and flushing work, cleaning foreign matters remained in the pipeline, and finishing the dredging work;

t4: and after the damaged point is accurately positioned, repairing the heat exchange pipeline, wherein the pipeline section where the damaged point is located is screwed out along the thread to replace the damaged pipeline section or an automatic repairing agent is adopted to repair the damaged section.

The automatic mending agent is one or more of polyethylene powder, polyethylene particles, polyamide fiber yarns, glycol and a binder; when the automatic mending agent is nylon fiber yarn, the fineness is 200-300D, and the length of the nylon fiber yarn is 3-5 mm, so that the mixture is prevented from agglomerating in the tube, and the effect of the mending agent is reduced. The nylon fiber yarns, the polyethylene powder, the rubber powder and the particles are twisted together, so that the polyethylene powder is difficult to fall off and participates in filling the holes of the heat exchange pipeline. Because the pipe wall is of a double-layer structure, the fiber yarns penetrate through two sides of the hole, so that the polyethylene microparticles are firmly fixed.

The repairing process and method of the automatic repairing agent are as follows:

p1: after the damaged point is positioned, plugging any one end of the inlet and outlet of the pipeline under the pressure test standard, and injecting an automatic mending agent into the pipe at the other pipeline port;

p2: after the repairing agent is injected, the pipe is pressurized, but the pressure is not more than the rated pressure which can be born by the adopted polyethylene pipeline;

p3: when the reading of the pressure gauge is stable, the repair of the damaged point is primarily completed; and after the automatic repairing agent completely meets the strength requirement, carrying out pressure test on the pipeline, and under the test pressure, stabilizing the pressure for 15-20 min, wherein the pressure after pressure stabilization is reduced to 1% -3%, and the repairing work is considered to be finished if no leakage phenomenon exists.

The invention has the beneficial effects that: compared with the prior control technology, the invention has the following technical advantages:

(1) The bridge floor buried pipe and the energy pile have the characteristics of permanence and concealment, the traditional sensor is high in laying difficulty and cannot predict abnormal information, and the intelligent operation and maintenance control method based on the machine learning method conducts advanced prediction on the abnormal information by analyzing various historical data.

(2) The invention provides a heat exchange tube overhauling and intelligent operation and maintenance control method of an energy pile active bridge deck deicing and snow melting system, which realizes comprehensive regulation and control among a plurality of modules of a bridge deck buried tube end, unit equipment and an energy pile, makes a decision, has a scientific and reasonable control scheme, and improves the stability and service life of the bridge deck deicing and snow melting system.

(3) The heat exchange tube overhauling device based on the electromagnetic induction principle combines the double-layer intelligent heat exchange tube structure with the miniature magnetoelectric sensor, and has the advantages of simple operation steps, strong operability, convenient control and easy realization. The metal ring embedded in the double-layer intelligent heat exchange tube structure is simple in structure, and the prefabrication method is simple and convenient; the miniature magnetoelectric sensor has the effect of monitoring and mediation pipeline concurrently, and the function is various.

(4) The invention provides a method for detecting blockage and damage of a heat exchange pipeline by using a heat exchange pipe overhauling device based on an electromagnetic induction principle in the whole operation and maintenance process. In the testing process, if the discovery pipeline blocks up, then start the control module immediately and dredge the module, carry out the pipeline mediation, when the pipeline is damaged, can be immediately with inlayer heat transfer pipeline screw-out change, improve the efficiency of discovery and solution problem.

(5) The automatic repairing agent is easy to obtain and economical in material, the using method of the repairing agent is easy to operate and good in repairing effect, and the structural strength of the heat exchange pipe is enhanced.

Drawings

FIG. 1 is a conceptual diagram of an alternative energy pile active bridge deck deicing and snow melting federal learning operation and maintenance control model according to an embodiment of the invention;

FIG. 2 is a schematic diagram of a partial model structure of an alternative energy pile pipe burying control parameter decision tree according to an embodiment of the present invention;

fig. 3 is a schematic flow chart of an alternative decision tree partial model of energy pile pipe burying control parameters according to the embodiment of the present invention.

FIG. 4 is a schematic view of a partial model flow of an alternative bridge deck buried pipe end control support vector machine according to an embodiment of the present invention;

FIG. 5 is a schematic diagram of a partial model structure of an alternative bridge deck buried pipe end control support vector machine according to an embodiment of the invention;

FIG. 6 is a schematic diagram of an algorithm flow of an alternative SVM improved by a simulated annealing method according to an embodiment of the present invention;

FIG. 7 is a schematic diagram of a random forest local model structure of an optional control parameter of the unit equipment according to the embodiment of the present invention;

FIG. 8 is a schematic diagram of a random forest local model structure of an optional crew device control parameter according to an embodiment of the present invention;

fig. 9 is a schematic diagram of an alternative heat exchange pipe inspection and repair device based on the electromagnetic induction principle according to an embodiment of the present invention;

FIG. 10 is a schematic diagram of an alternative double-layer intelligent heat exchange tube structure according to an embodiment of the invention;

FIG. 11 is a schematic cross-sectional structural view of an alternative double-layer intelligent heat exchange tube structure according to an embodiment of the invention;

fig. 12 is a schematic structural view of an alternative miniature magnetoelectric sensor with a dredging module according to an embodiment of the present invention;

fig. 13 isbase:Sub>A schematic cross-sectional viewbase:Sub>A-base:Sub>A of an alternative micro-magnetoelectric sensor structure includingbase:Sub>A dredging module according to an embodiment of the present invention.

Fig. 14 is a schematic structural diagram of an optional isolated forest algorithm for detecting an abnormal signal in the embodiment of the present invention.

In the figure: the device comprises 1-A-inner polyethylene pipe, 1-B-outer polyethylene pipe, 2-magnetic induction rings, 3-miniature magnetoelectric sensors containing dredging modules, 3-1-spiral stainless steel metal wires, 3-2-connecting members, 3-3-gears, 3-4-surrounding cameras, 3-5-series excitation motors, 3-6-magnetoelectric sensors, 3-7-driving wheels, 3-8-cables, 3-9-illuminating lamps, 4-cable containing discs, 5-automatic data acquisition instruments and 6-control ends.

Detailed Description

The following detailed description of the embodiments of the present invention will be described in conjunction with the accompanying drawings, and the scope of the invention is not limited to the description of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.

The embodiment of the invention provides a control method of an energy pile active type bridge deck deicing and snow melting federal learning operation and maintenance control model formed by an optional central server-local client, and fig. 1 is a structural schematic diagram of the energy pile active type bridge deck deicing and snow melting federal learning operation and maintenance control model, which is mainly formed by field equipment, a data acquisition part, a server and a control signal 4 part. Learning calculation of all sub-modules is realized at a local client, and the learning calculation comprises 3 client modules: the device comprises a bridge deck buried pipe end control module, a unit equipment control module and an energy pile buried pipe end control module. And finally, summarizing the desensitization parameters calculated by the client to a central server for calculation, and then issuing the desensitization parameters to each data holder (client) to update the local model of the client until the global model is stable. The method comprises the following specific steps:

step D1: collecting and arranging running parameters and control parameters of a bridge floor buried pipe end, running parameters and control parameters of unit equipment, running parameters and control parameters of an energy pile bridge floor deicing system in the Jiangyun city, performing simple quality analysis and cleaning treatment on sample data, extracting a data distribution rule from the cleaned sample data, analyzing characteristics of the sample data, and distributing the sample data to corresponding calculation module clients.

The operation parameters of the buried pipe end of the bridge deck comprise: the system comprises a water collector, a bridge floor buried pipe, a water supply flow, a bridge floor buried pipe backwater flow, a bridge floor buried pipe water supply pressure and a bridge floor buried pipe backwater pressure, wherein the water collector is used for collecting water; the unit equipment operation parameters comprise: working liquid level of a condenser, inlet temperature of the condenser, working liquid level of an evaporator, inlet temperature of the evaporator and pressure of a compressor; the operating parameters of the energy pile pipe burying end respectively comprise: the soil temperature, the water level of the water collector at the buried pipe end of the energy pile, the pressure difference, the flow rate of a collecting pipe, the pressure, the temperature, the water supply temperature of the energy pile, the return water temperature, the flow rate and the water pressure.

The control parameters of the buried pipe end of the bridge deck comprise: the opening degree of a bridge floor buried pipe backwater control valve, the opening degree of a bridge floor buried pipe water supply control valve, the opening degree of a collector pipe control valve, the opening degree of a collector branch pipe control valve, the opening degree of a water distributor branch pipe control valve and the opening degree of a water replenishing pump control valve; the unit equipment control parameters comprise: the opening degree of a water replenishing pump control valve, the opening degree of a cooling circulating pump control valve, the opening degree of a freezing circulating pump control valve, the opening degree of an expansion valve and a host machine; the control parameters of the energy pile pipe burying end comprise: the device comprises an energy pile water collector manifold control valve opening, a water collector branch pipe control valve opening, an energy pile water distributor manifold control valve opening, a water distributor branch pipe control valve opening, an energy pile backwater control valve opening and a water supply control valve opening.

TABLE 1 training sample space and sample quality of operation control model of energy pile bridge floor deicing system in Jiangyin City

Step D2: the local models of 3 clients of the bridge deck buried pipe end control module, the unit equipment control module and the energy pile buried pipe end control module are respectively as follows:

local client 1 (energy pile pipe burying control module): fig. 2 is a partial CART decision tree model structure diagram of an optional energy pile operation and maintenance control parameter. And establishing a CART decision tree local model of energy pile pipe burying control parameters according to various operation parameters of the energy pile pipe burying end, unit equipment operation parameters updated by a central server and bridge deck pipe burying end operation parameters as sample attribute data, determining the heat taking amount of the energy pile and the energy pile pipe burying end control parameters, encrypting and uploading the heat taking amount and the energy pile pipe burying end control parameters to a central processing unit. Fig. 3 is a C4.5 decision tree flowchart of an energy pile operation and maintenance control parameter according to an embodiment, where the specific steps of the s-th round of the client are as follows:

step D2.1.1: and decrypting the public key of the encrypted transmission data by using a private key, updating the bridge floor buried pipe end control module, the unit equipment control module and the energy pile buried pipe end control module of the 3 local client sides according to the global sharing model of the central server, and taking an iteration result as a new sample data attribute parameter. Calculate each value s that attribute A might take for _i The coefficient of kini of (a).

Step D2.1.2: and selecting the attribute with the minimum Keyny coefficient and the corresponding attribute value as the optimal attribute and the optimal segmentation point from all possible attributes A and all attribute values corresponding to the attributes A. And generating two child nodes by using the optimal attribute and the optimal segmentation point, and distributing the corresponding samples to the nodes.

Step D2.1.3: and then, taking each root node as a complete data set, calling D2.1.1-D2.1.2 in an iterative manner, dividing samples according to a suboptimal attribute as a basis, taking samples with the same suboptimal attribute value as the same sample set to form leaf nodes, and sequentially iterating until a REP decision tree pruning condition or a stopping condition is met (the number of samples in the nodes and the coefficient of the kinship are less than a threshold value) to form a decision tree.

Step D2.1.4: and comparing the decided energy pile buried pipe end control parameters with an equipment control threshold, and if the calculation result meets the equipment control threshold, converting the decision calculation result serving as advanced prediction control information into encryption parameters by using a public key and uploading the encryption parameters to a central server for global sharing model iteration of federal learning. If the calculation exceeds the device control threshold, step D2.1.5 is performed.

Step D2.1.5: and outputting the decided control parameters of the energy pile pipe burying end and giving an alarm, taking a control threshold value as advanced prediction control information, converting the advanced prediction control information into encryption parameters by using a public key, and uploading the encryption parameters to a central server for global sharing model iteration of federal learning.

Local client 2 (bridge deck buried pipe end control module): fig. 4 is a diagram of a prediction model structure of an optional support vector machine for a bridge floor buried pipe end control module according to an embodiment of the present invention, where an operation parameter of a bridge floor buried pipe end, a unit equipment operation parameter updated by a central server, and an energy pile buried pipe end operation parameter are used as sample attribute data of a bridge floor buried pipe end control local model, the support vector machine local model of the bridge floor buried pipe end control module is established, and a control scheme is predicted, calculated, encrypted and uploaded to the central processor. FIG. 5 is a flow chart of a prediction model of a support vector machine of a bridge deck buried pipe end control module, and the concrete steps of the model in the s-th round are as follows:

step D2.2.1: and decrypting the public key of the encrypted transmission data by using a private key, updating relevant clients of the energy pile pipe burying end control module, the unit equipment control module and the bridge floor pipe burying end control module according to a global sharing model of the central server, and taking an iteration result as a new sample data attribute parameter. According to sample data, taking the environmental temperature, the environmental humidity, the wind speed, the snowfall amount, the surface snow-free rate, the water level of the water collector, the pressure difference, the temperature difference, the supply and return water temperature of the buried pipe of the bridge deck, the flow and the pressure as sample attributes X _i Forming sample data X = { X) including a plurality of features ₁ ,X ₂ ,…X _n With the relevant control parameters as learning target y = { y = ₁ ,y ₂ ,…y _n }。

Step D2.2.2: introducing a relaxation variable xi _i And constructing a nonlinear segmentation support vector classifier considering soft intervals by using the penalty coefficient C, and representing the relationship between the running state parameter of the bridge deck pipe burying end and the control valve parameter of the bridge deck pipe burying end. The prediction accuracy of the model and the self-stability of the model can be represented by a loss function (L = max (0, | z | - ∈)).

Step D2.2.3: and converting the conditions into a multivariate function by using a Lagrange function to solve, and enabling the partial derivative of the Lagrange function to the optimization target w, b and xi to be 0 to obtain a Lagrange multiplier, so that the original condition most-valued function is converted into a dual function, and the minimum value in the constraint area is found.

Step D2.2.4: inner product phi (X) of mapping function of dual function number _i ) ^T φ(X _j ) A "Sigmoid core: phi (X) _i ) ^T φ(X _j )＝κ(X _i ,X _j )＝tanh(αX _i ^T X _j + c) "is processed.

Step D2.2.5: optimizing support vector machine model parameters according to training set data: relaxation variable xi _i Penalty coefficient C, and hyper-parameters gamma, lambda, alpha, C and d in the kernel function. Preferably, a simulated annealing method is adopted for model parameter optimization, and fig. 6 is a flow chart of a local model of an improved support vector machine algorithm of an optional simulated annealing method, which includes the following steps:

step D2.2.5.1: randomly generating an initial parameter set for interactive verification, and recording an error value EEP as a current annealing system state E ₀ Initial temperature T ₀ The annealing end temperature is T ₁ 。

Step D2.2.5.2: according to a disturbance algorithm m' _i ＝m _i +s*(μ-0.5)(B _i -A _i ) Disturbing the parameters to form a new parameter set, and obtaining the state E of the current annealing system through interactive verification _n Calculate Δ E = E _n -E _n-1 。

Step D2.2.5.3: if the delta E is less than 0, a new parameter set is accepted, and the step is jumped to the step D2.2.5.5; otherwise, receiving the corresponding parameter set according to the Metropolis criterion exp (delta E/KT) -mu > 0, and jumping to the step D2.2.5.4; and if the above conditions are not met, refusing to accept the critical state, returning to the step D2.2.5.2, generating a new parameter set in a disturbance mode again, and carrying out interactive verification until the parameter set acceptance condition in the step D2.2.5.3 is met.

Step D2.2.5.4: set end temperature set to T ₁ As algorithm exit, global maximum computation errorThe difference EEP number is set to N. When T is reached ₁ Or N, stopping annealing, at which time critical state cross validation error E is accepted _n Should be the lowest E _n The corresponding parameter should be the best prediction parameter.

And D2.2.6, inputting target operation parameters of the bridge floor buried pipe, and predicting the adjusting parameters of the frozen water circulation loop at the end of the bridge floor buried pipe and the opening degree of a related control valve by using the trained support vector machine model.

Step D2.2.7: and comparing the calculation result with a control threshold of the chilled water circulation loop control valve, and if the calculation result meets the control threshold of the chilled water circulation loop control valve, converting the prediction calculation result serving as advanced prediction control information into encryption parameters by using a public key and uploading the encryption parameters to a central server for global sharing model iteration of federal learning. If the calculation exceeds the chilled water loop control valve control threshold, step D2.2.8 is executed.

Step D2.2.8: and outputting the calculated control result of the chilled water circulation loop, giving an alarm, taking a control threshold value as advanced prediction control information, converting the advanced prediction control information into encryption parameters by using a public key, and uploading the encryption parameters to a central server for global sharing model iteration of federal learning.

Local client 3 (unit device control module): fig. 7 is a flow chart of a random forest local model of an optional crew device control parameter according to an embodiment of the present invention. And establishing a random forest local model of the control parameters of the unit equipment according to various operation parameters of the unit equipment, the energy pile operation parameters updated by the central server and the operation parameters of the buried pipe end of the bridge floor as sample attribute data, determining the control parameters of the unit equipment, and encrypting and uploading the control parameters to the central processor. Fig. 8 is a flowchart of a random forest local model according to an embodiment of the present invention, which includes the following specific steps:

step D2.3.1: and decrypting the public key of the encrypted transmission data by using a private key, and using the energy pile operation parameter updated according to the global sharing model of the central server and the bridge floor buried pipe end operation parameter as a new sample data attribute parameter. According to the local sample data, each state parameter of each unit device is used as a sample attribute space set, and a unit device control parameter is used as a learning target. And randomly generating a sample subset from the samples by using a Bootstrap sampling method, taking the sample subset as a training sample of one decision tree model, and repeatedly sampling for k times to form k decision tree training samples.

Step D2.3.2: and performing decision tree training according to the attribute subsets in the k training samples to form k mutually independent random decision trees.

Step D2.3.3: and voting the control schemes of the unit equipment predicted by the k decision trees, wherein the voting result is used as the optimal control scheme of the unit equipment.

Step D2.3.4: and comparing the determined control parameters of the unit equipment (an evaporator, a condenser, a cooling pump, a freezing pump and an expansion valve) with the control threshold of the unit equipment, and if the calculation result meets the control threshold of the unit equipment, converting the decision calculation result serving as advanced prediction control information into encryption parameters by using a public key and uploading the encryption parameters to a central server for global sharing model iteration of federal learning. And if the calculation result exceeds the unit equipment control threshold value, executing a step D2.3.5.

Step D2.3.5: and outputting the decided unit equipment control result and giving an alarm, and converting the unit equipment control threshold value as advanced prediction control information into encryption parameters by using a public key and uploading the encryption parameters to a central server for global sharing model iteration of federal learning.

And D3: local model results of 3 local clients of the bridge deck pipe burying end control module, the unit equipment control module and the energy pile control module are encrypted into desensitization parameters in a public key mode and uploaded to a central server.

Step D4: the server decrypts the encryption desensitization parameters uploaded by the 3 clients by using a private key, performs federate learning aggregation on the decryption parameters of all client models based on knowledge distillation, and updates the global model weights of the M clients according to the following formula:

wherein W _s+1 Is the global model parameter for the s-th round,

and uploading the client sub-model weight of the server for the s-th client i. After each round of model weight updating, the central server calculates the error and accuracy of the global sharing model.

Step D5: and the central server generates a public key for encrypting the transmission data by using the global sharing model and sends the public key to each client. And updating relevant client iteration results required by local model training according to each local client of the global sharing model to serve as new sample data attribute parameters.

Step D6: and D2-D5 are repeated for continuous iteration until the global sharing model is stable, and finally the client calculates according to the global sharing model to obtain a corresponding result. The control states matched with each other in the active bridge deck deicing and snow melting system of a certain energy pile in Jiangyun city are calculated, and the control states comprise the opening degrees of various regulating valves of a chilled water circulation loop at the end of a buried pipe of the bridge deck, the snow melting time consumption at the end of the buried pipe of the bridge deck, the opening degrees of various regulating valves at the equipment end of a unit and the opening degrees of various regulating valves of a cooling water circulation loop at the end of a buried pipe of the energy pile, and are shown in table 2.

TABLE 2 control parameter results for deicing system for pile and bridge surfaces of certain energy in Jiangyun city

The embodiment of the invention provides a heat exchange tube overhauling device based on an electromagnetic induction principle. The specific embodiment is as follows.

Fig. 10 is a schematic diagram of a heat exchange pipe inspection and repair device based on the electromagnetic induction principle. The maintenance device mainly comprises a double-layer intelligent heat exchange tube structure 1, a miniature magnetoelectric sensor 3 containing a dredging module, a cable storage disc 4, an automatic data acquisition instrument 5 and a control end 6. The detection principle is as follows: the magnetoelectric sensors 3-6 are connected with electric signals, when the magnetoelectric sensors 3-6 pass through the magnetic rings, the electromagnetic frequency will change, the electric signals are transmitted to the automatic data acquisition instrument 5 through the connected cables, the automatic data acquisition instrument 5 automatically records the time when the sensors pass through each magnetic ring in sequence and transmits the information to the control end 6, and if no electric signal transmission of the next measuring point exists for a long time after passing through a certain measuring point, the possibility of blockage between the two measuring points is judged.

As shown in fig. 11 and 12, the double-layer intelligent heat exchange tube structure 1 includes an inner polyethylene tube 1-a, an outer polyethylene tube 1-B, and a magnetic induction ring 2. The magnetic induction rings 2 are embedded in the pipe walls of the outer polyethylene pipes 1-B in a hot melting mode, the ring width is 5-8 mm, the magnetic induction rings and the outer polyethylene pipes 1-B are integrally formed, the magnetic induction rings are arranged at equal intervals along the axial direction of the heat exchange pipe, the interval setting is determined according to the design requirements of the energy piles and the bridge deck pavement layer, and the interval setting is 0.3-0.5 m, namely 0.5m. Threading the inner wall of the outer polyethylene pipe 1-B and the outer wall of the inner polyethylene pipe 1-A, smearing lubricating oil between the two polyethylene pipes, and screwing the two polyethylene pipes by utilizing the threads. Fig. 12 is a schematic cross-sectional view of a pipeline of the double-layer intelligent heat exchange tube structure of fig. 11.

The structure of the miniature magnetoelectric sensor 3 with the dredging module is schematically shown in fig. 13, and fig. 14 is an A-A cross section in fig. 13, and comprises the dredging module,base:Sub>A monitoring module, magnetoelectric sensors 3-6 andbase:Sub>A moving module. The dredging module comprises a spiral metal wire 3-1 or/and a hot melting rod, a connecting component 3-2, a gear 3-3 and a series excitation motor 3-5; the monitoring module comprises a surrounding camera 3-4 and an illuminating lamp 3-9, and the moving module comprises a driving wheel 3-7 and a corresponding motor.

The embodiment of the invention provides a method for detecting blockage and damage of a heat exchange pipeline by using a heat exchange pipe overhauling device based on an electromagnetic induction principle, which comprises the following steps:

step S1: after the energy pile bridge surface deicing and snow melting federal learning operation control gives out a failure preliminary alarm, firstly, a power switch 5 of a collecting instrument is turned on, a magnetic induction ring 2 is used for covering a magnetoelectric sensor 3-6 to move, when the magnetic induction ring 2 meets an induction point of the magnetoelectric sensor 3-6, an acousto-optic alarm is given out, and meanwhile, an instrument has an indication to indicate that the collecting instrument works normally;

step S2: during testing, the miniature magnetoelectric sensor 3 with the dredging module is placed into the magnetoelectric sensor 3-6 along a pipeline, the pipeline arranged in the vertical direction in the energy pile is slowly placed into the energy pile by utilizing the gravity action of the magnetoelectric sensor 3-6 and the rotation matching of the cable storage disc 4; a pipeline laid in the horizontal direction in the bridge deck pavement layer is used for starting a motor corresponding to the driving wheel 3-7 by utilizing a moving module of the miniature magnetoelectric sensor, so that the miniature magnetoelectric sensor 3-6 moves in the pipeline. When the center of the magnetoelectric sensor 3-6 intersects with the magnetic induction ring 2, the instrument sends out a buzzer sound accompanied with light indication. Meanwhile, the surrounding cameras 3-4 and the illuminating lamps 3-9 are started and transmitted to the control end 6 in real time, so that the cameras are aligned to the blocked positions. The data automatic acquisition instrument 5 automatically records the number of each magnetic induction ring measuring point and the interval time recording, real-time image and sound wave waveform information of two adjacent measuring points.

And step S3: the automatic data acquisition instrument 5 transmits the automatically recorded data to the control end 6, the control end 6 can analyze the acquired data by adopting an isolated forest algorithm and accurately position a damaged point, and the position of a blockage point is accurately positioned according to the analyzed abnormal time and position of the data.

And analyzing the real-time data by adopting an isolated forest algorithm. An isolated tree is established for an attribute space in a sample set to isolate attributes, in the same attribute space, the attribute values of abnormal values are relatively distant from the attribute values of most samples, the abnormal values are separated earlier through model training, normal values are located in deeper child nodes of the isolated tree, and the algorithm structure is shown in fig. 14. The method comprises the following specific steps:

step S3.1: the current signal, the sound wave signal and the electromagnetic signal which are transmitted to the control center on the arrangement dredging module are used as sample attributes to form a sample set which can be expressed as X = (X) ₁ ,x ₂ ,x ₃ ,…,x _n ) Wherein x is _n ＝(x _n1 ,x _n2 ,x _n3 ,…,x _nm )，x _nm Is the nth type attribute signal at the mth moment. Step S3.2: and segmenting the attribute type by taking the attribute dimension q as an attribute cutting point to form different isolated trees, and segmenting the sample space by taking the attribute value cutting point p as an abnormal value cutting point segmentation attribute value for each isolated tree. Q is a random value falling within the range (0, n) and p is a random value falling within min (x) _n )＜p＜max(x _n ) Random value of interval.

Step S3.3: and forming a hyperplane in the sample space according to the segmentation of the attribute cut points and the abnormal value cut points, classifying the samples smaller than the cut points into one class, forming a left node of the isolated tree, and otherwise forming a right node.

Step S3.4: and (4) repeating the step (S3.2) to the step (S3.3) until the leaf node of each isolated tree only contains one sample, and finishing the training of the isolated tree model.

Step S3.5: sampling data x at each moment _i Putting the sample into an isolated tree model for traversal calculation, and calculating to obtain the average height h (x) of each sample data _i ) And an abnormality score S (x, ψ) of the sample data. The abnormal signal is judged according to the following table.

TABLE 3 abnormal value judgment index

S (x, psi) value	Signal evaluation
		＞0.75	Abnormal signal
0.4～0.75	Suspected abnormal signal
		＜0.4	Normal signal

The embodiment of the invention provides a method for dredging and repairing a heat exchange pipeline by using a heat exchange pipe overhauling device based on an electromagnetic induction principle, which comprises the following specific steps of:

step T1: and (3) starting the surrounding cameras 3-4 and the illuminating lamps 3-9, and transmitting the surrounding cameras and the illuminating lamps to the control end 6 in real time to enable the cameras to be aligned to the blocked position.

Step T2: starting dredging work, if the blockage is found to be caused by hot melting of the pipeline, starting a series motor 3-5 for controlling a spiral stainless steel wire 3-1, after the spiral metal wire 3-1 rotates to a working rotating speed, if the sensor cable is slowly released by continuously utilizing the self gravity of a magnetoelectric sensor 3-6 in the pipeline in the vertical direction, if the cable is slowly released by utilizing a moving module to control the walking process of the magnetoelectric sensor 3-6 in the pipeline in the horizontal direction, and stirring the blocked foreign matters by the spiral metal wire 3-1 until the foreign matters are removed; if the blockage is caused by pipeline hot melting, the spiral metal wire 3-1 and the connecting component 3-2 of the dredging module can be replaced by a hot melting rod, the magnetoelectric sensor 3-6 is placed at the blockage again, the heating rod is started to reach the hot melting temperature of the polyethylene material, and the sensor cable is slowly placed by utilizing the self gravity of the sensor until the blockage is dredged.

And step T3: after the miniature magnetoelectric sensor is taken out of the pipeline, pipeline pressure testing and flushing work are carried out, residual foreign matters in the pipeline are cleaned, and dredging work is completed.

And step T4: after the intelligent heat exchange tube damage point is accurately positioned, because the lubricating oil for prefabricating the thread and reducing the friction force exists between the double-layer intelligent heat exchange tube structures, the tube section where the damage point is located can be screwed out along the thread to achieve the purpose of replacing the damaged tube section.

Claims (10)

1. A heat exchange tube overhauling and intelligent operation and maintenance control method of an energy pile active bridge deck deicing and snow melting system is characterized by comprising an energy pile active bridge deck deicing and snow melting federal learning operation and maintenance control model and a heat exchange tube overhauling device based on an electromagnetic induction principle, wherein the energy pile active bridge deck deicing and snow melting federal learning operation and maintenance control model is formed by a central server-local client; performing preliminary judgment and preliminary alarm positioning according to an energy pile active bridge deck deicing and snow melting federal learning operation and maintenance control model formed by a central server and a local client, and arranging a heat exchange tube overhauling device based on an electromagnetic induction principle to perform fine repairing operation on the basis;

the active bridge deck deicing and snow melting federal learning operation and maintenance control model of the energy pile formed by the central server and the local clients comprises 3 local client modules: the device comprises a bridge floor pipe burying end control module, a unit equipment control module and an energy pile pipe burying end control module; the local models of the local modules of the 3 local clients are subjected to learning training on the local clients, then calculation results are encrypted and uploaded to a central processing unit to pass through a global federal learning model, and results obtained by the global federal learning model are issued to the local models of the local clients again; through repeated iteration between the client and the central server until the global federal learning model is stable, different module regulation and control schemes matched with each other are finally made;

the method comprises the following specific steps:

d1: acquiring bridge deck pipe burying end operation parameters, bridge deck pipe burying end control parameters, unit equipment operation parameters, unit equipment control parameters, energy pile pipe burying end operation parameters and energy pile pipe burying end control parameters, performing sample data quality analysis and cleaning treatment, extracting a data distribution rule for the cleaned sample data, analyzing sample data characteristics, and distributing the sample data characteristics to corresponding calculation module client sides;

d2: update the local model of 3 local clients: a bridge deck buried pipe end control support vector machine local model, a unit equipment control random forest local model and an energy pile buried pipe end control decision tree local model;

d3: encrypting the local model results of the 3 local clients into desensitization parameters in a public key mode, and uploading the desensitization parameters to a central server;

d4: the central server decrypts the encryption desensitization parameters uploaded by the 3 local clients by using a private key, performs decoding operation on the encryption desensitization parameters, performs security aggregation, and then updates a global federated learning model; updating a global federal learning model by federal learning at the central server side according to one or more synchronous modes of gradient averaging, federal averaging and knowledge distillation; after the weight of the global federal learning model is updated in each round, the central server calculates the error and the accuracy of the global federal learning model;

d5: the central server generates a public key for encrypting transmission data from the global federal learning model and sends the public key to each client; updating iteration results of other clients by each local client as new sample data attribute parameters according to the global federated learning model;

d6: repeating the steps D2-D5 for continuous iteration until the global federal learning model is stable, and finally calculating by the client according to the global federal learning model to obtain a corresponding result; and calculating the control states matched with each other in the active bridge deck deicing and snow melting system of the energy piles, wherein the control states comprise the opening degrees of various regulating valves of a frozen water circulation loop at the end of the buried pipe of the bridge deck, the snow melting time consumption of the end of the buried pipe of the bridge deck, unit control parameters and energy pile buried pipe end control parameters.

2. The method for overhauling and intelligently controlling the operation and the maintenance of the heat exchange tube of the energy pile active bridge deck deicing and snow melting system according to claim 1, wherein the energy pile buried tube control module is used for establishing a decision tree local model of the energy pile buried tube control parameters according to the running parameters of the energy pile buried tube end, the running parameters of the unit equipment updated by the central server and the running parameters of the bridge deck buried tube end as sample attribute data; determining the heat taking quantity of the energy pile and control parameters of the buried pipe end of the energy pile, encrypting and uploading the parameters to a central processing unit; the decision tree model is formed by one or more of the following ID3 algorithm, C4.5 algorithm and CART algorithm; the method comprises the following specific steps:

d2.1.1: decrypting the public key of the encrypted transmission data by using a private key, updating a bridge floor buried pipe end control module, a unit equipment control module and an energy pile buried pipe end control module of 3 local client sides according to a global federal learning model of a central server, and taking an iteration result as a new sample data attribute parameter; calculating all attribute values A corresponding to any attribute A _i The coefficient of kini of (a);

d2.1.2: selecting the attribute with the minimum Gini coefficient and the corresponding attribute value as the optimal attribute and the optimal segmentation point from all the attributes A and all the attribute values corresponding to the attributes A; generating two child nodes by using the optimal attribute and the optimal segmentation point, and distributing corresponding samples to the root nodes;

d2.1.3: taking each root node as a complete data set, iteratively calling the step D2.1.1-the step D2.1.2, dividing samples according to a suboptimal attribute as a basis, taking the samples with the same suboptimal attribute value as the same sample set to form leaf nodes, and sequentially iterating by adopting a REP method, a PEP method or an MEP method until a decision tree pruning condition is met and a stop condition that the number of samples in the leaf nodes or the kiney coefficient is less than a threshold value is reached to form a decision tree;

d2.1.4: comparing the decided control parameters of the energy pile pipe burying end with an equipment control threshold, and when the calculation result meets the equipment control threshold, converting the decision calculation result into encryption parameters by using a public key as advanced prediction control information and uploading the encryption parameters to a central server for global federal learning model iteration; when the calculation result exceeds the device control threshold, executing a step D2.1.5;

d2.1.5: outputting the decided control parameters of the energy pile pipe burying end and giving an alarm, converting the control parameters into encryption parameters by using a public key with a control threshold as advance prediction control information and uploading the encryption parameters to a central server for global federal learning model iteration; arranging a heat exchange tube overhauling device based on an electromagnetic induction principle to carry out refined detection and positioning;

the bridge floor buried pipe end control module takes the running parameters of the bridge floor buried pipe end, the unit equipment running parameters updated by the central server and the running parameters of the energy pile buried pipe end as sample attribute data of a bridge floor buried pipe end control local model; establishing a bridge deck buried pipe end control support vector machine local model, predicting a calculation control scheme, encrypting and uploading to a central server; the method comprises the following specific steps:

d2.2.1: decrypting the public key of the encrypted transmission data by using a private key, updating the energy pile buried pipe end control module, the unit equipment control module and the bridge floor buried pipe end control module client according to a global federal learning model of a central server, and taking an iteration result as a new sample data attribute parameter; taking the environmental temperature, the environmental humidity, the wind speed, the snow fall amount, the snow-free rate of the surface, the water level, the pressure difference and the temperature difference of a water collector, the temperature, the flow and the pressure of the supply and return water of the buried pipe of the bridge deck as a sample attribute space set, and taking control parameters of the buried pipe end of the bridge deck as a learning target;

d2.2.2: introducing relaxation variables xi _i And penalty factor C construction considerationsThe soft-spaced nonlinear segmentation support vector classifier represents the relationship between the running parameters of the end of the bridge deck buried pipe and the control parameters of the end of the bridge deck buried pipe; the prediction precision and the self stability of the bridge deck buried pipe end control support vector machine local model are represented by a loss function L = max (0, | z | - ∈); converting the loss function into a condition maximum function;

wherein w is a hyperplane normal vector, C is a penalty coefficient, and xi are ^* Is a relaxation factor, and belongs to a hyper-parameter for determining the width of the boundary; y is _i The actual measurement result of the training sample is obtained; f (X) _i ) = w.phi (X) + b is a classification hyperplane of a bridge deck buried pipe end control support vector machine local model; i is the training sample number; n is the number of training set samples; phi (X) is a nonlinear mapping function;

d2.2.3: converting the condition most-valued function into a multivariate function through a Lagrange function to solve, making the partial derivative of the Lagrange function to the optimization target w, b and xi be 0 to obtain a Lagrange multiplier, and converting the condition most-valued function into a dual function so as to find the minimum value of the prediction boundary;

d2.2.4: nonlinear mapping function phi (X) contained in classification hyperplane in bridge floor buried pipe end control support vector machine local model, and inner product phi (X) of nonlinear mapping function phi (X) _i ) ^T φ(X _j ) Selecting one or more than one kernel functions of a Gaussian kernel, a linear kernel, a polynomial kernel and a Sigmoid kernel for combined processing;

d2.2.5: optimizing model parameters in a local model of a bridge deck buried pipe end control support vector machine: insensitive loss function E, penalty coefficient C, and hyper-parameters gamma, lambda, alpha, C and d in the kernel function; selecting one or more optimization methods of a simulated annealing method, a grid search method, a particle swarm optimization method, a PSO algorithm and a genetic algorithm;

d2.2.6: inputting target operation parameters of the bridge deck buried pipe, and predicting and calculating the bridge deck buried pipe end by using a trained bridge deck buried pipe end control support vector machine local model;

d2.2.7: comparing the calculation result with a control threshold of the chilled water circulation loop control valve, and when the calculation result meets the control threshold of the chilled water circulation loop control valve, taking the prediction calculation result as advanced prediction control information, converting the advanced prediction control information into encryption parameters by using a public key, and uploading the encryption parameters to a central server for global federal learning model iteration; when the calculation result exceeds the control threshold value of the chilled water circulation loop control valve, executing a step D2.2.8;

d2.2.8: outputting the calculated control result of the buried pipe end of the bridge deck and giving an alarm, taking an equipment control threshold value as advanced prediction control information, converting the advanced prediction control information into encryption parameters by using a public key, and uploading the encryption parameters to a central server for iteration of a global federal learning model; meanwhile, arranging a heat exchange tube overhauling device based on the electromagnetic induction principle to carry out fine detection and positioning;

the unit equipment control module is used for establishing a random forest local model of unit equipment control parameters according to various operation parameters of the unit equipment, energy pile operation parameters updated by a central server and bridge floor buried pipe end operation parameters as sample attribute data, determining the unit equipment control parameters and encrypting and uploading the unit equipment control parameters to the central processor; the method comprises the following specific steps:

d2.3.1: decrypting the public key of the encrypted transmission data by using a private key, and using the energy pile running parameter updated according to the global federal learning model of the central server and the running parameter of the buried pipe end of the bridge deck as a new sample data attribute parameter; according to local sample data, taking each state parameter of each unit device as a sample attribute space set, and taking a unit device control parameter as a learning target; randomly generating a sample subset from a unit equipment control training sample by using a Bootstrap sampling method, taking the sample subset as a training sample of one decision tree model, and repeatedly sampling for k times to form k decision tree training samples;

d2.3.2: performing decision tree training according to the attribute subsets in the k training samples to form k mutually independent random decision trees;

d2.3.3: voting the unit equipment control schemes predicted by the k decision trees by a random selection classifier of a random forest, and taking the voting result as the optimal control scheme of the unit equipment;

d2.3.4: comparing the decided unit equipment control parameters with a unit equipment control threshold, and when the calculation result meets the unit equipment control threshold, converting the decision calculation result into encryption parameters by using a public key as advanced prediction control information and uploading the encryption parameters to a central server for global federal learning model iteration; when the calculation result exceeds the control threshold value of the unit equipment, executing a step D2.3.5;

d2.3.5: and outputting the decided unit equipment control result and giving an alarm, and converting the unit equipment control threshold value as advanced prediction control information into an encryption parameter by using a public key and uploading the encryption parameter to a central server for global federal learning model iteration.

3. The method for overhauling and intelligently controlling the operation and the maintenance of the heat exchange tube of the active energy pile bridge deck deicing and snow melting system according to claim 2, wherein when a simulated annealing method is adopted for optimization in the step D2.2.5, the method comprises the following steps;

d2.2.5.1: randomly generating an initial parameter set for interactive verification, and recording an error value EEP as a current annealing system state E ₀ Initial temperature T ₀ Annealing end temperature of T ₁ ；

D2.2.5.2: m 'according to a perturbation algorithm' _i ＝m _i +s·(μ-0.5)(B _i -A _i ) Disturbing the parameters to form a new parameter set, and obtaining the state E of the current annealing system through interactive verification _n Calculate Δ E = E _n -E _n-1 ；

Wherein: m' _i For the perturbed variable, m _i For the current variable, s is the perturbation proportion, μ is [0,1 ]]Random number of (2), B _i 、A _i Is the current variable m _i A range of (d);

d2.2.5.3: when the delta E is less than 0, a new parameter set is received, and the step is jumped to the step D2.2.5.5; otherwise, receiving the corresponding parameter set according to Metropolis criterion exp (delta E/KT) -mu > 0, and jumping to the step D2.2.5.4; if the conditions are not met, refusing to accept the critical state, returning to the step D2.2.5.2, generating a new parameter set again by disturbance, and carrying out interactive verification until the parameter set acceptance condition in the step D2.2.5.3 is met;

d2.2.5.4: setting the end temperature to T ₁ Setting the number of times of the global maximum calculation error value EEP as N; when T is reached ₁ Or stopping annealing when N is reached, otherwise returning to the step D2.2.5.2; critical state cross validation error accepted at stop anneal E _n At the lowest, the corresponding parameter is the best prediction parameter.

4. The heat exchange tube overhauling and intelligent operation and maintenance control method of the energy pile active bridge deck deicing and snow melting system is characterized in that the heat exchange tube overhauling device based on the electromagnetic induction principle comprises a double-layer intelligent heat exchange tube structure (1), a miniature magnetoelectric sensor (3) with a dredging module, a cable storage disc (4), an automatic data acquisition instrument (5) and a control end (6); the double-layer intelligent heat exchange tube structure (1) is U-shaped or snakelike, the U-shaped double-layer intelligent heat exchange tube structure (1) is arranged in the energy pile, and the snakelike double-layer intelligent heat exchange tube structure (1) is arranged in the bridge floor; two ends of the double-layer intelligent heat exchange tube structure (1) are respectively connected with an automatic data acquisition instrument (5) through a cable storage disc (4), and the automatic data acquisition instrument (5) is connected with a control end (6); the double-layer intelligent heat exchange tube structure (1) comprises an inner polyethylene tube (1-A), an outer polyethylene tube (1-B) and a magnetic induction ring (2); the inner polyethylene pipe (1-A) is positioned in the inner wall of the outer polyethylene pipe (1-B), and lubricating oil is filled between the inner polyethylene pipe and the outer polyethylene pipe and is in threaded connection with the inner polyethylene pipe; the magnetic induction rings (2) are embedded in the pipe wall of the outer polyethylene pipe (1-B) and are arranged at equal intervals along the axial direction of the double-layer intelligent heat exchange pipe structure (1); a miniature magnetoelectric sensor (3) containing a dredging module moves in an inner polyethylene pipe (1-A), and comprises a dredging piece, a series excitation motor (3-5), a driving wheel (3-7) and a cable (3-8); one end of the series excited motor (3-5) is connected with the cable (3-8), and the other end of the series excited motor is fixedly connected with the dredging piece through the connecting component (3-2) by mounting the gear (3-3); the side surface of the series excited motor (3-5) is connected with a plurality of driving wheels (3-7) to slide on the inner wall of the inner polyethylene pipe (1-A); the magnetoelectric sensor (3-6), the surrounding camera (3-4) and the illuminating lamp (3-9) are all arranged on the surface of the series excitation motor (3-5); the magnetoelectric sensor (3-6) is introduced with an electric signal, when the magnetoelectric sensor passes through the magnetic induction ring (2), the electromagnetic frequency changes, and the electric signal is transmitted to the automatic data acquisition instrument (5) through the connected cable (3-8); the data automatic acquisition instrument (5) automatically records the time when the magnetoelectric sensors (3-6) pass through each magnetic induction ring (2) in sequence, transmits the information to the control end (6), and judges that the blockage exists between two magnetic induction rings (2) when the time of passing through a certain magnetic induction ring (2) exceeds the set time and no electric signal transmission of the next magnetic induction ring (2) is available.

5. The method for overhauling and intelligently controlling the operation and the maintenance of the heat exchange tube of the active energy pile bridge deck deicing and snow melting system according to claim 4, wherein the magnetic induction rings (2) are 5-8 mm wide and are integrally formed with the outer polyethylene tube (1-B); the distance between the magnetic induction rings (2) is 0.3-0.5 m; the dredging piece is a spiral metal wire (3-1) and/or a hot melting rod.

6. The method for overhauling and intelligently controlling the operation and the maintenance of the heat exchange tube of the energy pile active bridge deck deicing and snow melting system as claimed in claim 4 or 5, wherein the heat exchange tube overhauling device based on the electromagnetic induction principle is used for detecting blockage and damage of a heat exchange pipeline, and comprises the following steps:

s1: after the initial failure alarm is given out by the federal learning operation control of deicing and snow melting of the bridge surface of the energy pile, a switch of the automatic data acquisition instrument (5) is turned on, a magnetic induction ring is sleeved on the magnetoelectric sensors (3-6) to move, when the magnetic induction ring meets the induction points of the magnetoelectric sensors (3-6), an audible and visual alarm is given out, meanwhile, the automatic data acquisition instrument (5) has an indication, and the automatic data acquisition instrument (5) normally works to perform the subsequent steps;

s2: during testing, pipelines are arranged in the energy piles in the vertical direction, and the magnetoelectric sensors (3-6) are rotationally placed in the cable storage disc (4) by utilizing the gravity of the magnetoelectric sensors; pipelines are arranged in the bridge deck pavement layer in the horizontal direction, and the magnetoelectric sensors (3-6) move in the pipelines through driving wheels (3-7) and comprise miniature magnetoelectric sensors (3) with dredging modules; when the center of the magnetoelectric sensor (3-6) is intersected with the magnetic induction ring (2), the automatic data acquisition instrument (5) sends out a buzzer sound accompanied with light indication; opening a surrounding camera (3-4) to perform image real-time monitoring and sound wave monitoring in a pipeline, and automatically recording the number of the measuring points passing through each magnetic induction ring (2) and the interval time recording, real-time image and sound wave waveform information of the measuring points passing through two adjacent magnetic induction rings (2) by a data automatic acquisition instrument (5);

s3: the automatic data acquisition instrument (5) transmits the automatically recorded data to the control end (6), and one or more of the following methods are adopted to analyze the acquired data and accurately position the damaged point: isolated forest algorithm, X-ray digital imaging technology, ultrasonic detection, ultrasonic guided wave, ultrasonic C scanning, ultrasonic phased array and high-temperature thickness measurement method; and the control end (6) accurately positions the position of the blockage point according to the analyzed abnormal time and position of the data.

7. The method for overhauling the heat exchange tube and intelligently controlling the operation and the maintenance of the energy pile active bridge deck deicing and snow melting system is characterized in that when an isolated forest algorithm is adopted in the step S3 for analyzing the collected data, an isolated tree is established for isolating the attributes of the attribute space in the sample set, abnormal values are separated through model training in the same attribute space, and the normal values are located at deep sub-nodes of the isolated tree; the method comprises the following specific steps:

s3.1: the current signal, the sound wave signal and the electromagnetic signal transmitted to the control end (6) on the magnetoelectric sensors (3-6) are used as sample attributes to form a sample set, and the sample set is represented as X = (X) ₁ ,x ₂ ,x ₃ ,…,x _n ) Wherein x is _n ＝(x _n1 ,x _n2 ,x _n3 ,…,x _nm )，x _nm The nth type attribute signal at the mth moment;

s3.2: segmenting the attribute types by using the attribute dimension q as an attribute cutting point to form different isolated trees, and segmenting the attribute values in the attribute space by using the attribute value cutting point p as an abnormal value cutting point for each isolated tree; q is a random value falling within the range (0, n), and p is a random value falling within min (x) _n )＜p＜max(x _n ) A random value of the interval;

s3.3: forming a hyperplane in a sample space according to the segmentation of the attribute cutting point q and the abnormal value cutting point p, classifying samples smaller than the abnormal value cutting point p into one class to form a left node of the isolated tree, and otherwise forming a right node;

s3.4: repeating the step S3.2 to the step S3.3 until the leaf node of each isolated tree only contains one sample, and finishing the training of the isolated tree model;

s3.5: sampling data x at each moment _i Putting the sample into an isolated tree model for traversal calculation, and calculating to obtain the average height h (x) of each sample data _i ) Computing an anomaly score for the sample data according to:

wherein h (x) is the height of x in each isolated tree, c (n) is the average value of the path length at a given number of samples n, and E (h (x)) is the expected path length value of x in a plurality of isolated trees;

when the score S (x, psi) > 0.75, the signal corresponding to the sample at the moment is regarded as an abnormal signal;

when the score S (x, psi) is 0.4-0.75, the signal corresponding to the time sample is regarded as a suspected abnormal signal;

when the score S (x, ψ) < 0.4, the sample at that time is regarded as a normal signal.

8. The method for overhauling and intelligently controlling the operation and the maintenance of the heat exchange tube of the energy pile active bridge deck deicing and snow melting system according to claim 4 or 5, wherein the heat exchange tube overhauling device based on the electromagnetic induction principle is used for dredging and repairing a heat exchange pipeline and comprises the following steps:

t1: starting the surrounding cameras (3-4) and the illuminating lamps (3-9), and transmitting the surrounding cameras and the illuminating lamps to the control end (6) in real time to enable the surrounding cameras (3-4) to be aligned to the blockage;

t2: starting dredging work, starting a series excitation motor (3-5) for controlling the spiral metal wire (3-1) when a blockage except for blockage caused by pipeline hot melting is found, and rotating the spiral metal wire (3-1) to a working rotating speed; slowly putting down a sensor cable (3-8) in a vertical pipeline by using the self gravity of the magnetoelectric sensor (3-6); controlling the magnetoelectric sensor (3-6) to move by using a driving wheel (3-7) in the pipeline in the horizontal direction, and slowly lowering the sensor cable (3-8); the spiral metal wire (3-1) stirs the blocked foreign matters until the foreign matters are removed; when the blockage is caused by pipeline hot melting, the spiral metal wire (3-1) and the connecting member (3-2) of the dredging module are replaced by a hot melting rod, the magnetoelectric sensor (3-6) is placed at the blockage again, the heating rod is started to reach the hot melting temperature of the polyethylene material, and the sensor cable (3-8) is slowly lowered by utilizing the self gravity of the magnetoelectric sensor (3-6) until the blockage is dredged;

t3: taking out the miniature magnetoelectric sensor (3) containing the dredging module from the pipeline, carrying out pipeline pressure test and flushing work, cleaning foreign matters remained in the pipeline, and finishing the dredging work;

t4: and after the damaged point is accurately positioned, repairing the heat exchange pipeline, wherein the pipeline section where the damaged point is located is screwed out along the thread to replace the damaged pipeline section or an automatic repairing agent is adopted to repair the damaged section.

9. The method for overhauling and intelligently controlling the operation and the maintenance of the heat exchange tube of the energy pile active bridge deck deicing and snow melting system is characterized in that the automatic repairing agent is one or more of polyethylene powder, polyethylene particles, nylon fiber yarns, ethylene glycol and an adhesive; when the automatic mending agent is nylon fiber yarn, the fineness is 200-300D, and the length of the nylon fiber yarn is 3-5 mm.

10. The method for overhauling and intelligently controlling the operation and the maintenance of the heat exchange tube of the energy pile active bridge deck deicing and snow melting system is characterized in that the repairing process and the method of the automatic repairing agent are as follows:

p1: after the damaged point is positioned, plugging any one end of the inlet and outlet of the pipeline under the pressure test standard, and injecting an automatic repairing agent into the pipeline at the other pipeline port;

p2: after the mending agent is injected, the mending agent is pressed into the pipe, but the pressure is not more than the rated pressure which can be born by the adopted polyethylene pipe;

p3: when the reading of the pressure gauge is stable, primarily completing repair of the damaged point; and after the automatic repairing agent completely meets the strength requirement, carrying out pressure test on the pipeline, and regarding the repairing work as the completion of the pressure stabilization within 15-20 min under the test pressure, wherein the pressure drop after the pressure stabilization is 1% -3%, and no leakage phenomenon exists.

Images