Search

CN-122007351-A - Multi-sensor large casting pouring monitoring method and system

CN122007351ACN 122007351 ACN122007351 ACN 122007351ACN-122007351-A

Abstract

The application relates to the technical field of casting process monitoring, in particular to a multi-sensor large casting pouring monitoring method and system, and aims to solve the problems of data loss caused by sensor dead zones and response delay caused by distortion and centralization of a pure data interpolation model under a dynamic working condition. The system comprises a physical field sensing network, a marginal side physical field reconstruction unit, a multi-level decision unit and an actuator control unit, wherein the global physical field is reconstructed in real time by unevenly deploying sensors and fusing a simplified physical mechanism model, and the rapid identification and regulation of defect risks are realized based on a layered decision mechanism. The scheme can improve the data effectiveness, ensure the reconstruction precision, shorten the system response delay and meet the real-time monitoring requirement of the pouring process of the large casting.

Inventors

  • LI YUNLONG
  • DAI LICHANG
  • LI ZHAOSONG

Assignees

  • 玉溪锦福智能设备有限公司

Dates

Publication Date
20260512
Application Date
20260123

Claims (10)

  1. 1. A multi-sensor large casting pouring monitoring system, comprising: The physical field sensing network is used for collecting multisource physical quantity data in the casting process and comprises a plurality of sensor groups which are arranged on the outer wall of a casting mould, a casting system and preset internal detection points, wherein each sensor group comprises a thermocouple temperature sensor group, a vibration acceleration sensor group and a magnetic field intensity sensor group, the thermocouple temperature sensor groups are arranged according to a non-uniform topological structure, the deployment positions of all measuring points are determined through computational fluid mechanics simulation of the casting process performed in advance, and a high-density deployment area is extracted; The edge side physical field reconstruction unit is operated in industrial edge computing equipment deployed in an adjacent area of a pouring site and is used for receiving real-time data streams from a physical field sensing network, and executing a dynamic physical field reconstruction algorithm based on physical mechanism guidance, wherein the dynamic physical field reconstruction algorithm constructs a simplified physical field computing kernel fusing a fluid dynamics basic equation and a heat conduction law, and the simplified physical field computing kernel disperses the pouring area into a non-uniform sparse grid strictly corresponding to the deployment position of a sensor and a high-resolution uniform background grid covering the whole computing domain; The multi-level decision unit adopts a layered architecture and comprises a quick response layer integrated in the edge side physical field reconstruction unit and a depth analysis layer deployed on the central monitoring platform, wherein the quick response layer is integrated in the edge side physical field reconstruction unit; and the actuator control unit receives the instruction from the multi-level decision unit and drives the corresponding actuator to adjust the technological parameters.
  2. 2. The multi-sensor large casting pouring monitoring system according to claim 1, wherein the reconstruction process of the edge-side physical field reconstruction unit comprises: On a non-uniform sparse grid node, based on actual measurement data of a thermocouple temperature sensor group and a magnetic field intensity sensor group, carrying out initial interpolation by adopting a Gaussian kernel function considering space variability to obtain a temperature initial estimated value and a flow velocity initial estimated value on the non-uniform sparse grid node; Taking the temperature initial estimated value and the flow velocity initial estimated value as input boundary conditions and initial conditions of a simplified physical field calculation kernel; The simplified physical field calculation kernel performs rapid physical field evolution calculation by introducing a preset fluid viscosity coefficient of 0.001 Pa.s, a thermal diffusion coefficient of 5 multiplied by 10 < - 6 > m < 2 >/s and a slip-free boundary condition, wherein the calculation step length is set to be 2ms; the simplified physical field calculation kernel outputs a physical quantity predicted value of each node on the high-resolution uniform background grid at each time step, including temperature distribution and flow velocity distribution.
  3. 3. The multi-sensor large casting pouring monitoring system according to claim 2, wherein the edge-side physical field reconstruction unit further comprises a residual feedback module; the residual feedback module continuously compares the difference between the physical quantity predicted value obtained by physical field evolution calculation on the sparse grid node and the real-time measured value of the sensor; when the absolute value of the temperature residual exceeds 3K or the flow rate residual exceeds 0.05m/s, the residual feedback module generates a correction field based on a radial basis function; the correction field acts on the next time step of physical field evolution calculation in a weighted superposition mode, and the weight coefficient is in direct proportion to the residual error size and is used for dynamically correcting the model error.
  4. 4. The multi-sensor large casting pouring monitoring system according to claim 1, wherein a rule base of the quick response layer predefines the physical field characteristic modes of 12 defects pre-cursors; when detecting that the temperature gradient of a certain local area exceeds 15K/(s.mm) within 2s and the flow velocity predicted value of the local area is lower than 0.1m/s, judging that the local area has cold insulation risk; Once the fast response layer identifies the high risk defect pre-cursors, it immediately generates a primary regulatory instruction comprising location coordinates and risk levels.
  5. 5. A multi-sensor large casting placement monitoring system as defined in claim 1 wherein the depth analysis layer operates a pre-trained spatiotemporal convolutional neural network model; The space-time convolution neural network model takes a global temperature field, a flow velocity field and pouring technological parameters of a time sequence as input, and the space-time convolution neural network model structure comprises 5 convolution layers, 3 long-short-term memory network layers and 2 full-connection layers; the output is the evaluation of the overall quality grade of the casting and the prediction probability of 7 types of macroscopic defects; when the prediction probability of any defect exceeds 0.85, the depth analysis layer generates a secondary regulation strategy or process optimization suggestion.
  6. 6. The multi-sensor large casting pouring monitoring system according to claim 1, wherein the actuator control unit directly drives a corresponding quick actuating mechanism for a primary regulation command sent by the quick response layer, wherein the quick actuating mechanism comprises a ladle tilting servo motor or a local heater power controller for performing millisecond-level to second-level intervention, and the control precision of the ladle tilting servo motor is 0.1 degrees, and the response delay is less than 50ms; and for a secondary regulation strategy issued by the depth analysis layer, the actuator control unit coordinates a plurality of actuating mechanisms to carry out comprehensive process parameter adjustment, and the execution period is 2 s-5 s.
  7. 7. The multi-sensor large casting pouring monitoring system according to claim 1, wherein the data processing of the magnetic field intensity sensor group in the physical field sensing network comprises an eddy current compensation algorithm; The eddy current compensation algorithm calculates and deducts the background magnetic field interference generated by the eddy current effect of the molten metal in real time based on the known relation between the exciting current frequency of 10kHz and the conductivity of the molten metal; A magnetic field distortion signal caused purely by a change in flow rate was extracted, with a signal sampling rate of 2000Hz.
  8. 8. The multi-sensor large casting pouring monitoring system according to claim 1, wherein the rule base of the fast response layer supports online updating; The deep analysis layer of the central monitoring platform analyzes historical pouring data and a final casting quality inspection result monthly, and automatically refines a new defect pre-cursors characteristic mode; After the confirmation of the engineer, the local rule base is updated by sending the encrypted data to each edge side physical field reconstruction unit through the encrypted network channel.
  9. 9. The multi-sensor large casting pouring monitoring system according to claim 1, further comprising a data synchronization and clock calibration module; The data synchronization and clock calibration module ensures that data acquisition time stamps of all sensors in the physical field sensing network, a calculation period of the edge side physical field reconstruction unit and an instruction execution time of the actuator control unit are all kept in microsecond synchronization with the high-precision Beidou time source; The maximum clock deviation is controlled within +/-2 mu s, so that the strict time sequence consistency of the data flow and the control flow of the whole system is ensured.
  10. 10. A multi-sensor large casting pouring monitoring method is characterized by comprising the following steps: S1, acquiring multi-source physical quantity data in a pouring process through a physical field sensing network, wherein the multi-source physical quantity data comprises a plurality of sensor groups which are arranged on the outer wall of a casting mould, a pouring system and preset internal detection points, each sensor group comprises a thermocouple temperature sensor group, a vibration acceleration sensor group and a magnetic field intensity sensor group, the thermocouple temperature sensor groups are arranged according to a non-uniform topological structure, and the arrangement positions of all measuring points are determined through computational fluid mechanics simulation of the pouring process performed in advance; S2, in an edge side physical field reconstruction unit, receiving a real-time data stream from a physical field sensing network, and executing a dynamic physical field reconstruction algorithm based on physical mechanism guidance to reconstruct a global physical field in real time, wherein the dynamic physical field reconstruction algorithm constructs a simplified physical field calculation kernel fusing a fluid dynamics basic equation and a heat conduction law, and the simplified physical field calculation kernel disperses a pouring area into a non-uniform sparse grid strictly corresponding to a sensor deployment position and a high-resolution uniform background grid covering the whole calculation domain; S3, performing defect risk identification and regulation decision generation through a multi-level decision unit, wherein the multi-level decision unit comprises a quick response layer integrated in the edge side physical field reconstruction unit and a depth analysis layer deployed on a central monitoring platform; s4, receiving an instruction from the multi-level decision unit through the actuator control unit, and driving a corresponding actuator to adjust the technological parameters.

Description

Multi-sensor large casting pouring monitoring method and system Technical Field The invention belongs to the technical field of casting process monitoring, and particularly relates to a multi-sensor large casting pouring monitoring method and system. Background In the field of large-scale casting manufacturing, accurate monitoring and control of a pouring process are core links for determining the internal quality and service performance of a casting, the technology relates to a plurality of discipline branches such as casting technology, thermodynamics, hydrodynamics, automation control and the like, and the development level of the technology is directly related to the manufacturing quality and reliability of national important equipment such as energy equipment, heavy machinery and the like. The large casting pouring monitoring technology is used as a key component of automation of a casting process, aims to collect key physical parameters such as a temperature field, a flow field and the like in the casting process in real time through a multi-sensor system, and realizes closed-loop regulation and control of technological parameters based on data analysis, so that compactness and uniformity of internal tissues of the casting are ensured. In the prior art, a temperature sensor with uniform gridding deployment is commonly adopted for data acquisition, but when a large casting with a deep cavity, a small curvature radius corner and other complex geometric structures is faced, the sensor deployment is affected by the physical shielding of a die, so that a monitoring blind area is formed in the key areas of the cavity corner, the core head root and the like, and the air hole and shrinkage porosity defect omission ratio are high. In order to solve the problem of blind area data deletion, the existing scheme mostly adopts pure data driving models such as inverse distance weighting or Kriging interpolation to carry out deduction, however, the method completely ignores the physical laws of fluid dynamics and heat transfer in the casting process, the interpolation result is easy to generate obvious errors under the dynamic working condition that the flow rate of metal liquid is changed severely, so that the temperature field is reconstructed seriously, meanwhile, the existing monitoring system relies on a central server to carry out multi-source heterogeneous data fusion and decision, the end-to-end delay from data acquisition to the output of a regulation instruction is longer, the critical window period formed by casting defects is often exceeded, the failure rate of closed loop regulation is high, and the calculation complexity of the solving process is high, so that the millisecond response requirement of real-time monitoring cannot be met despite research attempts to introduce the computational fluid mechanics model to promote the prediction precision. Therefore, how to realize high-precision and low-delay dynamic monitoring of the whole pouring process of large castings on the premise of ensuring the consistency of physical laws has become a technical problem to be broken through in the field. Disclosure of Invention The invention aims to solve the technical problems of providing a multi-sensor large casting pouring monitoring method and system, and aims to overcome the technical contradiction that in the prior art, key area data is lost due to sensor deployment blind areas, reconstruction distortion is caused under dynamic working conditions due to the fact that a physical rule is ignored due to the fact that a pure data interpolation model is relied on, and a system response delay is too long and a defect regulation key window period is missed due to a centralized data processing mode. In order to achieve the purpose, the invention constructs a complete system integrating perception enhancement guided by a physical mechanism, edge side real-time physical field reconstruction and multi-level collaborative decision, and realizes omnibearing, high-precision and low-delay monitoring of the pouring process of a large casting by collaborative work of a plurality of sensing nodes, edge computing units and a central monitoring platform which are deployed on a pouring site. A multi-sensor large casting pouring monitoring system comprises a physical field sensing network, an edge side physical field reconstruction unit, a multi-level decision unit and an actuator control unit. The physical field sensing network is responsible for collecting multi-source physical quantity data in the pouring process, the edge side physical field reconstruction unit is based on the sensing network data and combined with an embedded simplified physical mechanism model to reconstruct the whole physical field of a pouring area in real time, the multi-level decision unit is responsible for carrying out defect risk identification and regulation decision generation on the reconstructed physical field data, and the actuator control unit is