Voltage monitoring system of megasonic generator

Abstract

The invention relates to the field of intelligent monitoring, and particularly discloses a voltage monitoring system of a megasonic generator, which comprises the following steps: firstly, carrying out conditioning and analog-to-digital conversion on a voltage analog signal, and converting the voltage analog signal into a digital signal; thirdly, establishing a dynamic mapping relation between the temperature drift effect and a voltage threshold value by analyzing historical voltage data and a time sequence fluctuation rule and synchronously analyzing an environment temperature change mode; the voltage tolerance requirements of different process stages are coded and quantified in combination with load characteristics, multivariable collaborative modeling is carried out by utilizing a response aggregation coding mechanism, and multi-dimensional interaction characteristics reflecting a real-time operation state are generated, so that a voltage preset range is dynamically adjusted, and a control instruction is generated. According to the method, the voltage preset range is dynamically optimized, different working conditions and dynamic factors are accurately adapted, the problem caused by a wide or too narrow threshold value is avoided, and therefore the high-precision manufacturing requirement is met.

Description

Technical Field

[0001] This application relates to the field of intelligent monitoring, and more specifically, to a voltage monitoring system for a megasonic wave generator. Background Art

[0002] In the fields of modern industrial production and scientific research, megasonic wave generators play an indispensable role in many key processes such as precision cleaning, material processing, and semiconductor manufacturing due to their unique high-frequency vibration characteristics. Their normal operation depends on a stable and appropriate supply voltage. Once the voltage is abnormal, it will not only reduce the working efficiency of the equipment, affect the processing accuracy, but may even cause equipment failures, shorten the service life, and result in serious economic losses. Therefore, it is crucial to accurately monitor the voltage of the megasonic wave generator and respond to abnormalities in a timely manner.

[0003] Traditional voltage monitoring systems rely on a fixed threshold determination mechanism. Although they can provide basic protection under extreme working conditions, it is difficult to balance the contradiction between equipment safety and operation efficiency: when the threshold is set too wide, the system can avoid false triggering caused by short-term power grid fluctuations, but it cannot capture the continuous slow voltage deviation during the light load operation stage, resulting in the transducer being in a non-optimal polarization state for a long time, accelerating the degradation of the piezoelectric material performance; while overly tightening the threshold improves the ability to detect subtle abnormalities, but is easily affected by conventional working conditions such as instantaneous surges during equipment startup and shutdown, and cross-interference during multi-machine collaborative operation, frequently triggering the protection mechanism and causing unplanned shutdowns. A more fundamental defect is that the existing system cannot dynamically analyze the parameter drift of the power supply module caused by equipment aging, the reference voltage deviation caused by environmental temperature and humidity changes, and the impact of load characteristics differences in different production stages on the voltage tolerance threshold, resulting in the disconnection between the equipment maintenance strategy and the real-time operating state, and it is difficult to meet the dual requirements of system robustness and process stability in high-precision manufacturing scenarios.

[0004] Therefore, an optimized voltage monitoring solution for a megasonic wave generator is desired. Summary of the Invention

[0005] To solve the above technical problems, this application is proposed.

[0006] According to one aspect of this application, a voltage monitoring system for a megasonic wave generator is provided, which includes: A signal receiving module for receiving voltage analog signals; A signal conditioning module for conditioning the voltage analog signal to obtain a conditioned voltage analog signal; An analog-to-digital conversion module for performing analog-to-digital conversion on the conditioned voltage analog signal to obtain a voltage digital signal; A voltage setting module for setting a voltage preset range, wherein the voltage setting module includes: a range extraction unit for extracting an initial upper limit value and an initial lower limit value of the voltage; a parameter acquisition unit for acquiring historical voltage data, the current device load type, and an environmental temperature time series; a variable interaction unit for performing response aggregation between device load - external variable features on the historical voltage data, the current device load type, and the environmental temperature time series to obtain a multi - variable internal collaborative interaction coding feature vector; and a range setting unit for optimizing the initial upper limit value and the initial lower limit value of the voltage based on the multi - variable internal collaborative interaction coding feature vector to obtain the voltage preset range. A voltage judgment module for judging whether the voltage digital signal is within the voltage preset range to obtain a judgment result. An instruction generation module for generating a control instruction for the megasonic wave generator based on the judgment result.

[0007] Compared with the prior art, a voltage monitoring system for a megasonic wave generator provided by the present application first conditions and performs analog - to - digital conversion on the voltage analog signal to convert it into a digital signal. Then, by analyzing the historical voltage data and the time - series fluctuation law, and synchronously analyzing the environmental temperature change pattern, a dynamic mapping relationship between the temperature drift effect and the voltage threshold is established. Combining the load characteristic coding, quantifying the voltage tolerance requirements in different process stages, and using the response aggregation coding mechanism for multi - variable collaborative modeling to generate multi - dimensional interaction features reflecting the real - time operating state, thereby dynamically adjusting the voltage preset range and generating a control instruction. This method dynamically optimizes the voltage preset range, precisely adapts to different working conditions and dynamic factors, avoids problems caused by broad or narrow thresholds, ensures that the device maintenance strategy matches the operating state, improves the system robustness and process stability, and meets the requirements of high - precision manufacturing. Description of the Drawings

[0008] By describing the embodiments of the present application in more detail in conjunction with the drawings, the above - mentioned and other objects, features, and advantages of the present application will become more obvious. The drawings are used to provide a further understanding of the embodiments of the present application, and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application, and do not constitute a limitation to the present application. In the drawings, the same reference numerals generally represent the same components or steps.

[0009] Figure 1 It is a block diagram of a voltage monitoring system for a megasonic wave generator according to an embodiment of the present application.

[0010] Figure 2 It is a block diagram of a voltage setting module in a voltage monitoring system for a megasonic wave generator according to an embodiment of the present application.

[0011] Figure 3Block diagram of the variable interaction unit in the voltage monitoring system of the megasonic wave generator according to an embodiment of the present application.

[0012] Figure 4 Block diagram of the multi-variable internal cooperation subunit in the voltage monitoring system of the megasonic wave generator according to an embodiment of the present application. Detailed implementation manners

[0013] Embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although some embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. On the contrary, these embodiments are provided to more thoroughly and completely understand the present disclosure. It should be understood that the drawings and embodiments of the present disclosure are only for exemplary purposes and are not used to limit the protection scope of the present disclosure.

[0014] It should be understood that the steps recited in the method embodiments of the present disclosure can be executed in different orders and / or in parallel. In addition, the method embodiments may include additional steps and / or omit the steps shown. The scope of the present disclosure is not limited in this regard.

[0015] The present application proposes a voltage monitoring system for a megasonic wave generator, which realizes intelligent adaptation of the voltage monitoring boundary of the megasonic wave generator by constructing a dynamic threshold generation engine. Specifically, first, the voltage analog signal is subjected to signal conditioning and analog-to-digital conversion to convert it into a voltage digital signal. Then, by extracting the temporal fluctuation law of the historical voltage data, the voltage baseline characteristics under the inherent working conditions of the device are captured; the time-varying mode of the ambient temperature sequence is synchronously analyzed, and a dynamic mapping relationship between the temperature drift effect and the voltage threshold is established; at the same time, combined with the characteristic coding of the current load type, the different requirements of different process stages for the voltage tolerance range are quantified. Then, with the help of the response aggregation coding mechanism, a multi-variable collaborative modeling of the implicit coupling relationship between the device load, environmental variables and voltage fluctuations is carried out to generate multi-dimensional interaction features reflecting the real-time operating state, so as to dynamically adjust the voltage preset range and generate corresponding control instructions. The present application dynamically optimizes the voltage preset range, accurately adapts to different working conditions and dynamic factors, avoids the problem of transducer performance degradation caused by broad thresholds, and prevents unplanned shutdowns caused by narrow thresholds. Moreover, it can dynamically analyze the effects of device aging, environmental changes and load characteristic differences, make the device maintenance strategy match the real-time operating state, and meet the requirements of high-precision manufacturing for system robustness and process stability.

[0016] Figure 1 Block diagram of the voltage monitoring system of the megasonic wave generator according to an embodiment of the present application. Specifically, as Figure 1As shown, the voltage monitoring system 100 of the megasonic wave generator according to an embodiment of the present application includes: a signal receiving module 110 for receiving a voltage analog signal; a signal conditioning module 120 for conditioning the voltage analog signal to obtain a conditioned voltage analog signal; an analog-to-digital conversion module 130 for performing analog-to-digital conversion on the conditioned voltage analog signal to obtain a voltage digital signal; a voltage setting module 140 for setting a voltage preset range; a voltage judgment module 150 for judging whether the voltage digital signal is within the voltage preset range to obtain a judgment result; and a command generation module 160 for generating a control command for the megasonic wave generator based on the judgment result.

[0017] Specifically, the signal receiving module 110 is configured to receive a voltage analog signal. It should be understood that during the operation of the megasonic wave generator, its supply voltage is a continuously changing physical quantity, and the voltage analog signal can reflect the actual voltage magnitude and change situation of the generator in real time and accurately. By receiving this signal, the monitoring system can obtain first-hand information about the supply voltage of the generator, providing raw data for subsequent analysis and processing.

[0018] Specifically, a high-precision voltage sensor is installed at the power input end of the megasonic wave generator. This sensor is the core component for receiving the voltage analog signal and has the ability to detect the voltage situation supplied to the megasonic wave generator in real time. During installation, it is necessary to ensure that the sensor is firmly connected to the power input end to ensure the stability and reliability of signal transmission and avoid signal loss or interference caused by loose connection.

[0019] After the sensor is installed, during its operation, it continuously monitors the voltage change at the power input end. When the megasonic wave generator is running, the magnitude and change of its supply voltage act on the voltage sensor in the form of an analog signal in real time. The sensing element inside the sensor responds to these voltage changes and converts them into corresponding electrical signals for output. Since there may be various interference factors in the actual environment, to ensure the quality of the obtained voltage analog signal, corresponding anti-interference measures are also equipped when connecting the output end of the sensor to the subsequent circuit, such as adding a shielding layer to reduce electromagnetic interference and setting up a filtering circuit to remove clutter, etc., to ensure that the voltage analog signal can be accurately and stably transmitted to the signal conditioning module, providing a reliable raw data basis for the subsequent conditioning and analog-to-digital conversion of the voltage analog signal and the accurate operation of the entire voltage monitoring system.

[0020] Specifically, the signal conditioning module 120 is configured to condition the voltage analog signal to obtain a conditioned voltage analog signal. In particular, the signal conditioning here includes signal amplification and filtering. Correspondingly, in actual voltage monitoring, the voltage analog signal obtained from the megasonic wave generator may have a small amplitude due to reasons such as transmission distance and sensor characteristics. For example, the signal output by the sensor may be only a few millivolts or even smaller. Such a weak signal is prone to being affected by noise and interference during subsequent processes such as analog-to-digital conversion, resulting in a decrease in measurement accuracy. Through amplification processing, the amplitude of the signal can be increased to an appropriate range, so that the subsequent circuit can better process and analyze it. Moreover, there are various interference factors in the industrial production environment, such as electromagnetic interference and power supply noise. These interferences will be superimposed on the voltage analog signal, making the signal unclear and affecting the accurate measurement of the true voltage value. For example, the operation of nearby motors and the working of high-frequency devices may generate electromagnetic interference, resulting in signal clutter. Filtering can filter out noise and interference signals according to the frequency characteristics of the signal, retain the useful voltage signal, and improve the quality and purity of the signal. Therefore, after amplification and filtering, the quality of the voltage analog signal is significantly improved. Amplification processing makes the signal amplitude within an appropriate range, reducing the influence of quantization errors, etc.; filtering removes noise and interference, making the signal more stable and accurate, so as to more accurately reflect the actual voltage situation of the megasonic wave generator, provide a high-quality signal for subsequent analog-to-digital conversion and voltage judgment, etc., and ultimately improve the measurement accuracy of the entire voltage monitoring system.

[0021] Specifically, the analog-to-digital conversion module 130 is configured to perform analog-to-digital conversion on the conditioned voltage analog signal to obtain a voltage digital signal. It should be understood that analog signals cannot be directly understood and processed by these digital systems, so it is necessary to convert the conditioned voltage analog signal into a digital signal. Analog-to-digital conversion can convert a continuously changing analog voltage signal into discrete digital quantities, and each digital quantity corresponds to a specific voltage value. In this way, the magnitude of the voltage can be accurately measured and quantified, facilitating precise analysis and calculation of the voltage. For example, parameters such as the average value, effective value, and peak value of the voltage can be calculated through digital signals, so as to more comprehensively understand the voltage characteristics of the megasonic wave generator.

[0022] The specific implementation process is as follows: After the signal conditioning module completes the amplification and filtering of the voltage analog signal, the conditioned voltage analog signal obtained will be transmitted to the analog-to-digital conversion module. The analog-to-digital conversion module usually consists of a sample-and-hold circuit, a quantizer, and an encoder. First, the sample-and-hold circuit samples the conditioned analog signal, obtains the instantaneous value of the analog signal within a specific time interval, and holds it for a period of time for subsequent processing. The sampling frequency needs to be reasonably set according to the characteristics of the signal and the accuracy requirements of the monitoring system. A higher sampling frequency can capture signal changes more accurately, but it will also increase the amount of data processing.

[0023] The sampled signal enters the quantizer, which divides the continuous analog voltage values into multiple discrete quantization levels. The voltage value of the analog signal is compared with these quantization levels to determine the quantization interval to which it belongs, achieving the preliminary conversion of the analog signal to a discrete digital quantity. However, the quantization process inevitably introduces quantization errors. To reduce the impact of such errors on the measurement accuracy, a more refined quantization method can be adopted, such as increasing the number of quantization levels.

[0024] The quantized signal is then converted by the encoder into a binary digital code, and finally a voltage digital signal is output. These digital signals can be easily recognized and processed by digital systems. For example, they can be used to calculate parameters such as the average value, effective value, and peak value of the voltage, facilitating the accurate analysis of the voltage characteristics of the megasonic generator and providing data support for subsequent judgment of whether the voltage is within the preset range and for generating control instructions.

[0025] Specifically, the voltage setting module 140 is used to set the voltage preset range. It can be understood that the megasonic wave generator requires an appropriate voltage range to ensure its normal operation under different working conditions and load conditions. If the voltage is too high, it may cause the electronic components inside the device to be overloaded, shortening the device life or even damaging the device; if the voltage is too low, the device may not achieve the expected working efficiency, affecting the production quality. Therefore, it is necessary to set a reasonable voltage preset range according to the characteristics and working requirements of the device to ensure that the device operates in a safe and efficient state. However, due to its static characteristics, the traditional fixed threshold mechanism cannot perceive the complex effects of device aging, environmental temperature and humidity drift, and load dynamic switching on the voltage tolerance boundary. For example, when the parameters of the power supply module decay with the use time, the original threshold may not accurately reflect the true tolerance ability of the current components; the reference voltage offset caused by environmental temperature fluctuations will change the effective working range, and the sensitivity differences of different load types (such as full load and no load in the cleaning tank) to voltage stability are significant. If the fixed preset range is continued to be used, it is difficult for the system to identify the sub-healthy state that has been on the threshold edge for a long time under light load, and it is also easy to cause misoperation due to threshold rigidity during load mutation or grid instantaneous disturbance. Therefore, in view of the problems in the above background technology, the following solutions are proposed in the voltage setting module.

[0026] Figure 2 It is a block diagram of the voltage setting module in the voltage monitoring system of the megasonic wave generator according to an embodiment of the present application. Specifically, as Figure 2 shown, the voltage setting module 140 includes: a range extraction unit 141 for extracting the initial upper limit value and the initial lower limit value of the voltage; a parameter acquisition unit 142 for acquiring historical voltage data, the current device load type, and the environmental temperature time series; a variable interaction unit 143 for performing response aggregation between device load-external variable features on the historical voltage data, the current device load type, and the environmental temperature time series to obtain a multi-variable internal collaborative interaction coding feature vector; a range setting unit 144 for optimizing the initial upper limit value and the initial lower limit value of the voltage based on the multi-variable internal collaborative interaction coding feature vector to obtain the voltage preset range.

[0027] Specifically, in the embodiments of the present application, the range extraction unit 141 is configured to extract the initial upper limit value and the initial lower limit value of the voltage. It should be understood that the initial upper limit value and the initial lower limit value of the voltage refer to the boundary values of the voltage range preset before the dynamic optimization of the voltage of the megasonic wave generator. These two values provide a starting reference range for subsequent dynamic optimization. The device has a roughly appropriate voltage operating range in design and theory. The initial upper limit value and the lower limit value define this basic range, enabling the system to have a clear voltage control framework at the initial stage of operation, ensuring that the voltage fluctuates within a relatively reasonable range and avoiding unlimited abnormal changes in the voltage. Based on the initial values and by analyzing multi-variable information such as the device load type and the fine-grained full-time domain correlation characteristics of voltage-environment temperature, the upper limit dynamic adjustment coefficient and the lower limit dynamic adjustment coefficient are obtained through corresponding algorithms and models, and then the initial upper limit value and the initial lower limit value are dynamically optimized to obtain a voltage preset range that better conforms to the actual operating conditions.

[0028] The implementation process is as follows: In the system initialization stage, technicians need to obtain the theoretical voltage range during normal operation of the device according to the product manual, design specifications, and relevant industry standards of the megasonic wave generator. This range is an important basis for determining the initial upper limit value and the initial lower limit value of the voltage. For example, if the generator is designed to require the supply voltage to operate stably within a certain fluctuation range near a certain nominal value, the upper and lower limits of this fluctuation range can be used as a preliminary reference.

[0029] Next, the initial values are further determined in combination with the actual application scenario of the generator. If the generator is used in semiconductor manufacturing processes with extremely high requirements for voltage stability, based on the theoretical range, the influence of load characteristics, environmental factors, etc. of the device under this process on the voltage needs to be comprehensively considered, and the voltage range is appropriately narrowed. For possible voltage fluctuation situations, the analysis also needs to be combined with the operation experience data of previous similar devices. If voltage fluctuations close to the theoretical limit value have occurred under certain specific working conditions and have not damaged the device or affected its normal operation, the initial values can be fine-tuned accordingly.

[0030] After completing the above comprehensive considerations, the determined voltage upper limit value is set as the initial upper limit value of the voltage, and the lower limit value is set as the initial lower limit value of the voltage. These two values will be stored in the memory of the system so that the subsequent range extraction unit can conveniently obtain them and use them for the optimization calculation of the voltage preset range.

[0031] Specifically, in the embodiments of the present application, the parameter acquisition unit 142 is configured to acquire historical voltage data, the current device load type, and the environmental temperature time series. Correspondingly, the historical voltage data includes the voltage values of the device at different past times. By analyzing these data, the change trend, fluctuation range, and abnormal situations of the voltage over time can be understood, which helps to discover the rules and potential patterns of voltage changes. Different device load types have different demands and impacts on voltage. Understanding the current device load type can more accurately evaluate its dynamic voltage demand, so as to make targeted adjustments when setting the voltage preset range, and avoid unstable operation or damage of the device caused by the mismatch between the load type and the voltage. The change in environmental temperature will affect the performance and parameters such as resistance of the device, and thus affect the voltage characteristics of the device. Generally speaking, an increase in temperature may lead to an increase in resistance, thereby causing a change in the voltage required by the device. Therefore, by comprehensively considering the information of historical voltage data, the current device load type, and the environmental temperature time series, the voltage demand of the device in various situations can be analyzed more comprehensively and accurately, so as to dynamically optimize the initial upper and lower limit values of the voltage and obtain a more reasonable voltage preset range.

[0032] Specifically, in the embodiments of the present application, the variable interaction unit 143 is configured to perform response aggregation between device load - external variable features on the historical voltage data, the current device load type, and the environmental temperature time series to obtain a multi - variable internal collaborative interaction coding feature vector. Figure 3 It is a block diagram of the variable interaction unit in the voltage monitoring system of the megasonic wave generator according to the embodiments of the present application. Specifically, as Figure 3 shown, more specifically, in the embodiments of the present application, the variable interaction unit 143 includes: a historical voltage time - series coding sub - unit 1431, configured to perform time - series coding on the historical voltage data to obtain a voltage time - series fluctuation pattern feature coding vector; an environmental temperature time - series coding sub - unit 1432, configured to perform time - series coding on the environmental temperature time series to obtain an environmental temperature fluctuation time - series pattern feature coding vector; a device load type coding sub - unit 1433, configured to perform one - hot coding on the current device load type to obtain a device load type one - hot coding vector; a voltage - environmental temperature fine - grained association sub - unit 1434, configured to calculate a voltage - environmental temperature full - time - domain fine - grained association feature matrix between the voltage time - series fluctuation pattern feature coding vector and the environmental temperature fluctuation time - series pattern feature coding vector; a multi - variable internal collaboration sub - unit 1435, configured to perform response aggregation coding between device load - external variables dominated by semantic query on the device load type one - hot coding vector and the voltage - environmental temperature full - time - domain fine - grained association feature matrix to obtain the multi - variable internal collaborative interaction coding feature vector.

[0033] Specifically, the historical voltage time series encoding subunit 1431 is configured to perform time series encoding on the historical voltage data to obtain a voltage time series fluctuation pattern feature encoding vector. More specifically, in the embodiment of the present application, the historical voltage time series encoding subunit is configured to: perform time series encoding based on causal dilated convolution on the historical voltage data to obtain the voltage time series fluctuation pattern feature encoding vector. It should be understood that considering that the historical voltage data implies key information about the inherent operating conditions of the device, the evolution of grid quality, and the cumulative effect of component aging. Traditional fixed-threshold monitoring systems only focus on the instantaneous value of the real-time voltage, ignoring the correlation and trend characteristics of voltage fluctuations in the time dimension. For example, under light load conditions, the transducer may cause a gradual drift of the voltage due to changes in the heat dissipation conditions. Although such a slow drift does not exceed the fixed threshold range within a single sampling period, long-term accumulation will cause the polarization state of the piezoelectric material to deviate from the optimal range, ultimately leading to a misalignment of the resonance frequency or a decrease in the conversion efficiency. If such cross-cycle fluctuation patterns cannot be mined from historical data, the monitoring system will be difficult to distinguish between instantaneous noise interference and the true performance degradation trend, resulting in a lag in the early warning of potential failures. Based on this, in the technical solution of the present application, time series encoding is performed on the historical voltage data to obtain a voltage time series fluctuation pattern feature encoding vector. Specifically, in a specific example of the present application, time series encoding based on causal dilated convolution is performed on the historical voltage data to obtain the voltage time series fluctuation pattern feature encoding vector. That is to say, "time series encoding based on causal dilated convolution" constructs multi-scale time series feature extraction capabilities by introducing a causal convolutional neural network with a dilated structure. Causal convolution ensures strict causality in the time dimension, that is, the features at the current moment only depend on historical and current input data, avoiding interference from future information leakage to real-time monitoring; dilated convolution increases the receptive field layer by layer through an exponentially expanding dilation rate, enabling the network to capture long-cycle fluctuation patterns without increasing the number of parameters. Specifically, the network synchronously analyzes high-frequency transient fluctuations (such as device start-stop surges) and low-frequency slow-changing trends (such as voltage baseline drift caused by environmental temperature rise) in the voltage signal by stacking multiple groups of convolutional layers with increasing dilation rates, and fuses these cross-scale time series features into a high-dimensional encoding vector. This vector not only retains the local detailed features of voltage fluctuations but also establishes a potential correlation mapping between short-term abnormal events and long-term performance degradation through cross-layer information transmission of the dilated structure.

[0034] Specifically, the environmental temperature time series encoding subunit 1432 is configured to perform time series encoding on the environmental temperature time series to obtain an environmental temperature fluctuation time series pattern feature encoding vector. Correspondingly, considering that the gradual increase in environmental temperature may lead to a decrease in the heat dissipation efficiency of power devices, causing a slow drift of the voltage baseline; while short-term severe temperature changes may change the equivalent series resistance (ESR) of the capacitor through thermal stress, inducing voltage transient fluctuations. If only the current temperature value is used for analysis, the non-linear effect of the temperature history fluctuation pattern cannot be accurately quantified, resulting in the deviation of the dynamic threshold adjustment from the true physical law. In addition, the impact of periodic temperature change patterns in different seasons or regions (such as diurnal temperature difference cycles, seasonal temperature rise trends) on the long-term operation stability of the device also varies, and it is necessary to capture their potential correlations through time series feature modeling. Therefore, in the technical solution of the present application, time series encoding is performed on the environmental temperature time series to obtain an environmental temperature fluctuation time series pattern feature encoding vector. In particular, the above-mentioned causal dilated convolution can be used to perform time series encoding on the environmental temperature time series to capture and extract the fluctuation patterns of environmental temperature, such as seasonal changes, diurnal temperature difference changes, etc., to obtain an environmental temperature fluctuation time series pattern feature encoding vector.

[0035] Specifically, the device load type encoding subunit 1433 is configured to perform one-hot encoding on the current device load type to obtain a device load type one-hot encoding vector. It should be understood that considering that the device load type belongs to categorical data. It is difficult for a computer to directly process these non-numerical categorical information. Therefore, in order to convert these categorical data into a numerical vector form that can be understood and processed by a computer, facilitating subsequent various mathematical operations and model processing, the present application performs one-hot encoding on the current device load type to obtain a device load type one-hot encoding vector. That is to say, the characteristic of one-hot encoding is that each category has a unique encoding vector, and only the position corresponding to the category in the vector is 1, and the other positions are all 0. In this way, different device load types can be clearly distinguished, avoiding confusion between categories, accurately representing the characteristics of each load type, and providing accurate input for data analysis and model training.

[0036] Specifically, the voltage-environment temperature fine-grained correlation subunit 1434 is configured to calculate a voltage-environment temperature full-time-domain fine-grained correlation feature matrix between the voltage time-sequence fluctuation pattern feature encoding vector and the environment temperature fluctuation time-sequence pattern feature encoding vector. More specifically, in the embodiment of the present application, the voltage-environment temperature fine-grained correlation subunit is configured to: multiply the voltage time-sequence fluctuation pattern feature encoding vector by the transposed vector of the environment temperature fluctuation time-sequence pattern feature encoding vector to obtain the voltage-environment temperature full-time-domain fine-grained correlation feature matrix. Correspondingly, considering that in actual situations, there is often a certain mutual influence between voltage and environment temperature. For example, a continuous high-temperature environment may cause a progressive shift in the voltage baseline by changing the heat dissipation efficiency of power devices, while a sudden temperature drop may lead to a mutation in the capacitor charging and discharging characteristics, generating an instantaneous voltage spike. Therefore, in order to capture this complex relationship hidden behind the data and deeply understand the interaction mechanism between voltage and environment temperature, the present application calculates the voltage-environment temperature full-time-domain fine-grained correlation feature matrix between the voltage time-sequence fluctuation pattern feature encoding vector and the environment temperature fluctuation time-sequence pattern feature encoding vector. In particular, in a specific example of the present application, the voltage time-sequence fluctuation pattern feature encoding vector is multiplied by the transposed vector of the environment temperature fluctuation time-sequence pattern feature encoding vector to obtain the voltage-environment temperature full-time-domain fine-grained correlation feature matrix. In detail, each dimension of the voltage time-sequence fluctuation pattern feature encoding vector represents the fluctuation pattern of the device at a specific time scale (such as second-level transient fluctuations, minute-level trend drifts), while the transpose of the environment temperature fluctuation time-sequence pattern feature encoding vector corresponds to the influence weight of temperature at different time granularities. The matrix multiplication operation quantifies the coupling strength between the voltage fluctuation pattern and the temperature fluctuation pattern at each time node through the alignment and matching of vector spaces. For example, a high-weight element in the matrix may reveal the potential causal relationship between "the current minute-level voltage drift and the temperature sudden rise two hours ago" or the synchronous correlation between "second-level voltage spikes and real-time temperature fluctuations". This fine-grained correlation analysis can decouple the immediate effect of temperature on voltage and the historical cumulative effect, providing a cross-modal time-sequence correlation basis for subsequent multi-variable collaborative modeling.

[0037] Specifically, the multi-variable intra-coordination subunit 1435 is configured to perform a device load-external variable response aggregation encoding dominated by semantic query on the one-hot encoded vector of the device load type and the voltage-environment temperature full-time-domain fine-grained correlation feature matrix to obtain the multi-variable intra-coordination interaction encoding feature vector. Further, considering the intertwined influence of multiple dynamic factors such as environmental temperature fluctuations and historical voltage patterns, the interaction between these variables is not a simple linear superposition, but there are deep semantic associations and non-linear couplings. Traditional monitoring systems using rule engines or shallow feature fusion strategies can only capture the surface statistical associations between device loads and voltage / temperature variables, and cannot analyze the modulation effect of load types on the voltage-temperature dynamic relationship. For example, the impedance characteristics of the liquid medium in the cleaning load mode will amplify the impact of temperature fluctuations on voltage stability, while the constant current demand of the etching load may weaken the interference of temperature on the voltage baseline. If a semantic-level interaction model between the load type and the voltage-temperature correlation matrix is not established, it will be difficult for the system to quantify the differential action intensity of environmental factors on the voltage threshold in different process stages, resulting in the deviation of the dynamic adjustment strategy from the true physical constraints. Therefore, in this application, a device load-external variable response aggregation encoding dominated by semantic query is performed on the one-hot encoded vector of the device load type and the voltage-environment temperature full-time-domain fine-grained correlation feature matrix to obtain the multi-variable intra-coordination interaction encoding feature vector.

[0038] Figure 4 It is a block diagram of the multi-variable intra-coordination subunit in the voltage monitoring system of the megasonic wave generator according to the embodiment of the present application. Specifically, as Figure 4 shown, the multi-variable intra-coordination subunit 1435 includes: an associated feature matrix local time-series splitting secondary subunit 1435-1, configured to split the voltage-environment temperature full-time-domain fine-grained correlation feature matrix based on row vectors to obtain a set of voltage-environment temperature local time-domain fine-grained correlation feature vectors; a device load type special enhancement secondary subunit 1435-2, configured to perform deconvolution feature encoding enhancement on the one-hot encoded vector of the device load type to obtain a one-hot encoded enhanced vector of the device load type; a device load-external variable response secondary subunit 1435-3, configured to perform single-body semantic query responses on each voltage-environment temperature local time-domain fine-grained correlation feature vector in the set of the one-hot encoded enhanced vector of the device load type and the set of voltage-environment temperature local time-domain fine-grained correlation feature vectors to obtain a set of device load-external variable single-body semantic query response score encoding vectors; a device load-external variable aggregation secondary subunit 1435-4, configured to perform dynamic aggregation based on self-attention weights on the set of device load-external variable single-body semantic query response score encoding vectors to obtain the multi-variable intra-coordination interaction encoding feature vector.

[0039] Specifically, the local time-series splitting secondary subunit 1435-1 of the correlation feature matrix is used to split the voltage-environment temperature full-time-domain fine-grained correlation feature matrix based on row vectors to obtain a set of voltage-environment temperature local time-domain fine-grained correlation feature vectors. It can be expressed by the formula: where, is the voltage-environment temperature full-time-domain fine-grained correlation feature matrix, is the splitting operation, is the set of voltage-environment temperature local time-domain fine-grained correlation feature vectors, , , and are respectively the 1st, 2nd, th, and th voltage-environment temperature local time-domain fine-grained correlation feature vectors in the set of voltage-environment temperature local time-domain fine-grained correlation feature vectors.

[0040] It should be understood that the full-time-domain correlation feature matrix of voltage and environment temperature contains the complex interaction patterns between the two at different time granularities. However, the mechanism by which the stability of voltage is affected by temperature is not evenly distributed across the entire time axis, but exhibits heterogeneous characteristics of strong correlation within local time periods and weak correlation across time periods. For example, the instantaneous temperature rise during the device startup phase may only affect the voltage transient response in the subsequent few seconds, while the gradual change in ambient temperature over several hours may lead to a cumulative shift in the voltage baseline. If the full-time-domain correlation matrix is processed as a whole, it is difficult for the model to distinguish the differences in local correlation strengths at different time nodes, resulting in the dynamic threshold adjustment strategy being unable to accurately match the actual physical laws, and may ignore abnormal signals during critical periods or over-respond to non-significant fluctuations. Based on this, the present application focuses on the voltage-temperature interaction characteristics at each time node by splitting the voltage-environment temperature full-time-domain fine-grained correlation feature matrix based on row vectors to obtain a set of voltage-environment temperature local time-domain fine-grained correlation feature vectors.

[0041] Specifically, the device load type special enhancement secondary subunit 1435-2 is used to perform deconvolution feature encoding enhancement on the one-hot encoding vector of the device load type to obtain an enhanced one-hot encoding vector of the device load type. Specifically, here, the enhanced one-hot encoding vector of the device load type has the same feature scale as each voltage-environment temperature local time-domain fine-grained correlation feature vector in the set of voltage-environment temperature local time-domain fine-grained correlation feature vectors. It can be expressed by the formula: where, is the one-hot encoding vector of the device load type, To calculate the Euclidean norm of a vector For deconvolution coding For the deconvolution coding weight matrix Is the one-hot encoding enhanced vector of the device load type

[0042] Accordingly, since the one-hot encoding vector of the device load type (discrete categorical feature) and the voltage-environment temperature local time-domain fine-grained correlation feature vector (continuous time-series correlation feature) essentially belong to heterogeneous data modalities, there are significant differences in the feature expression dimensions and semantic spaces between the two. If cross-modal feature interaction is directly performed, due to the mismatch in feature scale between the sparsity of the load type encoding and the density of the local time-domain correlation features, it will be difficult for the model to capture the modulation effect of the load characteristics on the voltage-temperature coupling relationship in a specific period. Therefore, in this application, the one-hot encoding vector of the device load type is enhanced by deconvolution feature encoding to map it to the same feature space as the voltage-environment temperature local time-domain fine-grained correlation feature, obtaining the one-hot encoding enhanced vector of the device load type

[0043] Specifically, the device load-external variable response secondary subunit 1435-3 is used to perform monomer semantic query responses on each voltage-environment temperature local time-domain fine-grained correlation feature vector in the set of the one-hot encoding enhanced vector of the device load type and the voltage-environment temperature local time-domain fine-grained correlation feature vectors, so as to obtain a set of device load-external variable monomer semantic query response score encoding vectors. It can be expressed by the formula as follows Wherein Is the one-hot encoding enhanced vector of the device load type Is the th voltage-environment temperature local time-domain fine-grained correlation feature vector in the set of voltage-environment temperature local time-domain fine-grained correlation feature vectors Is the vector concatenation operation Is the th monomer semantic query response weight matrix in the set of monomer semantic query response weight matrices Is the th monomer semantic query response bias vector in the set of monomer semantic query response bias vectors Is the hyperbolic tangent activation function Is the th device load-external variable monomer semantic query response score encoding vector in the set of device load-external variable monomer semantic query response score encoding vectors

[0044] It should be understood that the dynamic adjustment of the voltage threshold of the megasonic wave generator requires precise quantification of the modulation effect of the device load type on the voltage-environment temperature correlation. As a discrete operating condition identifier, the device load type has a modal gap with the continuous time-domain environment-voltage correlation characteristics - simple feature splicing or weighted fusion will destroy the semantic independence of the load type, resulting in the model confusing the boundary between process requirements and environmental interference. Therefore, in this application, the one-hot encoded enhanced vector of the device load type and each voltage-environment temperature local time-domain fine-grained correlation feature vector are respectively subjected to single-body semantic query responses to achieve precise decoupling of the load demand and the environmental coupling effect through semantic-driven cross-modal feature interaction, and a set of device load-external variable single-body semantic query response score encoding vectors is obtained.

[0045] Specifically, in the embodiment of this application, the device load-external variable aggregation secondary subunit 1435-4 is used for: Performing canonical constraint optimization of intrinsic alignment in the time-varying feature space on the set of device load-external variable single-body semantic query response score encoding vectors to obtain a set of device load-external variable single-body semantic query response score optimized encoding vectors, which can be expressed by the formula: Among them, is the th device load-external variable composite feature vector in the set of device load-external variable composite feature vectors, is the th device load-external variable interaction potential energy vector in the set of device load-external variable interaction potential energy vectors, is the th device load-external variable coupling constant in the set of device load-external variable coupling constants, is the th device load-external variable covariance matrix in the set of device load-external variable covariance matrices, is the th device load-external variable single-body semantic query response score optimized encoding vector in the set of device load-external variable single-body semantic query response score optimized encoding vectors; Based on the self-distribution characteristics of the feature set of the set of optimized coding vectors for device load-external variable monomer semantic query response scores, calculate the monomer semantic response matching degree for each optimized coding vector of device load-external variable monomer semantic query response scores in the set of optimized coding vectors for device load-external variable monomer semantic query response scores to obtain a set of device load-external variable monomer semantic response matching degrees, which can be expressed by the formula: where, is the th optimized coding vector of device load-external variable monomer semantic query response scores in the set of optimized coding vectors for device load-external variable monomer semantic query response scores, is the th optimized coding vector of device load-external variable monomer semantic query response scores, is the transpose operation, is the vector multiplication, is the exponential function value with the natural constant as the base, is the number of vectors in the set of optimized coding vectors for device load-external variable monomer semantic query response scores, is the normalization function, is the th device load-external variable monomer semantic response matching degree in the set of device load-external variable monomer semantic response matching degrees; Perform relational gating selection on the set of device load-external variable monomer semantic response matching degrees to obtain a set of device load-external variable monomer semantic response self-attention weights, which can be expressed by the formula: where, is the preset threshold, is the gating masking operation, is the th device load-external variable monomer semantic response self-attention weight in the set of device load-external variable monomer semantic response self-attention weights; Aggregate the set of optimized coding vectors for device load-external variable monomer semantic query response scores based on the set of device load-external variable monomer semantic response self-attention weights to obtain the multivariate intra-collaborative interaction coding feature vector, which can be expressed by the formula: where, is the multivariate intra-collaborative interaction coding feature vector.

[0046] Specifically, when the single - entity semantic query unit performs a single - entity semantic query response through the direct feature splicing of the one - hot encoded enhanced vector of the device load type and the corresponding voltage - ambient temperature local time - domain fine - grained correlation feature vector it is expected to improve the representation accuracy of the device load - external variable single - entity semantic response matching degree by optimizing the time - varying feature representation from the feature space to the semantic query encoding space and enhancing the intrinsic alignment characteristics between the feature space and the semantic query encoding space.

[0047] Based on this, if the one - hot encoded enhanced vector of the device load type and the corresponding voltage - ambient temperature local time - domain fine - grained correlation feature vector are concatenated to construct a device load - external variable composite feature vector that is then first calculate the device load - external variable interaction potential energy vector under the mapping representation: That is, taking the device load - external variable composite feature vector itself as the time - varying eigenvector representation to obtain the projection representation to the eigenstate under the space mapping process.

[0048] Then, perform the intrinsic alignment between the feature space and the semantic query encoding space: where is the device load - external variable coupling constant, calculated in the same way as during deconvolution enhancement, for example, to maintain gauge symmetry, that is: And introduce the device load - external variable covariant matrix that is, let to establish a dynamic constraint relationship.

[0049] Thus, with the device load - external variable interaction potential energy vector as the intrinsic alignment generation operator, through the adaptation trajectory under gauge - invariant constraints as the projection normalization constraint guarantee, the representation accuracy of the device load - external variable single - entity semantic response matching degree is improved.

[0050] Accordingly, the semantic matching results of the device load type and the voltage-environment temperature local time-domain correlation features (i.e., the set of monomer semantic query response score encoding vectors) often contain a large amount of fine-grained interaction information. However, the key correlation patterns in different time periods may vary significantly due to the load characteristics and environmental dynamic changes. Therefore, in order to dynamically evaluate the semantic significance of each score vector in the global context, this application calculates the monomer semantic response matching degree of each device load-external variable monomer semantic query response score optimized encoding vector based on the self-distribution characteristics of the feature set of the set of device load-external variable monomer semantic query response score optimized encoding vectors to obtain the set of device load-external variable monomer semantic response matching degrees.

[0051] It should be understood that the set of semantic matching degrees of the device load type and the voltage-environment temperature correlation features contains the key interaction information in different time periods. However, the original distribution of these matching degrees often contains noise interference and redundant correlations. Therefore, in order to selectively enhance and suppress the set of semantic matching degrees, this application performs relationship gating selection on the set of device load-external variable monomer semantic response matching degrees to dynamically calibrate the contribution ratio of the matching degrees in each time period to the final attention weight, and obtains the set of device load-external variable monomer semantic response self-attention weights.

[0052] Finally, based on the set of device load-external variable monomer semantic response self-attention weights, the set of device load-external variable monomer semantic query response score optimized encoding vectors is aggregated to obtain the multi-variable internal collaborative interaction coding feature vector. That is, a high weight is assigned to the time period with high load sensitivity and significant environmental correlation (such as the minute-level temperature accumulation period under the cleaning load) to strengthen its contribution to the final feature vector; while for the instantaneous fluctuations or noise interferences irrelevant to the load (such as the accidental correlation caused by the second-level power grid transient), their impacts are suppressed through low weights. The aggregation process is achieved through weighted summation, so as to fuse the multi-dimensional local semantic correlation features into a unified global interaction coding vector, representing the comprehensive action mode of the load-environment-voltage dynamic coupling.

[0053] Specifically, in the embodiment of this application, the range setting unit 144 is used to optimize the voltage initial upper limit value and the voltage initial lower limit value based on the multi-variable internal collaborative interaction coding feature vector to obtain the voltage preset range. More specifically, in the embodiment of this application, the range setting unit includes: an adjustment coefficient generation sub-unit, which is used to perform feature decoding on the multi-variable internal collaborative interaction coding feature vector to obtain an upper limit dynamic adjustment coefficient and a lower limit dynamic adjustment coefficient; a voltage preset range optimization sub-unit, which is used to dynamically optimize the voltage initial upper limit value and the voltage initial lower limit value based on the upper limit dynamic adjustment coefficient and the lower limit dynamic adjustment coefficient to obtain the voltage preset range.

[0054] More specifically, in the embodiments of the present application, the adjustment coefficient generation subunit is configured to: respectively use an upper limit adjuster based on a decoder and a lower limit adjuster based on a decoder to perform feature decoding on the multi-variable intra-cooperative interaction coding feature vector to obtain the upper limit dynamic adjustment coefficient and the lower limit dynamic adjustment coefficient. It should be understood that the multi-variable intra-cooperative interaction coding feature vector contains complex cooperative interaction information among multiple variables such as device load, voltage, and ambient temperature. The decoder-based adjuster can deeply explore the potential features and relationships related to upper and lower limit adjustments by performing decoding operations on these high-dimensional feature vectors. In particular, during the actual operation of the device, factors such as device load, voltage, and ambient temperature are dynamically changing, so it is necessary to dynamically adjust the relevant upper and lower limits. The decoder-based adjuster can perform real-time decoding on the input multi-variable intra-cooperative interaction coding feature vector according to its changes and generate corresponding dynamic adjustment coefficients. This dynamic adjustment ability enables the system to better adapt to different operating conditions, respond in a timely manner to changes in various factors, and ensure the stable operation and performance optimization of the device.

[0055] The specific implementation process is as follows: After the multi-variable intra-cooperative interaction coding feature vector is generated, it is simultaneously input into the upper limit adjuster based on a decoder and the lower limit adjuster based on a decoder. These two adjusters are essentially designed based on the decoder architecture and can analyze high-dimensional and complex feature vectors to extract key information related to voltage upper and lower limit adjustments.

[0056] Taking the upper limit adjuster as an example, the internal decoder structure first extracts features from the input multi-variable intra-cooperative interaction coding feature vector. Through a series of decoding layer operations, the complex cooperative relationships among multiple variables such as device load, voltage fluctuation, and ambient temperature change contained in the vector are disassembled and analyzed. For example, the convolutional layer in the decoder can capture local patterns in the feature vector, and the fully connected layer is responsible for integrating information from different regions to gradually extract features closely related to voltage upper limit adjustment.

[0057] During the feature extraction process, the decoder performs weighted calculation and non-linear transformation on the feature vector according to pre-set model parameters and training rules. These parameters are optimized during the system training stage using a large amount of historical operation data and actual working conditions to ensure that the decoder can accurately identify the factors affecting the voltage upper limit from the feature vector. For example, if an increase in device load leads to an increase in voltage demand under certain working conditions, the decoder can accurately capture the correlation between the load type and voltage through the analysis of the feature vector and assign a higher weight to the corresponding feature.

[0058] After layers of decoding and calculation, the upper limit adjuster finally outputs a value representing the dynamic upper limit adjustment coefficient. This coefficient reflects the magnitude and direction of adjustment required relative to the initial voltage upper limit value under the current device operating state.

[0059] The working principle of the lower limit adjuster is similar to that of the upper limit adjuster. It also performs feature extraction and decoding operations on the multi-variable intra-cooperative interaction coding feature vector, but focuses on information related to voltage lower limit adjustment. Through comprehensive analysis of factors such as historical voltage data, load type, and ambient temperature changes, the decoder identifies the key features affecting the voltage lower limit and calculates the corresponding dynamic lower limit adjustment coefficient. These two adjusters work independently but cooperate with each other to generate the dynamic upper limit adjustment coefficient and the dynamic lower limit adjustment coefficient respectively.

[0060] Specifically, the voltage preset range optimization subunit is used to dynamically optimize the voltage initial upper limit value and the voltage initial lower limit value based on the dynamic upper limit adjustment coefficient and the dynamic lower limit adjustment coefficient to obtain the voltage preset range. That is, the operation of the megasonic wave generator is affected by various factors such as device load type, voltage timing fluctuation, and ambient temperature fluctuation. The multi-variable intra-cooperative interaction coding feature vector obtained in the previous steps reflects the complex relationships among these factors, and the dynamic upper limit adjustment coefficient and the dynamic lower limit adjustment coefficient are specific manifestations of this relationship in terms of voltage adjustment. Optimizing the voltage initial limit values based on these coefficients can comprehensively consider the comprehensive influence of multiple variables and make the voltage preset range more in line with the actual operating conditions. In particular, here the voltage preset range is obtained by multiplying the dynamic upper limit adjustment coefficient and the dynamic lower limit adjustment coefficient by the corresponding voltage initial upper limit value and the voltage initial lower limit value respectively.

[0061] After the system obtains the dynamic upper limit adjustment coefficient and the dynamic lower limit adjustment coefficient, it starts the optimization process for the voltage initial upper limit value and the voltage initial lower limit value. These two adjustment coefficients are decoded based on the multi-variable intra-cooperative interaction coding feature vector and reflect the comprehensive influence of various factors such as device load, voltage timing fluctuation, and ambient temperature change on the voltage preset range.

[0062] First, for the optimization of the voltage initial upper limit value, the system will perform a specific operation on the dynamic upper limit adjustment coefficient and the voltage initial upper limit value. Generally, multiplication operation is used to achieve the adjustment. For example, if the dynamic upper limit adjustment coefficient is greater than 1, it means that under the current operating conditions, considering various dynamic factors, the megasonic wave generator can withstand a higher voltage than the initial upper limit value, and the system will correspondingly increase the upper limit of the voltage preset range; if the dynamic upper limit adjustment coefficient is less than 1 but greater than 0, it indicates that the initial upper limit value needs to be appropriately reduced to ensure that the device operates within the safe voltage range and avoid damage to the device due to excessive voltage.

[0063] Similarly, for the initial lower limit value of the voltage, the system calculates the lower limit dynamic adjustment coefficient with the initial lower limit value of the voltage. If the lower limit dynamic adjustment coefficient is greater than 1, it indicates that the device's demand for the minimum voltage has increased under the current working conditions, and the system will increase the lower limit of the voltage preset range to ensure that the device obtains sufficient electrical energy for driving; when the lower limit dynamic adjustment coefficient is less than 1 but greater than 0, it means that the lower limit value can be appropriately reduced. In this way, under light load or other specific working conditions, the voltage can be adjusted more flexibly, improving the energy utilization efficiency without affecting the normal operation of the device.

[0064] After completing the adjustment of the initial upper limit value and the lower limit value of the voltage, the system will perform a rationality check on the newly obtained upper limit value and lower limit value. This is to prevent unreasonable situations from occurring in the adjusted voltage preset range, such as the upper and lower limit values being too close or even reversed. If unreasonable values are found during the check process, the system will activate the correction mechanism and fine-tune the adjustment result according to the preset rules and historical data to ensure that the finally obtained voltage preset range not only conforms to the current operating state of the device but also guarantees the long-term stable and safe operation of the device. After this series of operations, the finally determined voltage preset range will be used for subsequent judgment of the voltage digital signal and become an important basis for ensuring the normal operation of the megasonic wave generator.

[0065] Specifically, the voltage judgment module 150 is used to judge whether the voltage digital signal is within the voltage preset range to obtain a judgment result. It should be understood that when the device is designed and manufactured, it has its specific voltage operating range. Beyond this range, it may cause the electronic components inside the device to be damaged due to excessive voltage or unable to work properly due to too low voltage. By continuously judging whether the voltage digital signal is within the preset range, voltage abnormalities can be detected in a timely manner, and corresponding measures can be taken to protect the device from damage and extend the service life of the device.

[0066] After receiving the voltage digital signal, special isolation devices are used to process the voltage digital signal. These isolation devices generally adopt electromagnetic isolation or optoelectronic isolation technology, which can effectively block the electrical connection between the mains power and the low-voltage circuit while ensuring signal integrity, prevent damage to the low-voltage monitoring circuit caused by high-voltage mains power, avoid the risk of electric shock to operators, and ensure the safety of the entire judgment process.

[0067] After the isolation process of the voltage digital signal is completed, it enters the formal judgment stage. The voltage judgment module continuously reads the isolated voltage digital signal and compares it with the preset voltage range. During the comparison process, the judgment module compares the voltage digital signal with the upper and lower limit values of the preset voltage range respectively. When the voltage digital signal is greater than or equal to the lower limit value and less than or equal to the upper limit value, the judgment module determines that the voltage is within the normal range at this time, generates a judgment result representing normal voltage, and promptly feeds back this result to the instruction generation module of the system to indicate that the power supply voltage of the megasonic wave generator meets the equipment operation requirements and the equipment can work normally. Conversely, if the voltage digital signal is less than the lower limit value or greater than the upper limit value, the judgment module will generate a judgment result of abnormal voltage.

[0068] Specifically, the instruction generation module 160 is used to generate a control instruction for the megasonic wave generator based on the judgment result. That is to say, the megasonic wave generator works under different voltage conditions. When the voltage digital signal exceeds the preset range, it may cause damage to the equipment. For example, too high voltage may cause the electronic components of the equipment to overheat and break down, shortening the service life of the equipment; too low voltage may cause the equipment to fail to start normally or work unstably. By generating a control instruction based on the judgment result, the equipment can be adjusted in time to restore the voltage to the safe range, thus ensuring the safe and stable operation of the equipment.

[0069] After the voltage judgment module gives the judgment result, the instruction generation module will immediately receive this information and parse it. If the judgment result indicates that the voltage digital signal is within the preset rated voltage range, it means that the current power supply voltage of the megasonic wave generator is normal and can ensure the stable operation of the equipment. At this time, the control instruction generated by the instruction generation module is to maintain the current working state of the equipment, let the megasonic wave generator continue to work normally, and ensure that the production process is not affected.

[0070] Once the judgment result shows that the voltage is too low or too high and exceeds the preset range, the instruction generation module will quickly make a series of responses to ensure the safety of the equipment. To prevent the equipment from being burned out due to abnormal voltage input, the instruction generation module will first generate an instruction to prohibit the equipment from working, immediately cut off the working circuit of the equipment, and stop the operation of the megasonic wave generator to avoid continuous damage to the equipment caused by abnormal voltage.

[0071] Meanwhile, the instruction generation module triggers the alarm mechanism to generate an instruction for displaying undervoltage or overvoltage alarm information. This instruction controls the display interface of the device or an external alarm device to prominently show the current voltage abnormality. For example, it displays prompts such as "Undervoltage, please check the power supply" or "Overvoltage, dangerous!" on the device operation panel, and may be accompanied by methods such as light flashing and sound alarms to promptly notify the operator of the voltage problem in the device, so that the operator can quickly take corresponding measures, such as checking the power supply and troubleshooting circuit faults.

[0072] In addition, the instruction generation module may also generate some auxiliary control instructions for recording relevant data of voltage abnormalities, including the time when the abnormality occurs, the specific value of the abnormal voltage, etc. These data are of great value for subsequent analysis of the reasons for equipment failures, optimization of the voltage monitoring system, and formulation of equipment maintenance plans. Throughout the process, based on the judgment results, the instruction generation module performs corresponding operations to comprehensively ensure the safety of the megasonic wave generator, prevent the equipment from being damaged due to voltage problems, and ensure the safety and stability of the production process.

[0073] In summary, the voltage monitoring system 100 of the megasonic wave generator based on the embodiments of the present application is elucidated. It first conditions and performs analog-to-digital conversion on the voltage analog signal to convert it into a digital signal. Then, by analyzing historical voltage data and timing fluctuation patterns, and synchronously analyzing the environmental temperature change pattern, a dynamic mapping relationship between the temperature drift effect and the voltage threshold is established. Combining the load characteristic coding, quantifying the voltage tolerance requirements in different process stages, and using the response aggregation coding mechanism for multi-variable collaborative modeling to generate multi-dimensional interaction features reflecting the real-time operating state, thereby dynamically adjusting the voltage preset range and generating control instructions. This method dynamically optimizes the voltage preset range, accurately adapts to different working conditions and dynamic factors, avoids problems caused by broad or narrow thresholds, ensures that the equipment maintenance strategy fits the operating state, improves the system robustness and process stability, and meets the requirements of high-precision manufacturing.

[0074] As described above, the voltage monitoring system 100 of the megasonic wave generator according to the embodiments of the present application can be implemented in various wireless terminals, such as a server with a voltage monitoring algorithm for the megasonic wave generator. In a possible implementation, the voltage monitoring system 100 of the megasonic wave generator according to the embodiments of the present application can be integrated into the wireless terminal as a software module and / or a hardware module. For example, the voltage monitoring system 100 of the megasonic wave generator can be a software module in the operating system of the wireless terminal, or can be an application program developed for the wireless terminal; of course, the voltage monitoring system 100 of the megasonic wave generator can also be one of the many hardware modules of the wireless terminal.

[0075] Alternatively, in another example, the voltage monitoring system 100 of the megasonic wave generator and the wireless terminal may also be separate devices, and the voltage monitoring system 100 of the megasonic wave generator can be connected to the wireless terminal through a wired and / or wireless network and transmit interaction information in accordance with a predefined data format.

[0076] The implementations of the present disclosure have been described above. The above description is exemplary and not exhaustive. And it is not limited to the disclosed implementations. Many modifications and variations are obvious to those of ordinary skill in the art without departing from the scope and spirit of the described implementations. The choice of terms used herein is intended to best explain the principles of the implementations, practical applications, or improvements to the technology in the market, or to enable other ordinary skill in the art to understand the various implementation manners disclosed herein.

Claims

1. A voltage monitoring system for a megasonic wave generator, characterized in that, Comprising: A signal receiving module, configured to receive a voltage analog signal; A signal conditioning module, configured to condition the voltage analog signal to obtain a conditioned voltage analog signal; An analog-to-digital conversion module, configured to perform analog-to-digital conversion on the conditioned voltage analog signal to obtain a voltage digital signal; A voltage setting module, configured to set a voltage preset range. Wherein, the voltage setting module includes: a range extraction unit, configured to extract an initial voltage upper limit value and an initial voltage lower limit value; a parameter acquisition unit, configured to acquire historical voltage data, the current device load type, and the environmental temperature time series; a variable interaction unit, configured to perform response aggregation between device load - external variable features on the historical voltage data, the current device load type, and the environmental temperature time series to obtain a multi-variable internal collaborative interaction coding feature vector; a range setting unit, configured to optimize the initial voltage upper limit value and the initial voltage lower limit value based on the multi-variable internal collaborative interaction coding feature vector to obtain the voltage preset range; A voltage judgment module, configured to judge whether the voltage digital signal is within the voltage preset range to obtain a judgment result; An instruction generation module, configured to generate a control instruction for a megasonic wave generator based on the judgment result.

2. The voltage monitoring system for a megasonic wave generator according to claim 1, characterized in that, The variable interaction unit includes: A historical voltage time series coding sub-unit, configured to perform time series coding on the historical voltage data to obtain a voltage time series fluctuation pattern feature coding vector; An environmental temperature time series coding sub-unit, configured to perform time series coding on the environmental temperature time series to obtain an environmental temperature fluctuation time series pattern feature coding vector; A device load type coding sub-unit, configured to perform one-hot coding on the current device load type to obtain a device load type one-hot coding vector; A voltage - environmental temperature fine-grained association sub-unit, configured to calculate a voltage - environmental temperature full-time domain fine-grained association feature matrix between the voltage time series fluctuation pattern feature coding vector and the transposed vector of the environmental temperature fluctuation time series pattern feature coding vector; A multi-variable internal collaboration sub-unit, configured to perform response aggregation coding between device load - external variables dominated by semantic query on the device load type one-hot coding vector and the voltage - environmental temperature full-time domain fine-grained association feature matrix to obtain the multi-variable internal collaborative interaction coding feature vector.

3. The voltage monitoring system for a megasonic wave generator according to claim 2, characterized in that, The historical voltage time series coding sub-unit is configured to: perform time series coding based on causal dilated convolution on the historical voltage data to obtain the voltage time series fluctuation pattern feature coding vector.

4. The voltage monitoring system for a megasonic wave generator according to claim 3, characterized in that, The voltage - environmental temperature fine-grained association sub-unit is configured to: multiply the voltage time series fluctuation pattern feature coding vector by the transposed vector of the environmental temperature fluctuation time series pattern feature coding vector to obtain the voltage - environmental temperature full-time domain fine-grained association feature matrix.

5. The voltage monitoring system for a megasonic wave generator according to claim 2, characterized in that, The multi-variable internal collaboration sub-unit includes: An associated feature matrix local time series splitting secondary sub-unit, configured to split the voltage - environmental temperature full-time domain fine-grained association feature matrix based on row vectors to obtain a set of voltage - environmental temperature local time domain fine-grained association feature vectors; The device load type special enhanced secondary subunit is used to perform deconvolution feature encoding enhancement on the one-hot encoded vector of the device load type to obtain a one-hot encoded enhanced vector of the device load type; The device load-external variable response secondary subunit is used to perform single-body semantic query response on the one-hot encoded enhanced vector of the device load type and each voltage-environment temperature local time-domain fine-grained association feature vector in the set of voltage-environment temperature local time-domain fine-grained association feature vectors to obtain a set of device load-external variable single-body semantic query response score encoding vectors; The device load-external variable aggregation secondary subunit is used to perform dynamic aggregation based on self-attention weights on the set of device load-external variable single-body semantic query response score encoding vectors to obtain the multi-variable internal collaborative interaction encoding feature vector.

6. The voltage monitoring system for a megasonic wave generator according to claim 5, characterized in that, The one-hot encoded enhanced vector of the device load type has the same feature scale as each voltage-environment temperature local time-domain fine-grained association feature vector in the set of voltage-environment temperature local time-domain fine-grained association feature vectors.

7. The voltage monitoring system for a megasonic wave generator according to claim 6, characterized in that, The device load-external variable aggregation secondary subunit is used for: Performing canonical constraint optimization of time-varying feature space intrinsic alignment on the set of device load-external variable single-body semantic query response score encoding vectors to obtain a set of device load-external variable single-body semantic query response score optimized encoding vectors; Based on the feature set self-distribution characteristics of the set of device load-external variable single-body semantic query response score optimized encoding vectors, calculating the single-body semantic response matching degree for each device load-external variable single-body semantic query response score optimized encoding vector in the set of device load-external variable single-body semantic query response score optimized encoding vectors to obtain a set of device load-external variable single-body semantic response matching degrees; Performing relationship gating selection on the set of device load-external variable single-body semantic response matching degrees to obtain a set of self-attention weights of the device load-external variable single-body semantic response; Aggregating the set of device load-external variable single-body semantic query response score optimized encoding vectors based on the set of self-attention weights of the device load-external variable single-body semantic response to obtain the multi-variable internal collaborative interaction encoding feature vector.

8. The voltage monitoring system for a megasonic wave generator according to claim 7, characterized in that, The range setting unit includes: An adjustment coefficient generation subunit is used to perform feature decoding on the multi-variable internal collaborative interaction encoding feature vector to obtain an upper limit dynamic adjustment coefficient and a lower limit dynamic adjustment coefficient; A voltage preset range optimization subunit is used to dynamically optimize the initial voltage upper limit value and the initial voltage lower limit value based on the upper limit dynamic adjustment coefficient and the lower limit dynamic adjustment coefficient to obtain the voltage preset range.

9. The voltage monitoring system of the megasonic wave generator according to claim 8, wherein The adjustment coefficient generation subunit is used for: respectively using an upper limit adjuster based on a decoder and a lower limit adjuster based on a decoder to perform feature decoding on the multi-variable internal collaborative interaction encoding feature vector to obtain the upper limit dynamic adjustment coefficient and the lower limit dynamic adjustment coefficient.

Images