An electronic component intelligent production control method and system, an electronic device, and a storage medium

Abstract

The application provides an electronic component intelligent production control method and system, electronic equipment and storage medium, and relates to the technical field of electronic component intelligent manufacturing. The method comprises the following steps: synchronously collecting second-level power consumption data, temperature field distribution information and environmental temperature and humidity data of a chip mounter, a reflow soldering device and an optical detection device, and fusing to generate multi-source time sequence flow data. Thermal imaging data is generated based on the temperature field distribution to identify abnormal power consumption areas. Combined with the multi-source time sequence flow data, cross-device cycle correlation features are extracted through an inflation convolution operation to predict future energy efficiency state change points. The time sequence target cascade algorithm is used to decompose the global energy efficiency target into chip mounter start-stop time sequence constraints, reflow soldering power parameter constraints and optical detection sampling frequency constraints to realize multi-device collaborative regulation. The application improves the energy efficiency management level and product yield stability of the electronic component production line.

Description

Technical Field

[0001] This application relates to the field of intelligent manufacturing technology for electronic components, and in particular to an intelligent production control method, system, electronic device, and storage medium for electronic components. Background Technology

[0002] In the intelligent manufacturing process of electronic components, efficient collaboration and energy consumption optimization of production equipment are key to improving production efficiency and product quality. As SMT assembly lines develop towards higher precision and higher integration, the energy consumption fluctuations and thermal management issues between equipment are becoming increasingly prominent. There is an urgent need for an intelligent control method that can monitor the operating status of multiple devices in real time, dynamically adjust production parameters, and achieve global energy efficiency optimization to ensure production stability and reduce energy costs.

[0003] Currently, some advanced production lines employ independent control strategies based on individual equipment energy consumption models. This involves real-time collection of operational data from key equipment and adjustments to local parameters based on preset energy consumption thresholds. For example, for reflow soldering equipment, temperature sensors monitor heat distribution in critical areas, and heating power is adjusted according to fixed rules to maintain temperature stability. Simultaneously, some systems incorporate time-series data analysis to make short-term predictions of equipment energy consumption trends, assisting in human decision-making.

[0004] This solution relies on independent optimization of individual devices, resulting in poor energy consumption coordination between devices and limited overall energy efficiency improvement. Due to the lack of deep fusion of multi-source data, the accuracy of identifying abnormal power consumption areas is insufficient, making it difficult to support accurate prediction of long-term energy efficiency trends. Furthermore, the parameter adjustment flexibility under fixed rules is low, failing to adapt to dynamic changes in complex production scenarios and affecting the yield stability of the final product. Summary of the Invention

[0005] This application provides an intelligent production control method, system, electronic device, and storage medium for electronic components, which aims to solve the problems of poor energy efficiency management and low product yield stability in existing electronic component production lines.

[0006] To address the aforementioned technical problems, in a first aspect, this application provides an intelligent production control method for electronic components, comprising: The system collects second-level power consumption data, temperature field distribution information of multiple devices, and ambient temperature and humidity data of the production environment during the electronic component manufacturing process. The multiple devices include a chip mounter, a reflow soldering machine, and an optical inspection machine. By fusing the second-level power consumption data with the ambient temperature and humidity data, multi-source time-series stream data is generated; Based on the temperature field distribution information of the multiple devices, thermal imaging data is generated, and abnormal power consumption areas are determined based on the thermal imaging data. Based on the multi-source time-series data and the abnormal power consumption region, periodic correlation features are generated through dilated convolution operations. Based on the periodic correlation features, the energy efficiency state change points of future time windows are predicted. Based on the energy efficiency state change points, the global energy efficiency target is decomposed into the process unit level of the chip mounter, the reflow soldering equipment, and the optical inspection equipment using a time-series target cascade algorithm, thereby generating unit-level energy efficiency constraints. Based on the unit-level energy efficiency constraints, the start-stop sequence of the chip mounter, the power parameters of the reflow soldering equipment, and the sampling frequency of the optical inspection equipment are adjusted in a coordinated manner to achieve optimal power consumption and yield in the electronic component manufacturing process.

[0007] Optionally, the step of generating periodic correlation features based on the multi-source time-series data and the abnormal power consumption region through dilated convolution operations, and predicting energy efficiency state change points in future time windows based on the periodic correlation features, includes: Extract the power consumption sequences of the reflow soldering equipment, the pick-and-place machine, and the optical inspection equipment corresponding to the abnormal power consumption regions from the multi-source time-series data. The power consumption sequences of the reflow soldering equipment, the pick-and-place machine, and the optical inspection equipment are horizontally spliced ​​together. Perform dilated convolution operation on the spliced ​​multi-device power consumption sequence to capture the periodic correlation features of cross-device power consumption fluctuations; Based on the aforementioned periodic correlation characteristics, energy efficiency status change points for N future production batch time windows are generated, where N is greater than or equal to 3.

[0008] Optionally, performing a dilated convolution operation on the spliced ​​multi-device power consumption sequence to capture the periodic correlation features of cross-device power consumption fluctuations includes: Define a three-level dilated convolution kernel structure; By scanning the spliced ​​multi-device power consumption sequence at second-level intervals using the first-level convolutional kernel, the first-cycle fluctuation characteristics within adjacent production batches are captured. The second-level convolutional kernel scans the spliced ​​multi-device power consumption sequence at minute-level intervals to capture the second-cycle variation features across batches. The third-level convolutional kernel scans the spliced ​​multi-device power consumption sequence at hourly intervals to capture the third-cycle trend characteristics of the daily production plan. The first periodic fluctuation feature, the second periodic change feature, and the third periodic trend feature are vertically superimposed to generate a multi-scale feature combination of cross-device power consumption fluctuation. Based on the weight allocation results of features at each level in the multi-scale feature combination, periodic correlation features are generated.

[0009] Optionally, generating periodic correlation features based on the weight allocation results of features at each level in the multi-scale feature combination includes: Based on the multi-scale feature combination, the variance contribution of each feature is calculated respectively; Based on the proportion of the variance contribution of each feature, the first weight value of the first periodic fluctuation feature, the second weight value of the second periodic change feature, and the third weight value of the third periodic trend feature are dynamically allocated. The first weighted feature is obtained by multiplying the first periodic fluctuation feature by the first weight value; the second weighted feature is obtained by multiplying the second periodic change feature by the second weight value; and the third weighted feature is obtained by multiplying the third periodic trend feature by the third weight value. The first weighted feature, the second weighted feature, and the third weighted feature are superimposed. The time dimension of the superimposed results is smoothed to generate periodic correlation features.

[0010] Optionally, based on the energy efficiency state change points, the global energy efficiency target is decomposed into process unit levels of the chip mounter, the reflow soldering equipment, and the optical inspection equipment using a time-series target cascade algorithm to generate unit-level energy efficiency constraints, including: A global energy efficiency target is generated based on the deviation between the preset unit output power consumption benchmark and the current actual power consumption. The time interval for target decomposition is determined based on the time coordinates of the energy efficiency state change points. Within the time interval, the global energy efficiency target is broken down into the maximum allowable start-stop number constraints of the pick-and-place machine mounting unit, the upper limit constraint of the power fluctuation of the reflow soldering equipment temperature control unit, and the minimum sampling interval constraint of the optical inspection equipment scanning unit according to the preset equipment function weights. The maximum allowable number of start-stop cycles, the upper limit of power fluctuation, and the minimum sampling interval constraint together constitute the unit-level energy efficiency constraint.

[0011] Optionally, the step of fusing the second-level power consumption data with the ambient temperature and humidity data to generate multi-source time-series stream data includes: The second-level power consumption data of the pick-and-place machine is marked as the first timing sequence, the second-level power consumption data of the reflow soldering equipment is marked as the second timing sequence, and the second-level power consumption data of the optical inspection equipment is marked as the third timing sequence. The environmental temperature and humidity data are split into a temperature time series and a humidity time series according to the collection timestamp; Perform timestamp alignment on the first time series, the second time series, the third time series, the temperature time series, and the humidity time series; The aligned five types of time series are concatenated into multi-source time series stream data according to the device dimension.

[0012] Optionally, generating thermal imaging data based on the temperature field distribution information of the multiple devices, and determining abnormal power consumption areas based on the thermal imaging data, includes: The surface temperature distribution data of the chip mounter, the surface temperature distribution data of the reflow soldering equipment, and the surface temperature distribution data of the optical inspection equipment are separated from the temperature field distribution information of the multiple devices. The surface temperature distribution data of each device are divided into a two-dimensional matrix containing the temperature values ​​of all monitoring points according to the preset monitoring grid division rules, and the two-dimensional matrix is ​​used as thermal imaging data. The thermal imaging data is used to calculate the temperature difference between adjacent monitoring grids to obtain the temperature change intensity value of each monitoring grid. When the temperature change intensity value of the same monitoring grid exceeds the preset fluctuation threshold in N consecutive collection cycles, the monitoring grid is marked as an abnormal power consumption area.

[0013] Secondly, this application provides an intelligent production control system for electronic components, comprising: The data acquisition module is used to collect second-level power consumption data of multiple devices, temperature field distribution information of multiple devices, and ambient temperature and humidity data of the production environment during the production process of electronic components. The multiple devices include a chip mounter, a reflow soldering machine, and an optical inspection machine. The fusion module is used to fuse the second-level power consumption data with the ambient temperature and humidity data to generate multi-source time-series stream data; The generation module is used to generate thermal imaging data based on the temperature field distribution information of the multiple devices, and to determine abnormal power consumption areas based on the thermal imaging data. The prediction module is used to generate periodic correlation features based on the multi-source time-series data and the abnormal power consumption region through dilated convolution operation, and predict the energy efficiency state change points in future time windows based on the periodic correlation features. The decomposition module is used to decompose the global energy efficiency target to the process unit level of the chip mounter, the reflow soldering equipment, and the optical inspection equipment based on the energy efficiency state change points using a time-series target cascading algorithm, thereby generating unit-level energy efficiency constraints. The adjustment module is used to coordinately adjust the start-stop timing of the chip mounter, the power parameters of the reflow soldering equipment, and the sampling frequency of the optical inspection equipment according to the unit-level energy efficiency constraints, so as to achieve the optimal balance between power consumption and yield in the electronic component manufacturing process.

[0014] Thirdly, this application provides an electronic device, comprising: Memory, used to store computer programs; A processor is used to execute the computer program to implement the steps of the intelligent production control method for electronic components as described in the first aspect above.

[0015] Fourthly, this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, can implement the steps of the intelligent production control method for electronic components as described in the first aspect above.

[0016] This application provides an intelligent production control method for electronic components. The method includes: collecting second-level power consumption data from multiple devices during the electronic component production process, temperature field distribution information from multiple devices, and ambient temperature and humidity data of the production environment. The multiple devices include a surface mount device (SMT), a reflow soldering machine, and an optical inspection device. The method also includes fusing the second-level power consumption data with the ambient temperature and humidity data to generate multi-source time-series data; generating thermal imaging data based on the temperature field distribution information from the multiple devices; determining abnormal power consumption regions based on the thermal imaging data; and using the multi-source time-series data and the abnormal power consumption regions to... The dilated convolution operation generates periodic correlation features. Based on these features, the energy efficiency state change points of future time windows are predicted. Based on these energy efficiency state change points, a time-series target cascade algorithm is used to decompose the global energy efficiency target to the process unit level of the pick-and-place machine, the reflow soldering equipment, and the optical inspection equipment, generating unit-level energy efficiency constraints. According to the unit-level energy efficiency constraints, the start-stop timing of the pick-and-place machine, the power parameters of the reflow soldering equipment, and the sampling frequency of the optical inspection equipment are adjusted in a coordinated manner to achieve optimal power consumption and yield in the electronic component manufacturing process.

[0017] The technical solution provided in this application has the following beneficial effects: This application provides a comprehensive and real-time data foundation for energy efficiency optimization by simultaneously collecting second-level power consumption, temperature field distribution, and ambient temperature and humidity data from multiple devices. Heterogeneous data is fused to generate multi-source time-series streaming data, enabling a comprehensive characterization of device operating status and environmental factors. Abnormal power consumption areas are identified based on thermal imaging data analysis, accurately locating energy efficiency anomalies. Long-term correlation features across devices are extracted through dilated convolution, accurately predicting future energy efficiency trends. A time-series target cascade algorithm is used to achieve a refined decomposition of global energy efficiency targets, ensuring the optimization coordination of each device unit. Based on unit-level constraints, device operating parameters are collaboratively optimized to achieve a dynamic balance between energy consumption and yield.

[0018] Furthermore, this application also extracts the device power consumption sequence associated with abnormal regions and performs horizontal stitching, and uses dilated convolution operation to capture the periodic correlation features of cross-device power consumption fluctuations, and predicts the energy efficiency status change points of at least three future production batches based on these features.

[0019] Furthermore, this technology can effectively identify the energy consumption correlation patterns between equipment, accurately predict the energy efficiency change trend of the production line, provide a reliable basis for formulating optimization strategies in advance, and improve the foresight and initiative of energy efficiency management.

[0020] These or other aspects of this application will become more apparent in the following description of the embodiments. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 A flowchart of an intelligent production control method for electronic components provided in this application embodiment; Figure 2 This application provides a schematic diagram illustrating a specific implementation of an intelligent production control method for electronic components. Figure 3 This application provides a schematic diagram illustrating a specific implementation of an intelligent production control method for electronic components. Figure 4 This is a schematic diagram of the structure of an intelligent production control system for electronic components provided in an embodiment of this application. Detailed Implementation

[0023] In the intelligent manufacturing process of electronic components, existing energy efficiency control technologies mainly rely on independent optimization strategies for individual devices, resulting in insufficient energy consumption coordination between devices. Under this model, the accuracy of identifying abnormal power consumption areas is limited, making it difficult to detect potential problems in a timely manner; at the same time, parameter adjustments under fixed rules lack flexibility and cannot adapt to complex dynamic changes during the production process, ultimately affecting production efficiency and product quality.

[0024] To address the aforementioned issues, this application proposes an intelligent production control method for electronic components. This method constructs a comprehensive monitoring system by real-time acquisition of equipment operating data, temperature distribution, and environmental parameters. First, it utilizes thermal imaging technology to accurately locate abnormal power consumption areas. Then, it combines time-series data analysis to predict energy efficiency trends. Finally, it dynamically adjusts the operating parameters of each piece of equipment using intelligent algorithms. This method overcomes the limitations of traditional single-equipment optimization, achieving collaborative control of multiple devices—accurately identifying potential problem areas and flexibly adjusting parameters based on real-time production status. This effectively solves the problems of poor coordination and delayed adjustments in existing technologies, improving the energy efficiency management level of the production line and the stability of product quality.

[0025] To enable those skilled in the art to better understand the present application, the present application will be further described in detail below with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are merely some embodiments of the present application, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0026] The core of this application is to provide an intelligent production control method for electronic components, and a flowchart of one specific implementation is shown below. Figure 1 As shown, the method includes: Step 101: Collect second-level power consumption data of multiple devices, temperature field distribution information of multiple devices, and ambient temperature and humidity data of the production environment during the production process of electronic components. The multiple devices include pick-and-place machines, reflow soldering equipment, and optical inspection equipment.

[0027] In step 101, the multi-device second-level power consumption data refers to the power consumption values ​​recorded per second by the pick-and-place machine, reflow soldering equipment, and optical inspection equipment, reflecting the real-time energy consumption status of the equipment. The multi-device temperature field distribution information is two-dimensional temperature distribution data obtained through an array of temperature sensors arranged on the equipment surface, used to monitor the heat generation of the equipment during operation. The ambient temperature and humidity data are the monitored values ​​of air temperature and humidity in the production workshop, affecting equipment heat dissipation and operational stability. The pick-and-place machine, reflow soldering equipment, and optical inspection equipment are processing and testing equipment used in the production of electronic components, not the electronic components themselves. Electronic components refer to final products such as resistors, capacitors, and integrated circuits.

[0028] In this embodiment, power consumption data is collected in real time per second by power meters installed on the pick-and-place machine, reflow soldering equipment, and optical inspection equipment. Simultaneously, a network of temperature sensors evenly distributed across the equipment surface is used to obtain the temperature distribution of each device. Temperature and humidity sensors are also deployed at key locations in the workshop to collect environmental parameters. All data is aggregated through an industrial IoT gateway to form a basic dataset containing equipment operating status and environmental factors.

[0029] For example, in an electronic component manufacturing workshop, a pick-and-place machine records its current operating power every second, such as continuous values ​​like 320 watts or 315 watts; 30 temperature monitoring points are arranged on the surface of the reflow soldering equipment to record the temperature at each point, such as 89.5 degrees Celsius or 88.7 degrees Celsius; at the same time, temperature and humidity sensors in the workshop record environmental parameters such as 26.5 degrees Celsius and 45% humidity. This data is uploaded to the control center in real time through the workshop's data acquisition system.

[0030] Step 102: Merge the second-level power consumption data with the ambient temperature and humidity data to generate multi-source time-series stream data.

[0031] In step 102, the multi-source time series data is a comprehensive dataset formed by aligning and integrating time series data from different sources and of different types according to a unified time reference.

[0032] In this embodiment, the second-level power consumption data of the pick-and-place machine, reflow soldering equipment, and optical inspection equipment are first marked as independent time series, and the ambient temperature and humidity data are split into temperature and humidity series. Then, using the unified clock of the equipment control system as a reference, the five time series are time-aligned to ensure that the data at each time point is completely corresponding. Finally, the aligned data is spliced ​​together according to the equipment dimension to form a time series data stream containing multi-dimensional information.

[0033] For example, following the previous example, the control system aligns the power consumption of the pick-and-place machine (320 watts), the power consumption of the reflow soldering equipment (2850 watts), the power consumption of the optical inspection equipment (180 watts), the ambient temperature (26.5 degrees Celsius), and the humidity (45%) at 8:00:00 with a unified time point, and combines them into a complete data record. The data at subsequent time points are also processed in the same way to form a continuous multi-source data stream.

[0034] Step 103: Based on the temperature field distribution information of the multiple devices, generate thermal imaging data, and determine the abnormal power consumption area based on the thermal imaging data.

[0035] In step 103, the thermal imaging data converts the surface temperature distribution information of the device into a visualized two-dimensional matrix representation. Abnormal power consumption regions are identified by analyzing temperature distribution characteristics, pinpointing areas of abnormal heat generation within the device. This region may correspond to a local component area of ​​a malfunctioning device (rather than the entire device). For example, an abnormal power consumption in a drive module of a pick-and-place machine might only create a unique thermal gradient characteristic at its local location, corresponding to a specific small area. This region may involve the interaction areas of multiple devices. For example, the superposition of heat dissipation anomalies in a reflow soldering machine and the adjacent optical inspection equipment results in an abnormal thermal gradient in the space between them, forming a cross-device abnormal power consumption region. This region may also be related to the interaction area between the device and its environment. For example, abnormal ambient temperature and humidity, combined with device heat dissipation, can create an abnormal heat distribution in a specific space around the device; this area may not necessarily be entirely within the physical footprint of the device itself.

[0036] In this embodiment, the data from the surface temperature monitoring points of each device are first organized into a two-dimensional matrix according to a preset grid layout. Then, the temperature difference between each grid point and its adjacent points is calculated to obtain the intensity of local temperature changes. When the temperature change of a certain grid point exceeds a set threshold for multiple consecutive cycles, the area is determined to be an abnormal power consumption area, and its location information is recorded.

[0037] For example, using the reflow soldering equipment data from the previous example, the temperature values ​​of 30 monitoring points are arranged into a 5x6 matrix. The temperature difference between point 15 (89.5 degrees Celsius) and its adjacent points is calculated. It is found that the temperature difference exceeds the 5-degree Celsius threshold for three consecutive cycles. Therefore, this area is marked as an anomaly, and its coordinates are recorded.

[0038] Step 104: Based on the multi-source time-series data and the abnormal power consumption region, generate periodic correlation features through dilated convolution operation, and predict the energy efficiency state change points of future time windows based on the periodic correlation features.

[0039] In step 104, the periodic correlation feature is a time-series feature reflecting the correlation between energy consumption fluctuations among different devices. The energy efficiency state change point is the predicted time node where energy efficiency characteristics may change in the future production process.

[0040] In this embodiment, the device power consumption sequence associated with abnormal regions is first extracted from multi-source time-series streams, and the sequences of the three types of devices are horizontally concatenated. Then, convolutional kernels with different time scales are used for feature extraction: small-scale kernels capture short-term fluctuations, medium-scale kernels analyze inter-batch correlations, and large-scale kernels identify long-term trends. Finally, based on the extracted features, multiple time points at which the energy efficiency status may change during future production processes are predicted.

[0041] For example, the power consumption sequences of the pick-and-place machine, reflow soldering machine, and optical inspection equipment corresponding to the abnormal area are extracted, spliced, and input into the feature extraction module. Through analysis, it is found that the energy consumption of the equipment has periodic fluctuation characteristics such as 25 minutes and 35 minutes. Based on this, it is predicted that the energy efficiency status may change at time points such as the 15th minute and 40th minute in the next production stage.

[0042] Step 105: Based on the energy efficiency state change points, the global energy efficiency target is decomposed into the process unit level of the chip mounter, the reflow soldering equipment, and the optical inspection equipment using the time-series target cascade algorithm, thereby generating unit-level energy efficiency constraints.

[0043] In step 105, the unit-level energy efficiency constraint is the limitation of the specific operating parameters of each device obtained from the decomposition of the global objective.

[0044] In this embodiment, the optimization time interval is first determined based on the predicted energy efficiency change points. Then, the overall energy-saving target is allocated according to equipment importance, translating into start-stop limits for the placement machine, power fluctuation range for reflow soldering, and sampling interval requirements for optical inspection. This ultimately forms the operating parameter constraints that each device must adhere to within the specified time interval.

[0045] For example, for the predicted energy efficiency change points, the system breaks down the overall energy saving target into specific constraints such as: the pick-and-place machine should not start and stop more than 5 times in the next 30 minutes, the reflow soldering power fluctuation should be controlled within ±3%, and the optical inspection sampling interval should not be less than 0.4 seconds.

[0046] Step 106: Based on the unit-level energy efficiency constraints, coordinately adjust the start-stop sequence of the placement machine, the power parameters of the reflow soldering equipment, and the sampling frequency of the optical inspection equipment to achieve optimal power consumption and yield in the electronic component manufacturing process.

[0047] In step 106, the start-stop sequence of the pick-and-place machine refers to controlling the working rhythm and pause time arrangement of the pick-and-place machine at different production stages. Its value is derived from the maximum allowable number of start-stop cycles specified in the unit-level energy efficiency constraints. It is calculated by analyzing the current production workload and predicted energy efficiency state change points to ensure energy consumption optimization while meeting production needs. The power parameters of the reflow soldering equipment refer to the range of power input required to control the temperature of the heating zone. Its value is determined according to the upper limit of power fluctuation set in the unit-level energy efficiency constraints. It is dynamically adjusted by real-time monitoring of temperature distribution and predicted energy efficiency change trends to achieve energy consumption optimization while ensuring soldering quality. The sampling frequency of the optical inspection equipment refers to the time interval for quality inspection of electronic components. Its minimum value is directly specified by the unit-level energy efficiency constraints. It is dynamically adjusted by analyzing the current production speed and predicted energy efficiency state change points to reduce unnecessary energy consumption while ensuring product quality inspection effects. Cooperative optimization refers to achieving the best balance between overall energy consumption and product quality by coordinating the operating parameters of each piece of equipment.

[0048] In this embodiment, the working rhythm of the pick-and-place machine, the heating power of the reflow soldering, and the scanning frequency of the optical inspection are adjusted in real time based on the constraints obtained from the decomposition. By dynamically balancing the operating status of each device, energy consumption optimization is achieved while ensuring product quality, ultimately achieving the best balance between production efficiency and energy utilization.

[0049] For example, the control system adjusts equipment operation according to constraints: slowing down the operation pace of some stations on the pick-and-place machine, fine-tuning the power output of the reflow soldering heating zone, and reducing the scanning frequency of the optical inspection equipment. Through these adjustments, the production line significantly reduces overall energy consumption while maintaining product qualification rates.

[0050] This method accurately identifies abnormal problem areas, predicts energy efficiency change trends, and intelligently adjusts equipment operating parameters by real-time monitoring of equipment operating status and environmental parameters. It achieves synergistic optimization of energy consumption control and quality assurance in the production process of electronic components, thereby improving the energy efficiency management level and operational stability of the production line.

[0051] To address the issue of coordinated energy consumption optimization across multiple devices during electronic component manufacturing, in some embodiments, step 104 involves generating periodic correlation features based on the multi-source time-series data and the abnormal power consumption region through dilated convolution operations, and predicting energy efficiency state change points for future time windows based on these periodic correlation features. This includes: Step 201: Extract the power consumption sequence of the reflow soldering equipment, the power consumption sequence of the pick-and-place machine, and the power consumption sequence of the optical inspection equipment corresponding to the abnormal power consumption region from the multi-source time-series data.

[0052] In step 201, the device power consumption sequence corresponding to the abnormal power consumption region refers to the power consumption data sequence of the pick-and-place machine, reflow soldering equipment, and optical inspection equipment that are temporally and spatially associated with the marked abnormal region, selected from multi-source data. These sequences reflect the changes in the operating status of each device when the abnormal region appears.

[0053] In this embodiment, the corresponding data segment is first located in the multi-source time-series data based on the location and time information of the identified abnormal power consumption area. Then, continuous power consumption readings of the three types of devices within the time period are extracted to form three independent time-series data chains, which retain the time correspondence of the original data.

[0054] Step 202: Horizontally splice the power consumption sequence of the reflow soldering equipment, the power consumption sequence of the chip mounter, and the power consumption sequence of the optical inspection equipment.

[0055] In step 202, horizontal splicing refers to the operation of connecting time-series data from different devices side-by-side according to the same time point, forming a composite data sequence containing information from multiple devices. This splicing method maintains the time synchronization between the data from each device.

[0056] In this embodiment, the three extracted device power consumption sequences are aligned along the time axis, with the three device data points at each time point forming a data unit, and arranged in chronological order to form a new composite sequence. This structure facilitates subsequent analysis of the energy consumption correlation between devices.

[0057] Step 203: Perform dilated convolution operation on the spliced ​​multi-device power consumption sequence to capture the periodic correlation features of cross-device power consumption fluctuations.

[0058] In step 203, dilated convolution is a special convolution calculation method that captures features at different time scales by controlling the spacing of the convolution kernels.

[0059] In this embodiment, multi-scale convolution calculations are applied to the spliced ​​composite sequence. Small-interval convolution kernels capture real-time interactions between devices, medium-interval kernels analyze batch correlations, and large-interval kernels identify long-term trends. Through this hierarchical processing, multi-scale feature combinations reflecting the collaborative working characteristics of the devices are extracted.

[0060] Step 204: Based on the aforementioned periodic correlation features, generate energy efficiency status change points for the next N production batch time windows, where N is greater than or equal to 3.

[0061] In step 204, the production batch time window is a time period division method based on a complete production batch.

[0062] In this embodiment, based on the extracted multi-scale features, the regularity of their changes over time is analyzed, and the inflection point positions on the feature curves are identified. These inflection points correspond to the moments when the energy efficiency status may change in the future production process, and at least three consecutive batches of transition nodes are predicted to ensure the reliability of the decision.

[0063] Here is a specific example: In a practical application at an electronic component manufacturing workshop, when the system detected an abnormal temperature at monitoring point 15 of the reflow soldering equipment, it immediately extracted relevant equipment data from the period between 8:00:00 and 8:05:00: the power consumption sequence for the pick-and-place machine was recorded at 320 watts, 318 watts, and 322 watts per second; the power consumption sequence for the reflow soldering equipment was recorded at 2850 watts, 2865 watts, and 2840 watts per second; and the power consumption sequence for the optical inspection equipment was recorded at 180 watts, 178 watts, and 183 watts per second. These sequences were then aligned and concatenated according to time points to form a composite sequence of 300 data points per minute, with each data point containing the real-time power consumption of the three types of equipment. When performing dilated convolution processing on this composite sequence, the first-level convolution kernel scanned and extracted second-level interaction features between equipment at 1-second intervals; the second level analyzed batch-to-batch correlations at 30-second intervals; and the third level identified production trends at 5-minute intervals. The convolution kernel size was set to 3, and the dilation coefficient increased in increments of 1, 3, and 5. Calculations revealed that the sequence exhibits periodicity, with its principal period T determined by the formula T = 2π / ω, where ω is the dominant frequency component obtained through Fourier transform. Based on this, it was predicted that energy efficiency status changes would occur in the next three production batches at 8:20:00, 8:50:00, and 9:20:00, respectively. These time points were calculated by superimposing the detected principal period T = 30 minutes with the current time.

[0064] In this embodiment of the application, the method achieves accurate prediction of future energy efficiency changes by deeply mining the energy consumption correlation characteristics between devices, providing a reliable basis for formulating optimization strategies in advance, and improving the predictability and initiative of production scheduling.

[0065] To further improve the accuracy of multi-device energy consumption feature extraction, in some embodiments, step 203: performing dilated convolution on the concatenated multi-device power consumption sequence to capture the periodic correlation features of cross-device power consumption fluctuations includes: Step 301: Set the three-level dilated convolution kernel structure.

[0066] In step 301, the three-level dilated convolution kernel structure refers to three sets of feature extraction units with different time scale characteristics, which are used to capture the short-term, medium-term and long-term equipment energy consumption fluctuation patterns, respectively.

[0067] In this embodiment, based on the characteristics of equipment energy consumption fluctuations during the production of electronic components, three sets of convolution calculation units with different time spans are pre-configured. These units are arranged according to a time scale from fine to coarse, forming a hierarchical processing structure.

[0068] Step 302: Using the first-level convolutional kernel, scan the spliced ​​multi-device power consumption sequence at second-level intervals to capture the first-cycle fluctuation characteristics within adjacent production batches.

[0069] In step 302, the first cycle fluctuation feature refers to the detailed features that reflect the second-level energy consumption interaction between devices, and embody the real-time energy consumption changes in a single batch production process.

[0070] In this embodiment, a convolution kernel with the smallest time scale is used to scan the stitched device data second by second to detect the rapid energy consumption response relationship between the pick-and-place machine, reflow soldering equipment and optical inspection equipment, and to capture the instantaneous fluctuation pattern directly caused by production operations.

[0071] Step 303: The second-level convolutional kernel scans the spliced ​​multi-device power consumption sequence at minute-level intervals to capture the second-cycle variation features across batches.

[0072] In step 303, the second cycle change characteristic refers to the mid-period characteristics that reflect the coordination of equipment energy consumption during cross-batch production, and embody the energy consumption adjustment pattern during batch switching.

[0073] In this embodiment of the application, a convolutional kernel with a medium time span is used to scan the data sequence in minutes to analyze the coordinated change trend of power consumption of each device when different production batches alternate, and to identify the energy consumption connection characteristics between batches.

[0074] Step 304: The third-level convolutional kernel scans the spliced ​​multi-device power consumption sequence at hourly intervals to capture the third cycle trend characteristics of the daily production plan.

[0075] In step 304, the third cycle trend characteristic refers to the long-term characteristics that reflect the macro-changes in equipment energy consumption under the daily production plan, and embody the impact of the overall production rhythm on energy consumption.

[0076] In this embodiment, data is sampled at hourly intervals using a convolutional kernel with the largest time span to analyze the overall energy consumption evolution of the equipment group from a longer time dimension and grasp the correlation between production scheduling and energy consumption distribution.

[0077] Step 305: Vertically superimpose the first periodic fluctuation feature, the second periodic change feature, and the third periodic trend feature to generate a multi-scale feature combination of cross-device power consumption fluctuation.

[0078] In step 305, multi-scale feature combination is a comprehensive feature representation formed by integrating features extracted at different time scales according to hierarchical relationships.

[0079] In this embodiment of the application, features extracted at three time scales of seconds, minutes and hours are aligned and superimposed along the time axis to form a composite feature map containing multi-level information, which fully presents the spatiotemporal variation characteristics of device energy consumption.

[0080] Step 306: Generate periodic correlation features based on the weight allocation results of features at each level in the multi-scale feature combination.

[0081] In step 306, the weight allocation result is the feature fusion ratio determined based on the contribution of each scale feature to the energy efficiency prediction.

[0082] In this embodiment of the application, by analyzing the stability and discriminative power of features at each scale, the weight ratio of each feature in the final feature is automatically determined, so that important periodic features can gain greater influence and finally generate representative periodic correlation features.

[0083] Here is a specific example: During the energy efficiency optimization implementation in an electronic component manufacturing workshop, when the system performed multi-scale feature extraction on the power consumption sequence of composite equipment collected from 8:00:00 to 8:05:00, it first configured a three-level dilated convolution kernel. The first-level convolution kernel scanned the data with a basic unit of 1 second, and found that the power consumption of the pick-and-place machine fluctuated by about 2 watts every 3 seconds. At the same time, the power consumption of the reflow soldering equipment followed the fluctuation of the pick-and-place machine by about 15 watts one second later. These second-level interaction features were recorded as the first-cycle fluctuation features. The second-level convolution kernel analyzed the data at 30-second intervals and observed that after each component assembly cycle was completed for about 30 seconds, the power consumption of the three types of equipment decreased synchronously by 3% to 5%. This cross-batch coordination feature was marked as the second-cycle change feature. The third-level convolution kernel scanned with a 1-hour window and identified a long-term trend in which the total power consumption of each device in the third hour after the start of production was 8% higher than the average, which was taken as the third-cycle trend feature. After aligning these three features along the time axis, weight allocation coefficients were calculated based on the stability of each feature. The weight for second-level features was 0.4, for minute-level features 0.4, and for hour-level features 0.2. The weight calculation formula is w_i = σ_i / Σσ, where σ_i represents the variance of the i-th level feature, and Σσ represents the total variance. The final multi-scale feature combination shows that the equipment group exhibits a significant tendency for energy efficiency state transitions at three time points: 8:20:00, 8:50:00, and 9:20:00. These predicted time points were calculated by substituting the detected main cycle feature T = 30 minutes into the periodic event prediction model. The accuracy of these predictions matches the energy efficiency fluctuation times in the actual production records of the workshop with over 90% consistency, providing precise time references for production line energy efficiency optimization.

[0084] In this embodiment of the application, the method achieves a comprehensive understanding of the energy consumption patterns of equipment groups through multi-timescale hierarchical feature extraction and intelligent fusion, providing accurate feature basis for energy efficiency optimization and improving the reliability and timeliness of energy consumption prediction.

[0085] To further improve the accuracy and reliability of periodic correlation features, in some embodiments, step 306: generating periodic correlation features based on the weight allocation results of features at each level in the multi-scale feature combination, includes: Step 401: Based on the multi-scale feature combination, calculate the variance contribution of each feature respectively.

[0086] In step 401, the variance contribution refers to the proportion of the fluctuation of each scale feature in the overall feature combination, reflecting the magnitude of the feature's contribution to the overall change.

[0087] In the embodiments of this application, the fluctuation amplitudes of features at the second, minute, and hour levels are calculated respectively, and their importance in the complete feature combination is quantified by comparing the degree of dispersion of each feature.

[0088] Step 402: Based on the proportion of variance contribution of each feature, dynamically allocate the first weight value of the first periodic fluctuation feature, the second weight value of the second periodic change feature, and the third weight value of the third periodic trend feature.

[0089] In step 402, the variance contribution ratio refers to determining the relative importance of each feature in the overall feature combination by calculating the proportion of the variance value of each scale feature (second-level, minute-level, hour-level) to the total variance of all features. The specific calculation process is as follows: first, calculate the variance value of each scale feature; then, divide the variance value of a single feature by the sum of the variance values ​​of all features to obtain the contribution ratio of that feature. This ratio reflects the degree of influence of features at different time scales on overall energy consumption fluctuations; the larger the value, the more critical that scale feature is to the formation of the final cycle-related features. Dynamic weight allocation refers to automatically adjusting the influence ratio of each scale feature in the final result based on feature importance, giving key features greater weight.

[0090] In this embodiment, the variance of each feature is divided by the total variance to obtain the weight ratio of that feature, ensuring that important features have a greater impact on the final result, while keeping the sum of the weights a fixed value.

[0091] Step 403: Multiply the first periodic fluctuation feature by the first weight value to obtain the first weighted feature, multiply the second periodic change feature by the second weight value to obtain the second weighted feature, and multiply the third periodic trend feature by the third weight value to obtain the third weighted feature.

[0092] In step 403, the weighted feature is a new feature representation obtained by multiplying the original feature by its importance weight, which highlights the role of the key feature.

[0093] In this embodiment, the second-level feature is multiplied by its weight to obtain the enhanced second-level feature. The same method is used to process the minute-level and hour-level features, so that the intensity of each scale feature is reasonably adjusted according to its importance.

[0094] Step 404: Perform a superposition operation on the first weighted feature, the second weighted feature, and the third weighted feature.

[0095] In step 404, the superposition operation refers to aligning the weighted features of different scales along the time axis and then adding them together to form a comprehensive feature representation.

[0096] In this embodiment of the application, the adjusted second-level, minute-level, and hour-level features are added together at the same time point to generate a composite feature sequence containing multi-scale information.

[0097] Step 405: Perform time-dimension smoothing on the superimposed results to generate periodic correlation features.

[0098] In step 405, time-dimensional smoothing is achieved by using filtering methods to eliminate random fluctuations in the feature sequence while preserving the main trend of change.

[0099] In this embodiment, a moving average is calculated on the superimposed feature sequence to remove short-term noise interference, making the feature curve smoother and more stable, and finally outputting a periodic correlation feature with clear physical meaning.

[0100] Here is a specific example: In the energy efficiency optimization process of an electronic component manufacturing workshop, when the system allocates weights based on the aforementioned multi-scale feature combination, it first calculates the variance of the second-level fluctuation feature (0.16), reflecting the stability of the second-level interaction between the placement machine and the reflow soldering equipment; the variance of the minute-level change feature (0.36), reflecting the degree of energy consumption coordination between batches; and the variance of the hour-level trend feature (0.08), representing the fluctuation amplitude of the long-term production trend. The total variance of the three is 0.6. According to the weight calculation formula w_i=σ_i / Σσ, where σ_i is the variance of each feature and Σσ is the total variance, the calculated weights are: second-level feature weight 0.16 / 0.6≈0.27, minute-level weight 0.36 / 0.6=0.6, and hour-level weight 0.08 / 0.6≈0.13. The initial composite feature sequence was obtained by multiplying the 3-second periodic features detected at the second level by 0.27, the 30-second periodic features identified at the minute level by 0.6, and the 3-hour trend detected at the hour level by 0.13. A 5-minute moving average was applied to this sequence to eliminate instantaneous fluctuations during equipment start-up and shutdown. The resulting periodic correlation features clearly showed significant energy efficiency state transition points at times such as 8:20:00 and 8:50:00. These time points highly coincided with key change times in the actual energy consumption records of the workshop, verifying the effectiveness of the weighting method.

[0101] In this embodiment, the method scientifically quantifies and rationally integrates the importance of features at various scales, and the generated periodic correlation features can accurately reflect the essential laws of energy consumption of equipment groups, providing a reliable basis for energy efficiency optimization decisions and improving the accuracy and effectiveness of production scheduling.

[0102] To further improve the accuracy and operability of energy efficiency target decomposition, in some embodiments, step 105: based on the energy efficiency state change points, the global energy efficiency target is decomposed to the process unit level of the chip mounter, the reflow soldering equipment, and the optical inspection equipment using a time-series target cascading algorithm to generate unit-level energy efficiency constraints, including: Step 501: Generate a global energy efficiency target based on the deviation between the preset unit output power consumption benchmark and the current actual power consumption.

[0103] In step 501, the unit output power consumption benchmark value refers to the standard energy value consumed in producing one unit quantity of electronic components, reflecting the energy efficiency design level of the production line. The current actual power consumption refers to the energy value actually consumed in producing the same quantity of products, as monitored in real time. The global energy efficiency target is the overall energy-saving requirement determined based on the difference between the two.

[0104] In this embodiment of the application, by comparing the standard energy consumption index of the production line with the actual operating data, the total energy consumption that needs to be optimized in the current production process is calculated, forming a clear overall energy-saving target value, which provides a basis for subsequent equipment-level decomposition.

[0105] Step 502: Determine the time interval for target decomposition based on the time coordinates of the energy efficiency state change points.

[0106] In step 502, the time coordinate of the energy efficiency state change point refers to the specific moment when the predicted energy efficiency characteristics change. The time interval of the target decomposition is the optimization period defined based on these key time points.

[0107] In this embodiment of the application, multiple energy efficiency state change points obtained by analysis and prediction are selected, and the time period between adjacent change points is selected as the basic optimization interval to ensure that the energy efficiency characteristics remain relatively stable within each interval, which facilitates the formulation of targeted optimization strategies.

[0108] Step 503: Within the time interval, the global energy efficiency target is divided into the maximum allowable start-stop number constraint of the pick-and-place machine mounting unit, the upper limit constraint of the power fluctuation of the reflow soldering equipment temperature control unit, and the minimum sampling interval constraint of the optical inspection equipment scanning unit according to the preset equipment function weight.

[0109] In step 503, the equipment function weight is an allocation coefficient determined based on the importance of each piece of equipment in the production process and its energy consumption proportion. The maximum allowable start-stop frequency constraint limits the start-stop operation frequency of the pick-and-place machine within a specified time period. The power fluctuation upper limit constraint specifies the maximum range of power adjustment for the reflow soldering equipment. The minimum sampling interval constraint sets the minimum time interval between two inspections by the optical inspection equipment.

[0110] In this embodiment, the total energy-saving target is allocated to three types of equipment according to a preset ratio based on the equipment type and process characteristics, and is converted into parameter constraints that can be executed by each process unit, so as to ensure that the optimization measures meet the energy-saving requirements without affecting normal production.

[0111] Step 504: The maximum allowable number of start-stop cycles constraint, the power fluctuation upper limit constraint, and the minimum sampling interval constraint are combined to form a unit-level energy efficiency constraint.

[0112] In this embodiment, the parameter constraints of the three types of equipment are integrated to form a complete constraint scheme, which serves as the direct basis for subsequent adjustment of equipment parameters.

[0113] Here is a specific example: During the energy efficiency optimization process in an electronic component manufacturing workshop, the system calculated a global energy efficiency target of 3 kWh to be optimized based on a preset standard energy consumption of 50 kWh per 1,000 products and the actual monitored energy consumption of 53 kWh. Based on the previously predicted energy efficiency status change points of 8:15:00 and 8:45:00, a 30-minute time interval was determined as the target decomposition time. According to the equipment function weight ratio of 40% for pick-and-place machines, 35% for reflow soldering, and 25% for optical inspection, the total target of 3 kWh was decomposed into 1.2 kWh to be saved for pick-and-place machines, 1.05 kWh for reflow soldering, and 0.75 kWh for optical inspection. Based on the equipment characteristics, the energy-saving target for the pick-and-place machine is transformed into a constraint of a maximum of 6 start-stop cycles within 30 minutes. This value is calculated by dividing 1.2 kWh by the average energy consumption per start-stop cycle of 0.2 kWh. The energy-saving target for reflow soldering is transformed into a constraint of power fluctuation not exceeding ±3.5%, obtained by dividing 1.05 kWh by the rated power of the equipment of 30 kW. The energy-saving target for optical inspection is transformed into a constraint of a minimum sampling interval of 0.45 seconds, determined by dividing 0.75 kWh by the energy consumption per unit time of inspection of 1.67 kWh / second. These constraints together constitute the unit-level energy efficiency optimization scheme for the time period from 8:15:00 to 8:45:00.

[0114] In this embodiment of the application, the method achieves precise conversion of global energy efficiency requirements into specific equipment parameters through scientific target decomposition and constraint generation, providing an operable execution standard for dynamic energy efficiency optimization of the production line, and effectively improving the refinement and effectiveness of energy efficiency management.

[0115] To further improve the accuracy and completeness of multi-source data fusion, in some embodiments, step 102: fusing the second-level power consumption data with the environmental temperature and humidity data to generate multi-source time-series stream data includes: Step 601: Mark the second-level power consumption data of the pick-and-place machine as the first timing sequence, the second-level power consumption data of the reflow soldering equipment as the second timing sequence, and the second-level power consumption data of the optical inspection equipment as the third timing sequence.

[0116] In step 601, the first time series refers to the set of continuous power consumption data recorded by the pick-and-place machine at the second level, reflecting the real-time energy consumption changes of the placement process. The second time series is the set of power consumption data at the second level of the reflow soldering equipment, reflecting the energy consumption characteristics of the soldering process. The third time series is the set of power consumption data at the second level of the optical inspection equipment, recording the energy consumption fluctuations in the quality inspection stage.

[0117] In this embodiment, the power monitoring module built into the device collects the power consumption readings per second of the pick-and-place machine, the reflow soldering equipment, and the optical inspection equipment, forming three independent but time-synchronized time-series data chains, laying the foundation for subsequent multi-source data fusion.

[0118] Step 602: The environmental temperature and humidity data are split into temperature time series and humidity time series according to the collection timestamp.

[0119] In step 602, the temperature time series is a set of air temperature data recorded by the ambient temperature sensor every second. The humidity time series is a set of air humidity data recorded by the ambient humidity sensor every second.

[0120] In this embodiment, the raw data from the temperature and humidity sensor is extracted from the workshop environment monitoring system, and the composite temperature and humidity signal is separated into two independent time series according to the acquisition time, maintaining the same time resolution as the equipment power consumption data.

[0121] Step 603: Perform timestamp alignment on the first time series, the second time series, the third time series, the temperature time series, and the humidity time series.

[0122] In step 603, the timestamp alignment operation is the process of matching and calibrating time-series data from different sources according to a unified time reference.

[0123] In this embodiment of the application, the time stamps of the five time series are calibrated based on the main control clock of the production line to ensure that each data point can accurately correspond to the same physical time and eliminate the time deviation caused by the asynchronous clock of the acquisition device.

[0124] Step 604: Concatenate the five aligned time series into multi-source time series stream data according to the device dimension.

[0125] In step 604, device-dimensional splicing refers to the operation of horizontally combining the time-series data of different devices at the same time point.

[0126] In this embodiment of the application, the five types of data after alignment are combined at each time point to form a composite data record including pick-and-place machine power consumption, reflow soldering power consumption, optical inspection power consumption, ambient temperature and ambient humidity, which are arranged in chronological order to form a complete multi-source time-series stream data.

[0127] Here is a specific example: During data acquisition in an electronic component manufacturing workshop, the system synchronously records the power consumption of the placement machine (320 watts) as the starting value of the first time series, the power consumption of the reflow soldering machine (2850 watts) as the starting value of the second time series, and the power consumption of the optical inspection equipment (180 watts) as the starting value of the third time series at 8:00:00. Simultaneously, it acquires temperature (26.5 degrees Celsius) and humidity (45%) data from environmental sensors, splitting them into starting values ​​for temperature and humidity time series. These data are timestamped using the time synchronization module of the workshop's central controller to ensure that all data precisely corresponds to the time of 8:00:00. The aligned five data categories are then concatenated in the order of [placement machine power consumption, reflow soldering power consumption, optical inspection power consumption, ambient temperature, ambient humidity] to form a complete data record for that moment [320, 2850, 180, 26.5, 45]. At the next moment of 8:00:01, the system continued to collect power consumption data for the pick-and-place machine (315 watts), reflow soldering (2845 watts), and optical inspection (182 watts), as well as temperature and humidity data of 26.7 degrees Celsius and 46%. After the same processing, a second record [315, 2845, 182, 26.7, 46] was generated. These records are arranged in chronological order to form a continuous multi-source time-series data stream.

[0128] In this embodiment of the application, the method generates a comprehensive dataset containing equipment operating status and environmental factors through strict time alignment and multi-dimensional data fusion, providing a complete and accurate data foundation for subsequent energy efficiency analysis and improving the utilization value and reliability of monitoring data.

[0129] To further improve the accuracy of abnormal power consumption region detection, in some embodiments, step 103: generating thermal imaging data based on the multi-device temperature field distribution information, and determining abnormal power consumption regions based on the thermal imaging data, includes: Step 701: Separate the surface temperature distribution data of the chip mounter, the surface temperature distribution data of the reflow soldering equipment, and the surface temperature distribution data of the optical inspection equipment from the multi-device temperature field distribution information.

[0130] In step 701, the surface temperature distribution data refers to the set of two-dimensional temperature measurements obtained by an array of temperature sensors arranged on the surface of the equipment, which reflects the heat distribution during equipment operation.

[0131] In this embodiment, real-time readings of temperature measurement points on the surfaces of the chip mounter, reflow soldering equipment, and optical inspection equipment are extracted from the temperature monitoring system, and stored according to equipment type, establishing an independent temperature dataset for each type of equipment.

[0132] Step 702: Generate a two-dimensional matrix containing the temperature values ​​of all monitoring points by dividing the surface temperature distribution data of each device according to the preset monitoring grid division rules, and use the two-dimensional matrix as thermal imaging data.

[0133] In step 702, the monitoring grid division rule is a scheme that divides the device surface into several regular regions.

[0134] In this embodiment, based on the physical dimensions of the device and the arrangement of the sensors, the surface of each device is divided into several square areas of the same size. A temperature measuring point is set at the center of each area, and the temperature values ​​of these points are arranged in a neat matrix according to their positional relationship.

[0135] Step 703: Calculate the temperature difference between adjacent monitoring grids on the thermal imaging data to obtain the temperature change intensity value of each monitoring grid.

[0136] In step 703, the temperature change intensity value is a heat flow activity index obtained by calculating the temperature difference between a certain grid point and its adjacent grid points, which reflects the intensity of local heat change.

[0137] In this embodiment of the application, for each grid point in the temperature matrix, the temperature difference between it and its four adjacent points (up, down, left, and right) is calculated, and the maximum difference is taken as the temperature change intensity value of that point to generate a change intensity map reflecting the heat flow distribution on the surface of the device.

[0138] Step 704: When the temperature change intensity value of the same monitoring grid exceeds the preset fluctuation threshold in N consecutive acquisition cycles, the monitoring grid is marked as an abnormal power consumption area.

[0139] In step 704, the preset fluctuation threshold is the allowable temperature fluctuation range set according to the equipment safety operation specifications.

[0140] In this embodiment of the application, the intensity of change of each grid point is continuously monitored. When the intensity of change of a certain point exceeds the safety threshold in multiple consecutive acquisition cycles, it is determined that there is abnormal power consumption in the area, and its position coordinates are marked in the temperature matrix.

[0141] Here is a specific example: During temperature monitoring in an electronic component manufacturing workshop, the system first acquired data from 30 temperature monitoring points on the surface of the reflow soldering equipment, including values ​​such as 89.5 degrees Celsius at point 15, 84.2 degrees Celsius at point 14, and 88.7 degrees Celsius at point 16. A temperature matrix was generated using a 5x6 grid layout. When calculating the temperature difference between point 15 and its adjacent points, it was found that the temperature difference with point 10 above it reached 5.3 degrees Celsius, and the temperature difference with point 16 to the right was 0.8 degrees Celsius. The maximum value of 5.3 was taken as the temperature change intensity value for this point. In the next two acquisition cycles, the temperature change intensity values ​​at this point were 6.1 degrees Celsius and 7.4 degrees Celsius, respectively, both exceeding the preset safety threshold of 5 degrees Celsius. Based on the judgment rule of exceeding the limit for three consecutive cycles, the system marked the area where point 15 was located as an abnormal power consumption area and recorded its coordinate position in the 3rd row and 4th column of the matrix. Simultaneously checking the temperature matrices of the pick-and-place machine and optical inspection equipment, no similar anomalies were found.

[0142] In this embodiment of the application, the method achieves accurate location of abnormally heated parts of the equipment through systematic temperature field analysis and change intensity detection, providing reliable problem area identification for subsequent energy efficiency optimization and improving the accuracy and timeliness of fault warning.

[0143] Figure 4 This is a schematic diagram of a specific implementation of an intelligent production control system for electronic components provided in this application, with reference to... Figure 4 The system may include: The acquisition module 41 is used to acquire second-level power consumption data of multiple devices, temperature field distribution information of multiple devices, and ambient temperature and humidity data of the production environment during the production process of electronic components. The multiple devices include a chip mounter, a reflow soldering machine, and an optical inspection machine.

[0144] The fusion module 42 is used to fuse the second-level power consumption data with the environmental temperature and humidity data to generate multi-source time-series stream data.

[0145] The generation module 43 is used to generate thermal imaging data based on the temperature field distribution information of the multiple devices, and to determine the abnormal power consumption area based on the thermal imaging data.

[0146] The prediction module 44 is used to generate periodic correlation features through dilated convolution operation based on the multi-source time-series data and the abnormal power consumption region, and to predict the energy efficiency state change points of future time windows based on the periodic correlation features.

[0147] The decomposition module 45 is used to decompose the global energy efficiency target to the process unit level of the chip mounter, the reflow soldering equipment and the optical inspection equipment respectively based on the energy efficiency state change point and using the time-series target cascading algorithm, thereby generating unit-level energy efficiency constraints.

[0148] The adjustment module 46 is used to coordinately adjust the start-stop sequence of the chip mounter, the power parameters of the reflow soldering equipment, and the sampling frequency of the optical inspection equipment according to the unit-level energy efficiency constraints, so as to achieve the optimal balance between power consumption and yield in the electronic component manufacturing process.

[0149] The intelligent production control system for electronic components in this application is used to implement the aforementioned intelligent production control method for electronic components. Therefore, the specific implementation of the intelligent production control system for electronic components can be found in the embodiment section of the intelligent production control method for electronic components above. The specific implementation can be referred to the description of the corresponding embodiments, which will not be repeated here.

[0150] This application also provides an electronic device, comprising: a memory for storing a computer program; and a processor for executing the computer program to implement the steps of any of the above-described intelligent production control methods for electronic components.

[0151] This application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of any of the above-described intelligent production control methods for electronic components.

[0152] In one exemplary embodiment, the aforementioned computer-readable storage medium may include, but is not limited to, various media capable of storing computer programs, such as USB flash drives, read-only memory, random access memory, portable hard drives, magnetic disks, or optical disks.

[0153] Embodiments of the present invention also provide a computer program product, which includes a computer program that, when executed by a processor, implements the steps in any of the embodiments of the intelligent production control method for electronic components described above.

[0154] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0155] The above provides a detailed description of an intelligent production control method, system, electronic device, and storage medium for electronic components provided in this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the embodiments above are merely for the purpose of helping to understand the method and its core ideas. It should be noted that those skilled in the art can make various improvements and modifications to this application without departing from its principles, and these improvements and modifications also fall within the protection scope of this application.

Claims

1. A method for intelligent production control of electronic components, characterized in that, include: The system collects second-level power consumption data, temperature field distribution information of multiple devices, and ambient temperature and humidity data of the production environment during the electronic component manufacturing process. The multiple devices include a chip mounter, a reflow soldering machine, and an optical inspection machine. By fusing the second-level power consumption data with the ambient temperature and humidity data, multi-source time-series stream data is generated; Based on the temperature field distribution information of the multiple devices, thermal imaging data is generated, and abnormal power consumption areas are determined based on the thermal imaging data. Based on the multi-source time-series data and the abnormal power consumption region, periodic correlation features are generated through dilated convolution operations. Based on the periodic correlation features, the energy efficiency state change points of future time windows are predicted. Based on the energy efficiency state change points, the global energy efficiency target is decomposed into the process unit level of the chip mounter, the reflow soldering equipment, and the optical inspection equipment using a time-series target cascade algorithm, thereby generating unit-level energy efficiency constraints. Based on the unit-level energy efficiency constraints, the start-stop sequence of the chip mounter, the power parameters of the reflow soldering equipment, and the sampling frequency of the optical inspection equipment are adjusted in a coordinated manner to achieve optimal power consumption and yield in the electronic component manufacturing process. The step of generating periodic correlation features through dilated convolution operations based on the multi-source time-series data and the abnormal power consumption region, and predicting energy efficiency state change points in future time windows based on the periodic correlation features, includes: Extract the power consumption sequences of the reflow soldering equipment, the pick-and-place machine, and the optical inspection equipment corresponding to the abnormal power consumption regions from the multi-source time-series data. The power consumption sequences of the reflow soldering equipment, the pick-and-place machine, and the optical inspection equipment are horizontally spliced ​​together. Perform dilated convolution operation on the spliced ​​multi-device power consumption sequence to capture the periodic correlation features of cross-device power consumption fluctuations; Based on the aforementioned periodic correlation characteristics, energy efficiency status change points for the next N production batch time windows are generated, where N is greater than or equal to 3. The process of performing dilated convolution on the concatenated multi-device power consumption sequence to capture the periodic correlation features of cross-device power consumption fluctuations includes: Define a three-level dilated convolution kernel structure; By scanning the spliced ​​multi-device power consumption sequence at second-level intervals using the first-level convolutional kernel, the first-cycle fluctuation characteristics within adjacent production batches are captured. The second-level convolutional kernel scans the spliced ​​multi-device power consumption sequence at minute-level intervals to capture the second-cycle variation features across batches. The third-level convolutional kernel scans the spliced ​​multi-device power consumption sequence at hourly intervals to capture the third-cycle trend characteristics of the daily production plan. The first periodic fluctuation feature, the second periodic change feature, and the third periodic trend feature are vertically superimposed to generate a multi-scale feature combination of cross-device power consumption fluctuation. Based on the weight allocation results of features at each level in the multi-scale feature combination, periodic correlation features are generated; Based on the energy efficiency state change points, the global energy efficiency target is decomposed into process unit levels for the chip mounter, reflow soldering equipment, and optical inspection equipment using a time-series target cascade algorithm, generating unit-level energy efficiency constraints, including: A global energy efficiency target is generated based on the deviation between the preset unit output power consumption benchmark and the current actual power consumption. The time interval for target decomposition is determined based on the time coordinates of the energy efficiency state change points. Within the time interval, the global energy efficiency target is broken down into the maximum allowable start-stop number constraints of the pick-and-place machine mounting unit, the upper limit constraint of the power fluctuation of the reflow soldering equipment temperature control unit, and the minimum sampling interval constraint of the optical inspection equipment scanning unit according to the preset equipment function weights. The maximum allowable number of start-stop cycles, the upper limit of power fluctuation, and the minimum sampling interval constraint together constitute the unit-level energy efficiency constraint. The step of generating thermal imaging data based on the temperature field distribution information of the multiple devices, and determining abnormal power consumption areas based on the thermal imaging data, includes: The surface temperature distribution data of the chip mounter, the surface temperature distribution data of the reflow soldering equipment, and the surface temperature distribution data of the optical inspection equipment are separated from the temperature field distribution information of the multiple devices. The surface temperature distribution data of each device are divided into a two-dimensional matrix containing the temperature values ​​of all monitoring points according to the preset monitoring grid division rules, and the two-dimensional matrix is ​​used as thermal imaging data. The thermal imaging data is used to calculate the temperature difference between adjacent monitoring grids to obtain the temperature change intensity value of each monitoring grid. When the temperature change intensity value of the same monitoring grid exceeds the preset fluctuation threshold in N consecutive collection cycles, the monitoring grid is marked as an abnormal power consumption area.

2. The method according to claim 1, characterized in that, The generation of periodic correlation features based on the weight allocation results of features at each level in the multi-scale feature combination includes: Based on the multi-scale feature combination, the variance contribution of each feature is calculated respectively; Based on the proportion of the variance contribution of each feature, the first weight value of the first periodic fluctuation feature, the second weight value of the second periodic change feature, and the third weight value of the third periodic trend feature are dynamically allocated. The first weighted feature is obtained by multiplying the first periodic fluctuation feature by the first weight value; the second weighted feature is obtained by multiplying the second periodic change feature by the second weight value; and the third weighted feature is obtained by multiplying the third periodic trend feature by the third weight value. The first weighted feature, the second weighted feature, and the third weighted feature are superimposed. The time dimension of the superimposed results is smoothed to generate periodic correlation features.

3. The method according to claim 1, characterized in that, The process of fusing the second-level power consumption data with the environmental temperature and humidity data to generate multi-source time-series stream data includes: The second-level power consumption data of the pick-and-place machine is marked as the first timing sequence, the second-level power consumption data of the reflow soldering equipment is marked as the second timing sequence, and the second-level power consumption data of the optical inspection equipment is marked as the third timing sequence. The environmental temperature and humidity data are split into a temperature time series and a humidity time series according to the collection timestamp; Perform timestamp alignment on the first time series, the second time series, the third time series, the temperature time series, and the humidity time series; The aligned five types of time series are concatenated into multi-source time series stream data according to the device dimension.

4. An intelligent production control system for electronic components, characterized in that, include: The data acquisition module is used to collect second-level power consumption data of multiple devices, temperature field distribution information of multiple devices, and ambient temperature and humidity data of the production environment during the production process of electronic components. The multiple devices include a chip mounter, a reflow soldering machine, and an optical inspection machine. The fusion module is used to fuse the second-level power consumption data with the ambient temperature and humidity data to generate multi-source time-series stream data; The generation module is used to generate thermal imaging data based on the temperature field distribution information of the multiple devices, and to determine abnormal power consumption areas based on the thermal imaging data. The prediction module is used to generate periodic correlation features based on the multi-source time-series data and the abnormal power consumption region through dilated convolution operation, and predict the energy efficiency state change points in future time windows based on the periodic correlation features. The decomposition module is used to decompose the global energy efficiency target to the process unit level of the chip mounter, the reflow soldering equipment, and the optical inspection equipment based on the energy efficiency state change points using a time-series target cascading algorithm, thereby generating unit-level energy efficiency constraints. The adjustment module is used to coordinately adjust the start-stop timing of the chip mounter, the power parameters of the reflow soldering equipment, and the sampling frequency of the optical inspection equipment according to the unit-level energy efficiency constraints, so as to achieve the optimal balance between power consumption and yield in the production process of electronic components. The step of generating periodic correlation features through dilated convolution operations based on the multi-source time-series data and the abnormal power consumption region, and predicting energy efficiency state change points in future time windows based on the periodic correlation features, includes: Extract the power consumption sequences of the reflow soldering equipment, the pick-and-place machine, and the optical inspection equipment corresponding to the abnormal power consumption regions from the multi-source time-series data. The power consumption sequences of the reflow soldering equipment, the pick-and-place machine, and the optical inspection equipment are horizontally spliced ​​together. Perform dilated convolution operation on the spliced ​​multi-device power consumption sequence to capture the periodic correlation features of cross-device power consumption fluctuations; Based on the aforementioned periodic correlation characteristics, energy efficiency status change points for the next N production batch time windows are generated, where N is greater than or equal to 3. The process of performing dilated convolution on the concatenated multi-device power consumption sequence to capture the periodic correlation features of cross-device power consumption fluctuations includes: Define a three-level dilated convolution kernel structure; By scanning the spliced ​​multi-device power consumption sequence at second-level intervals using the first-level convolutional kernel, the first-cycle fluctuation characteristics within adjacent production batches are captured. The second-level convolutional kernel scans the spliced ​​multi-device power consumption sequence at minute-level intervals to capture the second-cycle variation features across batches. The third-level convolutional kernel scans the spliced ​​multi-device power consumption sequence at hourly intervals to capture the third-cycle trend characteristics of the daily production plan. The first periodic fluctuation feature, the second periodic change feature, and the third periodic trend feature are vertically superimposed to generate a multi-scale feature combination of cross-device power consumption fluctuation. Based on the weight allocation results of features at each level in the multi-scale feature combination, periodic correlation features are generated; Based on the energy efficiency state change points, the global energy efficiency target is decomposed into process unit levels for the chip mounter, reflow soldering equipment, and optical inspection equipment using a time-series target cascade algorithm, generating unit-level energy efficiency constraints, including: A global energy efficiency target is generated based on the deviation between the preset unit output power consumption benchmark and the current actual power consumption. The time interval for target decomposition is determined based on the time coordinates of the energy efficiency state change points. Within the time interval, the global energy efficiency target is broken down into the maximum allowable start-stop number constraints of the pick-and-place machine mounting unit, the upper limit constraint of the power fluctuation of the reflow soldering equipment temperature control unit, and the minimum sampling interval constraint of the optical inspection equipment scanning unit according to the preset equipment function weights. The maximum allowable number of start-stop cycles, the upper limit of power fluctuation, and the minimum sampling interval constraint together constitute the unit-level energy efficiency constraint. The step of generating thermal imaging data based on the temperature field distribution information of the multiple devices, and determining abnormal power consumption areas based on the thermal imaging data, includes: The surface temperature distribution data of the chip mounter, the surface temperature distribution data of the reflow soldering equipment, and the surface temperature distribution data of the optical inspection equipment are separated from the temperature field distribution information of the multiple devices. The surface temperature distribution data of each device are divided into a two-dimensional matrix containing the temperature values ​​of all monitoring points according to the preset monitoring grid division rules, and the two-dimensional matrix is ​​used as thermal imaging data. The thermal imaging data is used to calculate the temperature difference between adjacent monitoring grids to obtain the temperature change intensity value of each monitoring grid. When the temperature change intensity value of the same monitoring grid exceeds the preset fluctuation threshold in N consecutive collection cycles, the monitoring grid is marked as an abnormal power consumption area.

5. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor is configured to execute the computer program to implement the steps of the intelligent production control method for electronic components as described in any one of claims 1 to 3.

6. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, enables the implementation of the intelligent production control method for electronic components as described in any one of claims 1 to 3.

Images