Multi-source data fusion processing method for fastener heat treatment

Abstract

The invention discloses a fastener heat treatment-oriented multi-source data fusion processing method, which relates to the field of data fusion and comprises five steps of multi-source data acquisition, data preprocessing, feature extraction, fusion modeling and result optimization. The multi-source data acquisition terminal acquires process data, material data, quality detection data and environment data; data cleaning and standardization are carried out in data preprocessing; the feature extraction terminal extracts a time sequence feature parameter, a component feature parameter, a quality feature parameter and an interference feature parameter; fusion modeling is combined with mechanism prior and data driving to realize fusion of a feature layer and a decision layer; and result optimization: smoothly correcting the fusion result, and outputting quality evaluation and process adjustment suggestions. According to the method, through deep fusion of multi-source data, the accuracy and stability of fastener heat treatment quality evaluation are improved, adaptive optimization of process parameters is realized, and the method is suitable for heat treatment whole-process management and control in a structured industrial scene.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of data fusion, in particular to a multi-source data fusion processing method for fastener heat treatment. BACKGROUND

[0002] In the field of fastener heat treatment production, product quality is closely related to process parameters, and key indicators such as hardness, strength, and deformation directly depend on the coordinated control of parameters such as heating temperature, holding time, and cooling rate. However, the traditional production mode has long relied on manual experience for process control, and there are outstanding problems such as quality evaluation lag and blind parameter optimization.

[0003] Current industry data collection is mostly limited to a single process link, such as monitoring only the temperature of the heating furnace or the flow of the cooling medium, lacking systematic integration of multi-source information such as process data (e.g. equipment power, furnace pressure), material data (e.g. composition content, original hardness), quality detection results (e.g. tensile strength, metallographic structure), and environmental interference (e.g. workshop temperature and humidity, power grid voltage), making it difficult to quickly trace the source when quality fluctuates.

[0004] Although existing data processing methods attempt to introduce simple statistical analysis, they lack the ability to handle noise interference, spatial and temporal dislocation, and other problems of multi-source heterogeneous data, and do not sufficiently mine feature correlations, making it impossible to establish a precise mapping relationship between process parameters and quality indicators. At the same time, process parameter adjustments are mostly based on fixed thresholds, making it difficult to adapt to material batch differences and environmental fluctuations, resulting in poor product consistency and large fluctuations in pass rates. Therefore, developing a method that can integrate multi-source data and achieve precise quality evaluation and process adaptive optimization throughout the entire process is key to improving fastener heat treatment levels. SUMMARY

[0005] In order to overcome the above-mentioned defects of the prior art, the embodiments of the present application provide a laser ground receiving system that can be quickly deployed to solve the problems raised in the background art.

[0006] To achieve the above-mentioned purpose, the present application provides the following technical scheme: a multi-source data fusion processing method for fastener heat treatment, comprising: S1, multi-source data acquisition: the fastener to be processed is denoted as a target object, and multi-source data in the heat treatment process of the target object is acquired in real time through a multi-source data acquisition terminal, wherein the multi-source data includes process data, material data, quality detection data, and environmental data; S2, data preprocessing: the multi-source data is cleaned and standardized by a data preprocessing terminal, and process data preprocessing parameters, material data preprocessing parameters, quality detection data preprocessing parameters, and environmental data preprocessing parameters are obtained respectively; S3, feature extraction: input the process data preprocessing parameters, material data preprocessing parameters, quality detection data preprocessing parameters and environmental data preprocessing parameters into the feature extraction terminal, and output the time series feature parameters, component feature parameters, quality feature parameters and interference feature parameters; S4, fusion modeling: input the component feature parameters, quality feature parameters and interference feature parameters into the fusion model through the fusion modeling terminal, combine the mechanism prior weight and the data-driven attention mechanism for feature layer fusion, and then perform decision layer fusion through the multi-quality index sub-model to output the fusion result; S5, result optimization: the fusion result is processed through the result optimization terminal, and finally the heat treatment quality evaluation result of the target object is output, the heat treatment quality evaluation result is judged, and if the preset result is not met, the process parameter adjustment is performed.

[0007] The technical effects and advantages of the present application are as follows: 1. The present application integrates process, material, quality and environmental data through multi-source data acquisition and fusion modeling, breaks through the limitation of traditional single link data, constructs the full-process quality mapping relationship, realizes the accurate evaluation of heat treatment quality, and solves the problem of quality fluctuation traceability; 2. The present application combines mechanism priori and data-driven attention mechanism, dynamically optimizes feature weight, and compared with fixed threshold method, can adapt to material batch difference and environmental fluctuation, improve product consistency, and reduce the fluctuation amplitude of qualified rate; 3. The present application outputs process adjustment suggestion through the result optimization terminal, quantifies parameter adjustment amount based on quality deviation, realizes closed-loop control from data acquisition to parameter optimization, reduces the dependence on artificial experience, and improves production efficiency and quality stability. BRIEF DESCRIPTION OF DRAWINGS

[0008] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed to be used for the embodiment description. Obviously, the drawings in the following description are only some embodiments of the present application, and for the basic field technical personnel, other drawings can also be obtained without creative labor on the basis of these drawings.

[0009] Figure 1 It is a terminal connection schematic diagram.

[0010] Figure 2 It is a process implementation schematic diagram.

[0011] Figure 3 It is a data processing process schematic diagram. DETAILED DESCRIPTION

[0012] With reference to the drawings of the embodiments of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described, obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the protection scope of the present application.

[0013] Please refer to Figure 1 As shown in the drawings, the present application provides a multi-source data fusion processing method for fastener heat treatment, comprising a multi-source data acquisition terminal, a data preprocessing terminal, a feature extraction terminal, a fusion modeling terminal and a result optimization terminal.

[0014] The multi-source data acquisition terminal is connected with the data preprocessing terminal, for transmitting the collected process data, material data, quality detection data and environment data to the data preprocessing terminal; the data preprocessing terminal is connected with the feature extraction terminal, for transmitting the preprocessed parameters to the feature extraction terminal; the feature extraction terminal is connected with the fusion modeling terminal, for transmitting the extracted time sequence features, component features, quality features and interference feature parameters to the fusion modeling terminal; the fusion modeling terminal is connected with the result optimization terminal, for transmitting the fusion results to the result optimization terminal, and finally outputting heat treatment quality evaluation results, process parameter adjustment suggestions and full-process data trajectories.

[0015] Please refer to Figure 2 The present application provides a multi-source data fusion processing method for fastener heat treatment, comprising: S1, multi-source data acquisition: taking a fastener to be processed as a target object, collecting multi-source data in a heat treatment process of the target object in real time through a multi-source data acquisition terminal, wherein the multi-source data comprises process data, material data, quality detection data and environment data; The process data comprises a heating furnace temperature, a cooling medium flow, an in-furnace pressure, a device running power and a running time. The material data comprises a material component content and an original hardness of the target object. The quality detection data comprises a hardness, a tensile strength, a metallographic structure image and a deformation amount after heat treatment. The environment data comprises a workshop temperature, a humidity and a power grid voltage.

[0016] The process data refers to data related to a device running state and a process parameter, which is collected in real time by various online sensors in a heat treatment process of the fastener, and directly reflects dynamic changes of each link in the heat treatment.

[0017] In this embodiment, the collection range and specific method of process data are as follows: the temperature of the heating furnace is collected by distributed thermocouples in the furnace, with 3 measuring points (9 thermocouples in total) arranged on the upper, middle and lower layers of the furnace chamber to ensure coverage of the temperature field in different areas of the furnace. The measurement range is 20-1200℃, the accuracy is ±1℃, and the sampling frequency is set to 1Hz. The collected raw data includes real-time temperature values of each measuring point, for example, the upper thermocouple collects 902℃, 899℃, 903℃, the middle layer collects 898℃, 897℃, 899℃, and the lower layer collects 900℃, 898℃, 902℃. These data are transmitted to the data acquisition card through the thermocouple dedicated signal line, and then uploaded to the data server through the industrial Ethernet for subsequent reflection of the temperature distribution and change trend in the heating furnace.

[0018] The cooling medium flow rate is collected by an electromagnetic flowmeter to collect the real-time flow rate of the cooling medium. The flowmeter is installed at the inlet of the cooling pipeline, the measurement range is 0-50L / min, the accuracy is ±0.5%, and the sampling frequency is 0.5Hz. Due to the influence of the equipment power supply system, the collected raw data has 50Hz power frequency interference, which shows periodic fluctuations, for example, alternating changes in the range of 25-27L / min. The flow data are transmitted to the PLC through 4-20mA analog signals, and then forwarded to the data server by the PLC for monitoring whether the cooling rate meets the process requirements.

[0019] The furnace pressure is collected by a pressure transmitter installed on the side of the furnace body, with a measurement range of -0.1-0.5MPa, an accuracy of ±0.2%, and a sampling frequency of 0.5Hz. Due to the influence of the sealing performance of the furnace body and the change of the environmental temperature, the raw data has baseline drift phenomenon, for example, increasing by 0.01MPa per hour, collecting 0.20MPa at 8:00, 0.21MPa at 9:00, and 0.22MPa at 10:00. The pressure data are transmitted to the data acquisition terminal through the RS485 bus, and finally uploaded to the server to ensure that the furnace pressure is stable within the process specified range.

[0020] The equipment running power is collected by a power sensor, which is installed on the power supply line of the equipment. The measuring range is 0-500 kW, the accuracy is ±1%, and the sampling frequency is 1 Hz. The equipment start-stop instant will produce a power peak, for example, the power can reach 600 kW when starting, which exceeds the normal operating range. The power data is transmitted to the gateway through wireless transmission, and then uploaded to the server after being aggregated by the gateway, which is used to monitor the equipment energy consumption and running state. The running time is recorded by the system clock for each process stage, with an accuracy of ±0.1 s. For example, the starting time of the heating stage is marked as t=0 s, and when the temperature reaches the austenitizing temperature, such as 850℃, it is marked as t=300 s, which is the start of the holding stage, and the starting time of the cooling stage is marked as t=900 s, which is the end of the holding stage. The running time data is used as a time reference and is stored in association with temperature, pressure, flow and other parameters to ensure the consistency of various process data in the time dimension.

[0021] Material data refers to the original attribute data of fasteners before heat treatment, reflecting the inherent characteristics of the material, and having a direct impact on the quality after heat treatment.

[0022] In this embodiment, the collection content and method of material data are as follows: The material composition content is detected by a spectrum analyzer for the raw material of the fastener offline, and the detected elements include C, Si, Mn, Cr, Ni, Mo and other key alloy elements, with a detection accuracy of ±0.01%. For example, the material composition of a batch of bolts is: C=0.45%, Si=0.25%, Mn=1.2%, Cr=0.8%, Ni=0.3%, Mo=0.1%. After detection, the composition data is entered into the material information management system to form a material composition database, which serves as the basis for subsequent analysis of material hardenability, hardness and other properties.

[0023] The original hardness is detected by a Rockwell hardness tester before heat treatment of the fastener, and the average value of 3 detections at different positions (such as head, shaft, tail) of the fastener is taken as the original hardness value. For example, the detection results of a nut are 25HRC, 26HRC and 25HRC, with an average value of 25.3HRC. The original hardness data is recorded in the quality inspection report and uploaded to the data server for comparison with the hardness after heat treatment to evaluate the heat treatment effect.

[0024] Quality detection data refers to the data related to product quality obtained by offline detection means after heat treatment of the fastener, which directly reflects the final effect of the heat treatment process.

[0025] In this embodiment, the collection content and method of quality detection data are as follows: the hardness after heat treatment is detected by using the same Rockwell hardness tester as that for original hardness detection, 5 detection points (uniformly distributed along the axial direction) are selected on the fastener after heat treatment, 3 detections are performed at each point, and the average value is taken as the hardness value of the position. For example, the detection results of a certain bolt are: head 44.8HRC, 45.2HRC, average 45.0HRC, middle 43.9HRC, 44.1HRC, average 44.0HRC, tail 42.8HRC, 43.2HRC, average 43.0HRC. The hardness data is input into the quality detection system, which is used to evaluate whether the hardness of the fastener meets the standard and uniformity.

[0026] The tensile strength is tested by using a universal testing machine on 5 fastener samples of the same batch, the maximum load at the time of sample fracture is recorded during the test, and then the tensile strength is calculated according to the cross-sectional area of the sample. For example, the test results of 5 samples are 1000MPa, 1020MPa, 990MPa, 1010MPa, and 1030MPa, which are used to evaluate whether the mechanical strength of the fastener meets the design requirements.

[0027] The metallographic structure image is collected by using a metallographic microscope to collect the metallographic structure image of the cross section of the fastener, the sample is first polished, polished, and etched, and then observed under the microscope, the image resolution is 2048x1536 pixels, and the magnification is 500x. For example, the collected image contains martensite, pearlite and other structures, which are used to analyze the grain size and phase composition ratio. The deformation is measured by using a laser diameter measuring instrument to measure the diameter and length of the fastener before and after heat treatment, and the size change is calculated as the deformation. For example, the diameter of a certain bolt before treatment is 10.000mm, and the diameter after treatment is 10.005mm, the deformation is 5μm; the length before treatment is 50.00mm, and the length after treatment is 50.02mm, the deformation is 20μm. The deformation data is used to evaluate whether the size stability of the fastener meets the assembly requirements.

[0028] The environmental data refers to the environmental parameters of the heat treatment workshop, which does not directly participate in the process control, but may indirectly affect the equipment operation and process stability. In this embodiment, the collection content and method of environmental data are as follows: the workshop temperature is collected by using a temperature and humidity sensor to collect the environmental temperature of the workshop, the sensor is installed in the corner of the workshop and near the equipment (5 measurement points in total), the measurement range is 0-50℃, the accuracy is ±0.5℃, and the sampling frequency is 0.1Hz. For example, the collected data is 25℃, 26℃, 25℃, 27℃, and 25℃, which is used to analyze the influence of environmental temperature on the heat dissipation of the heating furnace.

[0029] The humidity and the workshop temperature are collected by the same temperature and humidity sensor, the measurement range is 10%-90%RH, the accuracy is ±3%RH, and the sampling frequency is 0.1Hz. For example, the collected data is 40%RH, 42%RH, 39%RH, 41%RH, 40%RH, which is used to evaluate the influence of humidity on the electrical elements of the equipment and the evaporation of the cooling medium.

[0030] The grid voltage is collected by a voltage sensor, the sensor is installed in the power distribution cabinet, the measurement range is 380±50V, the accuracy is ±0.5V, and the sampling frequency is 0.1Hz. For example, the collected data is 380V, 379V, 381V, 380V, 378V, which is used to monitor the influence of voltage fluctuation on the power output of the heating furnace.

[0031] S2, data preprocessing: through the data preprocessing terminal, the multi-source data is cleaned and standardized, and the process data preprocessing parameters, material data preprocessing parameters, quality detection data preprocessing parameters and environmental data preprocessing parameters are obtained respectively; The process data preprocessing parameters are obtained by a smoothing algorithm; the material data preprocessing parameters are obtained by a range standardization method, and the calculation formula is: , Among them, is the output value after standardization, is the minimum value under the feature dimension, is the maximum value under the feature dimension, is the original data value.

[0032] The specific processing method of process data preprocessing is as follows: the heating furnace temperature preprocessing first carries out spatial consistency check on the original temperature data of 9 measuring points, and eliminates the abnormal measuring points obviously deviating from the overall trend, such as a measuring point showing 1200℃ due to failure, which is directly eliminated, and then calculates the arithmetic mean value of three layers of measuring points, such as the mean value of the upper layer 901.3℃, the mean value of the middle layer 898.0℃, and the mean value of the lower layer 900.0℃, and the overall mean value is (901.3+898.0+900.0) / 3≈899.8℃.

[0033] Then apply sliding window median filtering, the window size is set to 5 sampling points (i.e. 5 seconds), for example, the original data is 899℃, 900℃, 950℃, 901℃, 900℃, and after filtering, it is corrected to 899℃, 900℃, 900℃, 901℃, 900℃, and the instantaneous pulse interference is filtered out. Finally, the smoothed temperature time curve is obtained, which is used for subsequent extraction of temperature change rate and other characteristics.

[0034] The cooling medium flow rate preprocessing is aimed at the 50 Hz power frequency interference in the original data, and the data is converted to the frequency domain by Fourier transform, the 50 Hz and its harmonic components are identified and removed, and then the data is converted back to the time domain by inverse Fourier transform to obtain the denoised flow rate data. For example, the data before denoising fluctuates between 25-27 L / min, and after denoising it stabilizes at about 26 L / min. Then the 1 minute sliding window average value (i.e. the average value of 12 sampling points) is calculated to further smooth the data, for example, the 12 denoised data are 26.1, 25.9, 26.0, 26.2, 25.8, 26.0, 26.1, 25.9, 26.0, 26.1, 25.9, 26.0, and the average value is 26.0 L / min, which is used as the basic data of the flow rate average value feature.

[0035] The furnace pressure preprocessing is aimed at the baseline drift problem, and the linear fitting method is used to eliminate the drift. Assuming that the pressure data from 8:00 to 10:00 is 0.20 MPa, 0.205 MPa, 0.21 MPa, 0.215 MPa, 0.22 MPa, the drift amount formula is 0.01t, t is the number of hours, 8:00 corresponds to t=0, 9:00 corresponds to t=1, and so on. After subtracting the drift amount from the original data, the corrected pressure data is obtained: 0.20 MPa, 0.195 MPa, 0.20 MPa, 0.205 MPa, 0.21 MPa, which eliminates the linear drift. Then the pressure fluctuation amplitude within 5 minutes is calculated, which is the difference between the maximum value and the minimum value. For example, the corrected pressure data from 8:00 to 8:05 is 0.20 MPa, 0.198 MPa, 0.202 MPa, 0.199 MPa, 0.201 MPa, the maximum value is 0.202 MPa, the minimum value is 0.198 MPa, and the fluctuation amplitude is 0.004 MPa.

[0036] The device running power preprocessing sets the power threshold to 0-500 kW, and the peak data exceeding the threshold is corrected by replacing it with the average value of the adjacent 3 normal sampling points. For example, a certain segment of data is 480 kW (t=1 s), 600 kW (t=2 s, abnormal), 470 kW (t=3 s), then the power at t=2 s is corrected to (480+470) / 2=475 kW. The corrected data is then used to calculate the power standard deviation within 30 seconds (i.e. the dispersion degree of 30 sampling points), for example, the average value of 30 data is 490 kW, and the standard deviation is 5 kW, reflecting the stability of the power.

[0037] The running time preprocessing does not require complex processing, only the time stamp alignment with other process data is required to ensure that each temperature, pressure, flow rate, etc. data point corresponds to a unique time marker. For example, the heating stage starting time t=0 s is associated with the first temperature data 20℃, and t=300 s is associated with the holding stage starting temperature 850℃, forming a multi-parameter time sequence matrix with time as the horizontal axis.

[0038] The material data preprocessing is specifically processed as follows: the original composition content of C element is 0.45%, the minimum value under the characteristic dimension is 0.3%, and the maximum value is 0.6%. The normalized value of C element is calculated as (0.45-0.3) / (0.6-0.3)=0.5. Similarly, the normalized value of Si element is 0.5, the normalized value of Mn element is 0.4, the normalized value of Cr element is 0.6, the normalized value of Ni element is 0.33, and the normalized value of Mo element is 0.33.

[0039] Further, the normalized values are combined into a composition feature vector [0.5, 0.5, 0.4, 0.6, 0.33, 0.33], which is used for subsequent fusion modeling. The original hardness preprocessing does not need to be normalized, only the average value of multiple detections is calculated, and directly stored as the initial performance parameter of the material, which is used for comparison and analysis with the hardness after heat treatment. For example, the average value of the original hardness of a batch of fasteners is 25.3HRC, and the average value of the hardness after heat treatment is 45.0HRC. The hardness improvement amplitude can be directly calculated as 19.7HRC.

[0040] The specific processing method of the quality detection data preprocessing is as follows: the hardness preprocessing after heat treatment arranges the average value of the hardness of each detection point, calculates the hardness gradient along the axial direction of the fastener, and the formula is: For example, the head hardness of a bolt is 45.0HRC, the tail hardness is 43.0HRC, and the axial length is 20cm. The hardness gradient is 0.1HRC / cm, which reflects the change rate of the hardness along the axial direction and evaluates the hardness uniformity.

[0041] The tensile strength preprocessing calculates the average value and the standard deviation of the tensile strength data of 5 samples. The average value is (1000+1020+990+1010+1030) / 5=1010MPa, and the standard deviation is 15MPa. The average value is taken as the representative value of the tensile strength of the batch, and the standard deviation is taken as the strength dispersion degree to evaluate the performance consistency within the batch.

[0042] The metallographic structure image preprocessing first removes noise from the original image by Gaussian filtering, then identifies the grain boundary by edge detection algorithm, randomly selects 100 grains to calculate the equivalent diameter, and counts the martensite phase ratio by gray threshold segmentation. The two quantitative indicators are taken as image features.

[0043] The deformation amount preprocessing records the maximum size difference before and after heat treatment, takes the maximum value as the deformation amount peak value, and compares it with the standard value to evaluate the dimensional stability.

[0044] ​The specific processing mode of the environment data preprocessing is as follows: the temperature in the workshop is preprocessed to calculate the temperature fluctuation range within 1 hour, for example, the temperature data within a certain hour is 23-27℃, the fluctuation range is 4℃, and the time points exceeding the normal range (15-35℃) are marked as abnormal, for example, a certain time point reaches 36℃ in summer, which is marked as abnormal. Humidity preprocessing does not require additional transformation, only when the humidity exceeds the range of 10%-90%RH, such as 92%RH recorded in plum rain season, it is recorded as an environmental interference factor for reference. The voltage standard deviation within 10 minutes is calculated, and then divided by the rated voltage 380V to obtain the voltage stability coefficient to judge the stability of the power grid (required to be ≤2%).

[0045] S3, feature extraction: input the process data preprocessing parameters, material data preprocessing parameters, quality detection data preprocessing parameters and environment data preprocessing parameters into the feature extraction terminal, and output the time sequence feature parameters, composition feature parameters, quality feature parameters and interference feature parameters; The time sequence feature parameters include temperature change rate, pressure fluctuation amplitude, flow mean value and power standard deviation. The composition feature parameters include element content proportion and composition uniformity. The quality feature parameters include hardness gradient, strength dispersion and deformation peak value. The interference feature parameters include temperature fluctuation range and voltage stability coefficient.

[0046] The temperature change rate is calculated every 10 seconds based on the preprocessed heating furnace temperature time sequence curve, which is calculated by the temperature values of 5 consecutive sampling points, and the formula is: , Wherein, is the temperature value of the current sampling point, is the average temperature of the previous 4 sampling points, is the sampling time interval, is the temperature change rate; the composition uniformity is calculated by the ratio of the standard deviation of the element content to the mean value.

[0047] Taking this embodiment as an example, the temperature value of the current sampling point is 900.5℃, the average temperature of the previous 4 sampling points is 900.1℃, and the sampling time interval is 1 second. Substituting the formula, 0.4℃ / s is obtained. This feature reflects the heating rate, and if it exceeds the process threshold (such as 1℃ / s), it is prompted that the rapid heating may cause uneven organization.

[0048] The pressure fluctuation amplitude is calculated based on the pre-processed furnace pressure data, and the difference between the maximum and minimum values within 5 minutes is calculated. For example, in this example, the pressure data from 8:00 to 8:05, the maximum value is 0.202 MPa, the minimum value is 0.198 MPa, and the fluctuation amplitude is 0.004 MPa. If it exceeds 0.01 MPa, it indicates that the furnace body sealing may be abnormal, and needs to be checked in combination with the equipment log.

[0049] The flow average is calculated based on the pre-processed cooling medium flow data, and the average value of a 1-minute sliding window (12 sampling points) is calculated. For example, the flow data average value in a certain window is 26.0 L / min, which directly reflects the stability of the cooling medium supply. If it is lower than the process requirement of 25 L / min, the pump power needs to be adjusted.

[0050] The power standard deviation is calculated based on the pre-processed equipment operating power data, and the power dispersion within 30 seconds is calculated. For example, the average value of 30 data is 490 kW, and the standard deviation is 5 kW. The formula is: , wherein, is a single power value, is the average value, n=30, and the smaller the standard deviation, the more stable the equipment power output. If it exceeds 10 kW, it indicates that there is abnormal fluctuation in the power grid or equipment.

[0051] The element content ratio is based on the pre-processed material composition standardized data, and the standardized value is directly used as a feature to reflect the proportion of each element within the process allowable range. For example, the C element standardized value is 0.5, indicating that its content is at the medium level of historical batches, and the austenitizing temperature needs to be adjusted for the holding time. High C content needs to be extended for holding.

[0052] The composition uniformity is calculated based on the raw material composition data of 3 samples from the same batch, and the ratio of standard deviation to average value is calculated. The formula is: , wherein , , and the result is about 2.2%. If the uniformity exceeds 5%, it indicates that the raw material composition is segregated, and the heating temperature needs to be appropriately increased during heat treatment to promote composition diffusion.

[0053] The hardness gradient is calculated based on the pre-processed hardness data after heat treatment, and the hardness change rate along the axial direction is calculated. The formula is: , wherein the axial length is 20 cm, and the result is 0.1 HRC / cm. The smaller the gradient, the better the hardness uniformity. If it exceeds 0.3 HRC / cm, it indicates that the cooling medium distribution is uneven, and the nozzle position needs to be adjusted.

[0054] The intensity dispersion is based on the pretreated tensile strength data, and the standard deviation of 5 samples is 15 MPa. The smaller the dispersion, the more stable the batch performance. If it exceeds 30 MPa, the temperature field uniformity of the heating furnace needs to be checked.

[0055] The metallographic structure features are based on the pretreated metallographic image quantitative data, and the equivalent diameter of the grain and the proportion of the martensite phase are extracted. In this embodiment, 100 grains are randomly selected, and the average diameter is 15 μm. If it exceeds 20 μm, it indicates that the holding time is too long, resulting in coarse grains. Through gray scale segmentation, 85% is obtained. If it is less than 80%, it indicates that the cooling medium flow is insufficient, and the cooling medium flow needs to be increased.

[0056] The deformation peak is based on the pretreated deformation data, and the maximum value 20 μm is taken as the feature. If it exceeds the process threshold 10 μm, it indicates that the quenching stress distribution is uneven, and the cooling path needs to be optimized.

[0057] The temperature fluctuation range is based on the pretreated workshop temperature data, and the maximum temperature difference 4℃ within 1 hour is calculated. If it exceeds 5℃, it indicates that the workshop temperature control system is abnormal, which may cause unstable heat dissipation of the heating furnace, and needs to be corrected in combination with the furnace temperature data.

[0058] The voltage stability coefficient is based on the pretreated power grid voltage data, and the ratio of the voltage standard deviation within 10 minutes to the rated value is calculated. The formula is: , Wherein , , and the coefficient is about 0.4%. If the coefficient exceeds 2%, it indicates that the voltage fluctuation may affect the heating power output, and the weight of the power feature needs to be increased in the fusion model.

[0059] Finally, the above extracted feature parameters are integrated into feature vectors according to categories, such as time series feature vector: [0.4℃ / s, 0.004MPa, 26.0L / min, 5kW]; component feature vector: [0.5, 0.5, 0.4, 0.6, 0.33, 0.33, 2.2%]; quality feature vector: [0.1HRC / cm, 15MPa, 15μm, 85%, 20μm]; interference feature vector: [4℃, 0.4%]. These feature vectors will be used as inputs of the fusion modeling terminal, and the mechanism prior weight and data driven attention mechanism are used to realize the deep fusion of multi-source data.

[0060] S4, fusion modeling: input the component feature parameters, quality feature parameters and interference feature parameters into the fusion model through the fusion modeling terminal, combine the mechanism prior weight and data driven attention mechanism for feature layer fusion, and then perform decision layer fusion through multi-quality index sub-model to output the fusion result; The fusion model comprises feature layer fusion and decision layer fusion. The mechanism prior weight is set based on an austenitizing temperature- holding time correlation relationship of fastener heat treatment, a data-driven attention mechanism learns feature importance coefficients, and the fusion feature vector is an accumulation sum of the product of each type of feature parameter and the corresponding weight.

[0061] The multi-quality index sub-model comprises a hardness uniformity sub-model, a strength stability sub-model, a microstructure eligibility sub-model, and a size precision sub-model.

[0062] The hardness uniformity sub-model takes hardness gradient characteristics as core input and outputs hardness distribution uniformity scores, and the formula is: , wherein is a hardness gradient value, is a hardness influence coefficient, is a natural constant; The strength stability sub-model is based on strength dispersion characteristics and outputs batch strength consistency coefficients, and the formula is: , wherein is a strength standard deviation, is a maximum allowable standard deviation; The microstructure eligibility sub-model combines grain diameter and martensite phase proportion characteristics and outputs microstructure eligibility rates, and the formula is: , wherein d is an actual grain diameter, are upper and lower limits of the grain diameter, respectively, is a martensite phase proportion; The size precision sub-model is based on deformation amount peak characteristics and outputs size deviation levels, and the formula is: , wherein is a deformation amount peak value, is a deformation amount threshold value; The output values of each sub-model are weighted and summed according to preset weights wherein is a preset weight and the sum is 1.

[0063] Preferably, the implementation is as follows: the core of feature layer fusion is to fuse the component feature parameters, quality feature parameters and interference feature parameters in the spatial dimension, and highlight the contribution of key features through mechanism prior weight and data-driven attention mechanism. First, the feature parameter input is performed, and three types of feature parameter vectors are input: a component feature vector: [0.5, 0.5, 0.4, 0.6, 0.33, 0.33, 2.2%], a quality feature vector: [0.1HRC / cm, 15MPa, 15μm, 85%, 20μm] and an interference feature vector: [4℃, 0.4%]; then, the mechanism prior weight distribution is performed, and different features are given prior weights based on the mechanism of the heat treatment process: in the component feature, the C element (0.5) and the Cr element (0.6) have a significant impact on the hardenability, and the weights are set to 0.2 and 0.15 respectively, and the component uniformity is 2.2% and the weight is set to 0.1; In the quality feature, the martensite ratio of 85% and the hardness gradient of 0.1HRC / cm directly reflect the phase change quality, and the weights are set to 0.2 and 0.15 respectively; In the interference feature, the voltage stability coefficient of 0.4% has a greater impact on the heating power, and the weight is set to 0.1, and the temperature fluctuation range (4℃) weight is set to 0.05; the total weight is 1.0, which ensures that the feature contribution can be directly compared.

[0064] Further, the data-driven attention mechanism is used to construct an attention network, and the feature dynamic weight is learned based on historical data: the input layer receives the standardized feature vector, the hidden layer uses the ReLU activation function, and the output layer generates the attention weight through the Softmax function ; for example, when the C element content of a batch is relatively high (0.5) and the martensite ratio is relatively low (85%), the network automatically increases the attention weight of the C element feature (such as from 0.2 to 0.25), and reduces the weight of the temperature fluctuation feature (such as from 0.05 to 0.03); finally, the attention weight and the mechanism prior weight are weighted and fused, and the formula is: , wherein 0.7 and 0.3 are the balance coefficients of mechanism and data, is the mechanism prior weight, is the attention weight.

[0065] Finally, the feature fusion output is performed, and the weighted feature vector is spliced into a fusion feature matrix, and the embodiment is a 14-dimensional matrix.

[0066] The decision layer fuses the fusion feature matrix based on the feature layer output, and outputs the comprehensive quality evaluation result through the integrated learning of multiple quality index sub-models. First, multiple quality index sub-models are constructed, and sub-models are trained for key quality indexes: the hardness uniformity sub-model takes the fusion feature matrix as input, adopts the random forest regression model, and outputs the hardness uniformity score. For example, after inputting the features of a batch, the model outputs 85 points, which corresponds to a hardness gradient of 0.1 HRC / cm, which is better than the threshold of 0.3 HRC / cm; The strength stability sub-model adopts support vector machine regression, outputs the strength dispersion prediction value, and converts it into a score, such as 15 MPa, 90 points, because 15 MPa < 30 MPa threshold; The organization eligibility sub-model is based on a convolutional neural network, inputs the grain diameter of 15 μm and the martensite proportion of 85% in the fusion features, and outputs the organization eligibility rate of 92%; The size accuracy sub-model adopts linear regression, outputs the deformation peak prediction value of 20 μm, and converts it into a score of 60 points, because 20 μm > 10 μm threshold.

[0067] Further, the sub-model weight distribution is performed, and the sub-model weight is assigned according to the importance of the quality index: the hardness uniformity and the organization eligibility rate have higher weights, each set to 0.3; the strength stability weight is set to 0.25; the size accuracy weight is set to 0.15, and the weight sum is 1.0, reflecting the influence degree of different indexes on the final quality. Finally, the comprehensive decision output is calculated, the weighted sum of the sub-model output is taken as the fusion result, and 84.1 points are obtained. If the fusion result is greater than or equal to 80 points, it is determined that the quality of the batch of fasteners after heat treatment is qualified.

[0068] S5, result optimization: the fusion result is processed by the result optimization terminal, and the heat treatment quality evaluation result of the target object is finally output. The heat treatment quality evaluation result is judged, and if it does not meet the preset result of 80 points, the process parameter is adjusted.

[0069] The data processing process is shown in Figure 3 , and specifically is: S51, the mean value is calculated for a plurality of fusion results; S52, the mean value is detected for abnormal values; S53, the abnormal values are replaced by the median of adjacent results.

[0070] The process parameter adjustment mode is to derive the adjustment amount based on the deviation of the fusion result and the target quality threshold. When the predicted hardness value is lower than the target hardness value, the hardness index is optimized to adjust the holding time, and the adjustment amount calculation formula is: , wherein, as the increased holding time, as the target hardness value, as the hardness value predicted by the fusion model, as the material coefficient, as the temperature correction coefficient.

[0071] Further, the specific implementation is as follows: the timing smoothing processing is aimed at the instantaneous fluctuation in the fusion result, and the sliding average value is calculated for the continuous fusion results, such as 84.1 points, 83.9 points, 85.2 points, 84.5 points, 83.8 points and 84.7 points, to obtain 84.4 points, so as to eliminate the result jump caused by the instantaneous noise of the sensor.

[0072] The process of the abnormality detection is based on the smoothed fusion result, and a deviation threshold is set in combination with historical data, such as ±5 points, when a batch of fusion results is 78 points, lower than the threshold 80 points, it is determined as an abnormal value. When correcting, the median of the adjacent 5 normal batch results is replaced, for example, the adjacent batch results are 84.2 points, 84.5 points, 84.0 points, 84.7 points and 83.9 points, the median is 84.2 points, and the abnormal value 78 points is replaced, so as to ensure the continuity of the data track.

[0073] When the hardness value predicted by the fusion model is lower than the target value , the holding time for the hardness index is adjusted, and the adjustment amount is calculated according to the following formula: , wherein, is the material coefficient (1.2 for carbon steel), is the temperature correction coefficient (1.1 when the actual temperature is 5 DEG C lower than the standard temperature), and the holding time that needs to be increased is obtained by substitution , that is, it is suggested that the holding time is increased by about 2 minutes.

[0074] Secondly, only the structures related to the disclosed embodiments are involved in the drawings of the disclosed embodiments, other structures can be referred to the general design, and the same embodiments and different embodiments of the present application can be combined with each other under the condition of no conflict; Finally, the above-mentioned is only the preferred embodiment of the present application, and is not used for limiting the present application, any modification, equivalent replacement, improvement and the like made within the spirit and principle of the present application should be included in the protection scope of the present application.​​

Claims

1. A multi-source data fusion processing method for fastener heat treatment, characterized in that, Includes the following steps: S1. Multi-source data acquisition: The fastener to be processed is recorded as the target object. Multi-source data of the target object during the heat treatment process is collected in real time through the multi-source data acquisition terminal. The multi-source data includes process data, material data, quality inspection data and environmental data. S2. Data Preprocessing: Multi-source data is cleaned and standardized through a data preprocessing terminal to obtain process data preprocessing parameters, material data preprocessing parameters, quality inspection data preprocessing parameters, and environmental data preprocessing parameters. S3. Feature Extraction: Input the process data preprocessing parameters, material data preprocessing parameters, quality inspection data preprocessing parameters, and environmental data preprocessing parameters into the feature extraction terminal, and output the time series feature parameters, composition feature parameters, quality feature parameters, and interference feature parameters; S4. Fusion Modeling: The component feature parameters, quality feature parameters and interference feature parameters are input into the fusion model through the fusion modeling terminal. The feature layer is fused by combining the mechanism prior weight and the data-driven attention mechanism. Then, the decision layer is fused through the multi-quality index sub-model, and the fusion result is output. S5. Result Optimization: The fusion results are processed through the result optimization terminal, and the heat treatment quality assessment results of the target object are finally output. The heat treatment quality assessment results are judged, and if the preset results are not met, the process parameters are adjusted.

2. The multi-source data fusion processing method for fastener heat treatment according to claim 1, characterized in that: The process data includes furnace temperature, cooling medium flow rate, furnace pressure, equipment operating power, and operating time; the material data includes the material composition content and original hardness of the target object; the quality inspection data includes the hardness, tensile strength, metallographic image, and deformation after heat treatment; and the environmental data includes workshop temperature, humidity, and power grid voltage.

3. The multi-source data fusion processing method for fastener heat treatment according to claim 1, characterized in that: The time-series characteristic parameters include temperature change rate, pressure fluctuation amplitude, mean flow rate, and power standard deviation; The component characteristic parameters include the content ratio of each element and the component uniformity; the quality characteristic parameters include hardness gradient, strength dispersion, and peak deformation. The interference characteristic parameters include temperature fluctuation range and voltage stability coefficient.

4. The multi-source data fusion processing method for fastener heat treatment according to claim 3, characterized in that: The rate of temperature change is calculated using the temperature values ​​from five consecutive sampling points, using the following formula: ， in, This is the temperature value at the current sampling point. This is the average temperature of the first four sampling points. The sampling time interval, The temperature change rate is used; the component uniformity is calculated by the ratio of the standard deviation to the mean of the content of each element.

5. The multi-source data fusion processing method for fastener heat treatment according to claim 1, characterized in that: The mechanism prior weights are set based on the correlation between austenitizing temperature and holding time in the heat treatment of fasteners. The data-driven attention mechanism learns the feature importance coefficients, and the fused feature vector is the sum of the products of various feature parameters and their corresponding weights.

6. The multi-source data fusion processing method for fastener heat treatment according to claim 1, characterized in that: The multi-quality index sub-model includes a hardness uniformity sub-model, a strength stability sub-model, a microstructure conformity sub-model, and a dimensional accuracy sub-model.

7. A multi-source data fusion processing method for fastener heat treatment according to claim 6, characterized in that: The hardness uniformity sub-model takes hardness gradient features as its core input and outputs a hardness distribution uniformity score, as shown in the formula: ， in This represents the hardness gradient value. This is the hardness influence coefficient. It is a natural constant; The strength stability sub-model outputs a batch strength consistency coefficient based on the strength dispersion characteristic, using the following formula: ， in For strength standard deviation, This represents the maximum permissible standard deviation. The microstructure qualification sub-model, combining grain diameter and martensite phase ratio characteristics, outputs the microstructure qualification rate, as shown in the formula: ， Where d is the actual grain diameter. These are the upper and lower limits of the grain diameter, respectively. The proportion of martensite phase; The dimensional accuracy sub-model outputs the dimensional deviation level based on the peak deformation characteristics, using the following formula: ， in This represents the peak value of the deformation. This is the deformation threshold; The output values ​​of each sub-model are assigned preset weights. Weighted summation, where The weights are preset and the sum is 1.

8. The multi-source data fusion processing method for fastener heat treatment according to claim 1, characterized in that: The data processing procedure is as follows: S51. Calculate the mean value from multiple consecutive fusion results; S52. Perform anomaly detection on the mean to identify outliers; S53. Replace outliers with the median of multiple adjacent results.

9. A multi-source data fusion processing method for fastener heat treatment according to claim 8, characterized in that: The anomaly detection process is constructed based on the temporal distribution characteristics of the fusion results. The normal fluctuation range is determined by calculating the mean and standard deviation of N consecutive fusion results. When a fusion result exceeds 1.5 times the standard deviation of the range, it is judged as an anomaly. Where N is the preset sliding window size, and the normal fluctuation range is dynamically adjusted as the data within the window is updated.

10. A multi-source data fusion processing method for fastener heat treatment according to claim 1, characterized in that: The process parameters are adjusted by deriving the adjustment amount based on the deviation between the fusion result and the target quality threshold. When the predicted hardness value is lower than the target hardness value, the heat preservation time for optimizing the hardness index is adjusted. The formula for calculating the adjustment amount is as follows: ， in, To increase the required insulation time, The target hardness value, The hardness value is predicted by the fusion model. For material coefficient, This is the temperature correction factor.

Images