Method and device for calibrating scale body offset based on elevator current and medium

Abstract

The invention provides a method, device, equipment and medium for carrying out offset calibration on a scale based on elevator current, and the method comprises the steps: collecting a current signal, a vibration acceleration signal and an environment temperature and humidity signal of an elevator motor and a weight signal of a scale body in real time, and carrying out the preprocessing to form a data set; a prediction model based on multi-modal adversarial learning is constructed, the prediction model comprises a physical constraint flow and a process knowledge flow, and feature extraction irrelevant to working conditions is achieved through a gradient inversion layer; comparing the weight predicted by the prediction model with the weight signal of the scale body in real time, and when the continuous multiple deviations exceed a first preset value and the confidence coefficient is greater than a second preset value, determining that the scale body deviates; the zero point or range parameter of the scale body is automatically calibrated through a progressive parameter adjustment strategy according to the scale body offset type, and a calibration log is recorded, so that the problems that when manual equipment calibration is carried out, operation is tedious, consumed time is long, shutdown operation is needed, and equipment offset cannot be monitored in real time are solved.

Description

Technical Field

[0001] The present invention relates to the field of industrial intelligent measurement and control technology, and particularly to a method, device, equipment and medium for calibrating the offset of a weighing body based on the current of a hoist. Background Art

[0002] In the process of cement production, weighing devices such as belt scales and screw scales are widely used in the weighing of raw materials, semi-finished products and finished products. However, due to factors such as mechanical wear, material adhesion, and environmental temperature changes, the weighing devices will have zero-point drift and range drift, resulting in measurement errors and affecting production control and product quality.

[0003] Traditional calibration methods for weighing devices usually rely on manual calibration with weights or physical calibration. These methods are cumbersome to operate, time-consuming, require shutdown operations, and cannot monitor the device offset in real time, affecting production efficiency.

[0004] Therefore, there is an urgent need to propose a method for calibrating the offset of a weighing body based on the current of a hoist to solve the technical problems of cumbersome operation, long time consumption, requiring shutdown operations, and inability to monitor the device offset in real time when manually calibrating the device. Summary of the Invention

[0005] To overcome the problems existing in the related art, the present disclosure provides a method, device, equipment and medium for calibrating the offset of a weighing body based on the current of a hoist, so as to solve the technical problems of cumbersome operation, long time consumption, requiring shutdown operations, and inability to monitor the device offset in real time in the related art.

[0006] One or more embodiments of this specification provide a method for calibrating the offset of a weighing body based on the current of a hoist, including the following steps:

[0007] Real-time collect the current signal, vibration acceleration signal, ambient temperature and humidity signal of the hoist motor and the weight signal of the weighing body, and form a data set after preprocessing;

[0008] Construct a prediction model based on multi-modal adversarial learning. The prediction model includes a physical constraint flow and a process knowledge flow, and realizes feature extraction independent of working conditions through a gradient reversal layer, and train the prediction model through the data set;

[0009] Compare the weight predicted by the prediction model with the weight signal of the weighing body in real time. When the deviation exceeds the first preset value continuously for multiple times and the confidence level is greater than the second preset value, it is determined that the weighing body is offset;

[0010] According to the type of weighing body offset, automatically calibrate the zero point or range parameters of the weighing body through a progressive parameter adjustment strategy, and record the calibration log.

[0011] Preferably, the current signal, vibration acceleration signal, ambient temperature and humidity signal of the hoist motor, and the weight signal of the weighing body are collected in real time, and a data set is formed after preprocessing, which specifically includes the following steps:

[0012] Collect the current signal through a Hall sensor, perform wavelet packet decomposition on the current signal, and extract the ripple energy entropy;

[0013] Collect the vibration acceleration signal through a MEMS sensor, perform spectrum analysis on the vibration acceleration signal, and extract the main frequency amplitude ratio;

[0014] Collect the ambient temperature and humidity signal through an integrated sensor;

[0015] Form a data set from the preprocessed data.

[0016] Preferably, the prediction model based on multi-modal adversarial learning is constructed. The prediction model includes a physical constraint stream and a process knowledge stream, and feature extraction independent of working conditions is realized through a gradient reversal layer, which specifically includes the following steps:

[0017] Input the ripple energy entropy and the main frequency amplitude ratio into the physical constraint stream, and extract high-dimensional features through a residual convolutional network;

[0018] Input the humidity compensation coefficient and the set value of the material flow rate into the process knowledge stream, and obtain process features through a fully connected network;

[0019] The gradient reversal layer realizes feature extraction independent of working conditions.

[0020] Preferably, the confidence level is calculated through the Monte Carlo Dropout strategy, which specifically includes the following steps:

[0021] Randomly discard some neurons during the inference process of the prediction model, and repeat the inference 100 times;

[0022] Calculate the mean and standard deviation of the predicted values, and obtain the confidence interval as the confidence level.

[0023] Preferably, the zero point or range parameters of the weighing body are automatically calibrated through a progressive parameter adjustment strategy, which specifically includes the following steps:

[0024] Zero point calibration: New zero point = original zero point + (predicted value - measured value) × 0.8;

[0025] Range calibration: New range coefficient = original range coefficient × (measured value / predicted value);

[0026] The single adjustment amplitude does not exceed 0.2%, and the cumulative adjustment amplitude does not exceed 5%.

[0027] One or more embodiments of this specification provide a device for calibrating the offset of a weighing body based on the current of a hoist, including a data set acquisition module, a model construction module, a prediction module, and a calibration module;

[0028] The data set acquisition module is used to collect the current signal, vibration acceleration signal, ambient temperature and humidity signal of the hoist motor, and the weight signal of the weighing body in real time, and form a data set after preprocessing;

[0029] The model construction module is used to construct a prediction model based on multi-modal adversarial learning. The prediction model includes a physical constraint flow and a process knowledge flow, and realizes feature extraction independent of working conditions through a gradient reversal layer, and trains the prediction model through the data set;

[0030] The prediction module is used to compare the weight predicted by the prediction model with the weight signal of the weighing body in real time. When the deviation exceeds the first preset value continuously for multiple times and the confidence level is greater than the second preset value, it is determined that the weighing body is offset;

[0031] The calibration module is used to automatically calibrate the zero point or range parameters of the weighing body according to the offset type through a progressive parameter adjustment strategy, and record the calibration log.

[0032] Preferably, the data set acquisition module includes a current signal acquisition unit, a vibration acceleration signal acquisition unit, an ambient temperature and humidity signal acquisition unit, and a data set formation unit;

[0033] The current signal acquisition unit is used to collect the current signal through a Hall sensor, perform wavelet packet decomposition on the current signal, and extract the ripple energy entropy;

[0034] The vibration acceleration signal acquisition unit is used to collect the vibration acceleration signal through a MEMS sensor, perform spectrum analysis on the vibration acceleration signal, and extract the main frequency amplitude ratio;

[0035] The ambient temperature and humidity signal acquisition unit is used to collect the ambient temperature and humidity signal through an integrated sensor;

[0036] The data set formation unit is used to form a data set from the preprocessed data.

[0037] Preferably, the model construction module includes a high-dimensional feature extraction unit, a process extraction unit, and an irrelevant feature extraction unit;

[0038] The high-dimensional feature extraction unit is used to input the ripple energy entropy and the main frequency amplitude ratio into the physical constraint flow, and extract high-dimensional features through a residual convolutional network;

[0039] The process extraction unit is configured to input the humidity compensation coefficient and the set value of the material flow rate into the process knowledge stream, and obtain process features through a fully connected network.

[0040] The irrelevant feature extraction unit is configured to implement feature extraction independent of working conditions through the gradient reversal layer.

[0041] One or more embodiments of this specification provide a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the method for calibrating the offset of the weighing body based on the hoist current as described above.

[0042] One or more embodiments of this specification provide a computer-readable storage medium storing a computer program, and when the computer program is executed by a processor, it implements the steps of the method for calibrating the offset of the weighing body based on the hoist current as described above.

[0043] A method, device, equipment, and medium for calibrating the offset of a weighing body based on the hoist current provided by the present disclosure have the following advantages: By collecting the current signal, vibration acceleration signal, ambient temperature and humidity signal, and the weight signal of the weighing body of the hoist motor in real time, and forming a data set after preprocessing, the fusion of multi-dimensional data is realized, effectively eliminating sensor noise and data asynchronization problems, and improving the reliability and consistency of the data set; constructing a prediction model based on multi-modal adversarial learning, the prediction model includes a physical constraint flow and a process knowledge flow, and implements feature extraction independent of working conditions through the gradient reversal layer, and training the prediction model through the data set enables the model to capture the influence of both device states and process conditions simultaneously, enhancing the generalization ability of the model in different environments; comparing the weight predicted by the prediction model with the weight signal of the weighing body in real time, and when the deviation exceeds the first preset value continuously for multiple times and the confidence level is greater than the second preset value, it is determined that the weighing body is offset. The dual determination conditions avoid misjudgment and missed judgment, can timely detect the offset problem of the weighing body, avoid process fluctuations caused by measurement errors, can be dynamically adjusted according to the actual working conditions, enhancing the flexibility and adaptability of the system; according to the type of weighing body offset, automatically calibrate the zero point or range parameters of the weighing body through a progressive parameter adjustment strategy, and record the calibration log. The progressive adjustment strategy avoids over-adjustment or under-adjustment problems and can complete calibration without stopping the machine, significantly improving production efficiency and equipment utilization rate. Recording information such as the time, offset type, and adjustment amount of each calibration provides data support for equipment maintenance and process optimization, and can effectively solve the problem of weighing body offset in cement plants, having a wide range of application prospects. Description of the Drawings

[0044] To more clearly illustrate the technical solutions in one or more embodiments of this specification or the prior art, the following will briefly introduce the accompanying drawings required for the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description are only some embodiments recorded in this specification. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0045] Figure 1 A schematic flowchart of a method for calibrating the offset of a weighing body based on the current of a hoist provided for one or more embodiments of this specification;

[0046] Figure 2 A schematic structural diagram of a device for calibrating the offset of a weighing body based on the current of a hoist provided for one or more embodiments of this specification;

[0047] Figure 3 A schematic structural diagram of a computer device provided for one or more embodiments of this specification. Detailed implementation manners

[0048] In order to enable those skilled in the art to better understand the technical solutions in one or more embodiments of this specification, the following will clearly and completely describe the technical solutions in one or more embodiments of this specification in conjunction with the accompanying drawings in one or more embodiments of this specification. Obviously, the described embodiments are only a part of the embodiments of this specification, rather than all the embodiments. Based on one or more embodiments of this specification, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of this invention document.

[0049] The following will make a detailed description of the present invention in conjunction with the detailed implementation manners and the accompanying drawings of the specification.

[0050] Method embodiment

[0051] According to an embodiment of the present invention, a method for calibrating the offset of a weighing body based on the current of a hoist is provided. As Figure 1 shown, it is a schematic flowchart of the method for calibrating the offset of a weighing body based on the current of a hoist provided for this embodiment. The method for calibrating the offset of a weighing body based on the current of a hoist according to an embodiment of the present invention includes the following steps:

[0052] S110. Real-time collect the current signal, vibration acceleration signal, ambient temperature and humidity signal of the hoist motor, and the weight signal of the weighing body, and form a data set after preprocessing.

[0053] Specifically, on the hoist side, a Hall sensor (range 0 - 100A, accuracy ±0.5%) is used to collect current signals. The current signals are decomposed by wavelet packet decomposition to obtain the sub-band energy in the 5 - 50Hz frequency band. The probability distribution of each sub-band energy is calculated, and the ripple energy entropy is calculated through the Shannon entropy formula to achieve the extraction of the ripple energy entropy.

[0054] A vibration acceleration signal is collected by a MEMS (Micro-Electro-Mechanical System) sensor (0 - 5kHz). The vibration acceleration signal is subjected to spectral analysis using the fast Fourier transform, the maximum amplitude in the 5 - 50Hz frequency band is extracted, and the ratio of the maximum amplitude to the total energy is calculated to obtain the main frequency amplitude ratio.

[0055] An environmental temperature and humidity signal is collected by an integrated sensor (-20°C to 80°C).

[0056] On the scale body side, the 4 - 20mA weight signal and the speed encoder pulse signal of the existing scale body - belt scale / screw scale are read.

[0057] The preprocessed data is formed into a data set, and data synchronization is performed through an edge computing gateway.

[0058] The data collector selects an industrial-grade edge computing gateway (such as Advantech UNO-2484G), which supports the Modbus RTU / TCP protocol. The sampling frequency is uniformly 100Hz. The edge computing gateway uses the NVIDIA Jetson Nano platform, which supports TensorRT accelerated inference. The edge computing gateway communicates with the PLC system through the OPC UA protocol, and the calibration instruction is written into the scale body controller through Modbus RTU.

[0059] S120. Build a prediction model based on multi-modal adversarial learning. The prediction model includes a physical constraint stream and a process knowledge stream, and realizes condition-independent feature extraction through a gradient reversal layer. The prediction model is trained with the data set.

[0060] S130. Compare the weight predicted by the prediction model with the weight signal of the scale body in real time. When the deviation exceeds the first preset value continuously for multiple times and the confidence level is greater than the second preset value, it is determined that the scale body is offset. The first preset value is divided into three levels:

[0061] First-level threshold: deviation > 2%, triggering log recording;

[0062] Second-level threshold: deviation > 3% and lasting for 30 minutes, triggering automatic calibration;

[0063] Third-level threshold: deviation > 5%, triggering shutdown alarm.

[0064] S140. According to the type of scale body offset (zero - point offset, range drift, non - linear error, random error, and temperature drift), automatically calibrate the zero - point or range parameters of the scale body through a progressive parameter adjustment strategy. The adjustment range of the range coefficient for a single calibration does not exceed 5%, and the cumulative adjustment range does not exceed 15%. After calibration, automatically run the verification process. When the error is less than 1% for 10 consecutive minutes, it is determined that the calibration is successful, record the calibration log, and provide a visual monitoring interface to display the current - weight relationship curve, historical error trend, and calibration record in real - time.

[0065] The method provided in this embodiment realizes the fusion of multi - dimensional data by collecting the current signal, vibration acceleration signal, ambient temperature and humidity signal of the hoist motor and the weight signal of the scale body in real - time, and forming a data set after pre - processing, effectively eliminating sensor noise and data asynchronization problems, and improving the reliability and consistency of the data set; constructing a prediction model based on multi - modal adversarial learning, the prediction model includes a physical constraint flow and a process knowledge flow, and realizes feature extraction independent of working conditions through a gradient reversal layer. Train the prediction model with the data set so that the model can simultaneously capture the influence of equipment status and process conditions, enhancing the generalization ability of the model in different environments; compare the weight predicted by the prediction model with the weight signal of the scale body in real - time. When the deviation exceeds the first preset value continuously for multiple times and the confidence level is greater than the second preset value, it is determined that the scale body is offset. The dual - determination condition avoids misjudgment and missed judgment, can timely detect the scale body offset problem, and avoid process fluctuations caused by measurement errors. It can be dynamically adjusted according to the actual working conditions, enhancing the flexibility and adaptability of the system; according to the type of scale body offset, automatically calibrate the zero - point or range parameters of the scale body through a progressive parameter adjustment strategy, and record the calibration log. The progressive adjustment strategy avoids over - adjustment or under - adjustment problems and can complete calibration without stopping the machine, significantly improving production efficiency and equipment utilization rate. Record information such as the time, offset type, and adjustment amount of each calibration, providing data support for equipment maintenance and process optimization, and can effectively solve the problem of scale body offset in cement plants, with broad application prospects.

[0066] In one embodiment, S120. Construct a prediction model based on multi - modal adversarial learning. The prediction model includes a physical constraint flow and a process knowledge flow, and realizes feature extraction independent of working conditions through a gradient reversal layer, specifically including the following steps:

[0067] Input the ripple energy entropy and the main frequency amplitude ratio into the physical constraint flow, and extract high - dimensional features through a residual convolutional network (ResNet - 18).

[0068] Input the humidity compensation coefficient and the set value of material flow rate into the process knowledge flow, and obtain process features through a fully connected network. The set value of material flow rate is usually set by process engineers according to production plans and quality requirements, and comes from the following systems or devices:

[0069] DCS (Distributed Control System):

[0070] DCS is the core control system of a cement plant, responsible for monitoring and adjusting the operating parameters of each process link.

[0071] The set value of material flow rate is usually generated by the raw material proportioning module or the feeding control module in the DCS.

[0072] PLC (Programmable Logic Controller):

[0073] In a production line with a high degree of automation, the PLC is responsible for executing the control instructions issued by the DCS.

[0074] The set value of material flow rate is stored through the registers or variables of the PLC.

[0075] Manual setting:

[0076] In some cases, the operator can manually input the set value of material flow rate through the HMI (Human Machine Interface).

[0077] The gradient reversal layer realizes condition-independent feature extraction, which is used to eliminate the influence of condition changes on model prediction. Location of the gradient reversal layer (Gradient Reversal Layer, GRL): Before the fusion layer connecting the physical constraint flow and the process knowledge flow. Adversarial loss function: Minimize the cross-entropy loss of the minimal condition classifier (MLP), forcing the model to learn condition-independent features.

[0078] Training process:

[0079] Pre-train the physical constraint flow:

[0080] Freeze the process knowledge flow and train only with physical features for 50 epochs with a learning rate of 1e-3.

[0081] Objective: Initially establish the correlation between current / vibration and weight.

[0082] Joint adversarial training:

[0083] Unfreeze all layers, add the adversarial loss, and train for 200 epochs.

[0084] Early stopping strategy: terminate if the validation set loss does not decrease for 10 consecutive rounds.

[0085] The method provided in this embodiment reduces the mean absolute error between the predicted weight and the actual weight through the synergistic effect of the physical constraint flow and the process knowledge flow. The model can maintain stable performance under different working conditions, reduce the calibration frequency caused by environmental changes, and avoid process fluctuations caused by measurement errors.

[0086] In one embodiment, the confidence level is calculated by the Monte Carlo Dropout strategy, which specifically includes the following steps:

[0087] During the inference process of the prediction model, randomly discard some neurons, repeat the inference 100 times, perform active learning on low-confidence samples (<90%), and expand the training set.

[0088] Calculate the mean and standard deviation of the predicted values to obtain the confidence interval as the confidence level.

[0089] The method provided in this embodiment can effectively quantify the uncertainty of model prediction, reduce the contingency of single inference results, and enhance the credibility of the model in practical applications.

[0090] In one embodiment, the zero point or range parameters of the scale are automatically calibrated through a progressive parameter adjustment strategy, which specifically includes the following steps:

[0091] Zero point calibration: new zero point = original zero point + (predicted value - measured value) × 0.8;

[0092] Range calibration: new range coefficient = original range coefficient × (measured value / predicted value);

[0093] The single adjustment amplitude does not exceed 0.2%, and the cumulative adjustment amplitude does not exceed 5%. After calibration, the verification process is automatically run, and the calibration is determined to be successful when the error is less than 1% for 10 consecutive minutes.

[0094] The method provided in this embodiment calibrates the zero point and range parameters of the scale respectively, effectively improving the measurement accuracy of the scale, avoiding system instability caused by excessive parameter adjustment, realizing automatic calibration of the scale, reducing manual intervention, improving calibration efficiency, and reducing labor costs and operation errors.

[0095] Device embodiment

[0096] According to an embodiment of the present invention, a device for calibrating the offset of a scale based on the hoist current is provided, as Figure 2As shown in the figure, it is a schematic structural diagram of a device for calibrating the offset of a weighing body based on the hoist current. The device for calibrating the offset of a weighing body based on the hoist current according to an embodiment of the present invention includes a data set acquisition module 21, a model construction module 22, a prediction module 23, and a calibration module 24.

[0097] The data set acquisition module 21 is configured to collect in real time the current signal, vibration acceleration signal, ambient temperature and humidity signal of the hoist motor, and the weight signal of the weighing body, and form a data set after preprocessing.

[0098] The model construction module 22 is configured to construct a prediction model based on multi-modal adversarial learning. The prediction model includes a physical constraint flow and a process knowledge flow, and realizes feature extraction independent of working conditions through a gradient reversal layer, and trains the prediction model through the data set.

[0099] The prediction module 23 is configured to compare in real time the weight predicted by the prediction model with the weight signal of the weighing body. When the deviation exceeds a first preset value continuously for multiple times and the confidence level is greater than a second preset value, it is determined that the weighing body is offset.

[0100] The calibration module 24 is configured to automatically calibrate the zero point or range parameters of the weighing body according to the offset type through a progressive parameter adjustment strategy, and record a calibration log.

[0101] The device provided in this embodiment, the dataset acquisition module 21 forms a dataset after preprocessing by collecting the current signal, vibration acceleration signal, ambient temperature and humidity signal of the hoist motor and the weight signal of the weighing body in real time, realizing the fusion of multi-dimensional data, effectively eliminating sensor noise and data asynchronization problems, and improving the reliability and consistency of the dataset; the model construction module 22 constructs a prediction model based on multi-modal adversarial learning, the prediction model includes a physical constraint flow and a process knowledge flow, and realizes feature extraction independent of working conditions through a gradient reversal layer, and trains the prediction model through the dataset, enabling the model to capture the influence of both equipment status and process conditions simultaneously, enhancing the generalization ability of the model in different environments; the prediction module 23 compares the weight predicted by the prediction model with the weight signal of the weighing body in real time. When the deviation exceeds the first preset value continuously for multiple times and the confidence level is greater than the second preset value, it is determined that the weighing body is offset. The dual determination conditions avoid misjudgment and missed judgment, can timely detect the problem of weighing body offset, avoid process fluctuations caused by measurement errors, can be dynamically adjusted according to the actual working conditions, enhancing the flexibility and adaptability of the system; the calibration module 24 automatically calibrates the zero point or range parameters of the weighing body according to the type of weighing body offset through a progressive parameter adjustment strategy, and records the calibration log. The progressive adjustment strategy avoids over-adjustment or under-adjustment problems, and calibration can be completed without stopping the machine, significantly improving production efficiency and equipment utilization rate. Recording information such as the time, offset type, and adjustment amount of each calibration provides data support for equipment maintenance and process optimization, can effectively solve the problem of weighing body offset in cement plants, and has a wide application prospect.

[0102] In one embodiment, the dataset acquisition module 21 includes a current signal acquisition unit, a vibration acceleration signal acquisition unit, an ambient temperature and humidity signal acquisition unit, and a dataset formation unit.

[0103] The current signal acquisition unit is used to collect the current signal through a Hall sensor, perform wavelet packet decomposition on the current signal, and extract the ripple energy entropy.

[0104] The vibration acceleration signal acquisition unit is used to collect the vibration acceleration signal through a MEMS sensor, perform spectrum analysis on the vibration acceleration signal, and extract the main frequency amplitude ratio.

[0105] The ambient temperature and humidity signal acquisition unit is used to collect the ambient temperature and humidity signal through an integrated sensor.

[0106] The dataset formation unit is used to form a dataset from the preprocessed data.

[0107] The device provided in this embodiment obtains the operating state information of the device or system from different perspectives, provides rich basis for comprehensive analysis, extracts the main frequency amplitude ratio, highlights the key features reflecting the operating state in the signal, helps to accurately identify potential problems, improves the data quality and usability, and provides a solid foundation for subsequent state monitoring, fault diagnosis, etc. using methods such as machine learning and data analysis.

[0108] In one embodiment, the model construction module 22 includes a high-dimensional feature extraction unit, a process extraction unit, and an irrelevant feature extraction unit.

[0109] The high-dimensional feature extraction unit is configured to input the ripple energy entropy and the main frequency amplitude ratio into the physical constraint flow, and extract high-dimensional features through a residual convolutional network.

[0110] The process extraction unit is configured to input the humidity compensation coefficient and the material flow set value into the process knowledge flow, and obtain process features through a fully connected network.

[0111] The irrelevant feature extraction unit is configured to implement feature extraction independent of the working conditions through the gradient reversal layer.

[0112] The device provided in this embodiment reduces the mean absolute error between the predicted weight and the actual weight through the synergistic effect of the physical constraint flow and the process knowledge flow. The model can maintain stable performance under different working conditions, reduces the calibration frequency caused by environmental changes, and avoids process fluctuations caused by measurement errors.

[0113] The embodiment of the present invention is a device embodiment corresponding to the above method embodiment. The specific operations of each module processing step can be understood with reference to the description of the method embodiment, and will not be elaborated here.

[0114] As Figure 3 shown, the present invention also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it implements the method for calibrating the scale body offset based on the hoist current in the above embodiment, or when the computer program is executed by a processor, it implements the method for calibrating the scale body offset based on the hoist current in the above embodiment.

[0115] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, storage, database, or other medium used in the embodiments provided in the present application can include non-volatile and / or volatile memories. Non-volatile memories can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memories can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and Rambus dynamic RAM (RDRAM), etc.

[0116] Each embodiment in this specification is described in a progressive manner. For the same or similar parts among the embodiments, reference can be made to each other. The key point of each embodiment is to illustrate the differences from other embodiments. In particular, for the device or system embodiments, since they are basically similar to the method embodiments, the description is relatively simple. For the relevant parts, reference can be made to the description of the method embodiments. The device and system embodiments described above are only illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they can be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art can understand and implement it without creative efforts.

[0117] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some or all of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the various embodiments of the present invention, and the content not described in detail in the specification of the present invention belongs to the well-known technology of those skilled in the art.

Claims

1. A method for calibrating a scale offset based on a hoist current, characterized in that: The following steps are involved: Collect the current signal of the hoist motor, vibration acceleration signal, ambient temperature and humidity signal, and weight signal of the scale in real time, and form a data set after preprocessing; Constructing a prediction model based on multimodal adversarial learning, the prediction model includes a physical constraint flow and a process knowledge flow, and realizes feature extraction independent of working conditions through a gradient reversal layer, and training the prediction model through the data set; The weight predicted by the prediction model is compared with the weight signal of the scale in real time, and when the deviation exceeds a first preset value for multiple consecutive times and the confidence level is greater than a second preset value, it is determined that the scale is offset; According to the scale offset type, the scale zero point or span parameters are automatically calibrated through a progressive parameter adjustment strategy, and the calibration log is recorded.

2. The method for calibrating the offset of a scale body based on the hoist current according to claim 1, characterized in that: The real-time collection of the current signal of the hoist motor, the vibration acceleration signal, the ambient temperature and humidity signal and the weight signal of the scale body, and the pre-processing to form a data set specifically includes the following steps: The current signal is collected by a Hall sensor, and the current signal is decomposed by wavelet packets to extract ripple energy entropy; Collect vibration acceleration signals through MEMS sensors, perform spectrum analysis on the vibration acceleration signals, and extract the main frequency amplitude ratio; Collect ambient temperature and humidity signals through integrated sensors; The preprocessed data is formed into a dataset.

3. The method for calibrating the offset of a scale body based on the hoist current according to claim 2, characterized in that: The construction of a prediction model based on multimodal adversarial learning includes a physical constraint flow and a process knowledge flow, and realizes feature extraction independent of working conditions through a gradient reversal layer. Specifically, the following steps are included: Inputting the ripple energy entropy and the main frequency amplitude ratio into the physical constraint flow, and extracting high-dimensional features through a residual convolutional network; Input the humidity compensation coefficient and the material flow setting value into the process knowledge flow, and obtain the process characteristics through a fully connected network; The gradient reversal layer enables condition-independent feature extraction.

4. The method for calibrating the offset of a scale body based on the hoist current according to claim 1, characterized in that: The confidence is calculated by the Monte Carlo Dropout strategy, which specifically includes the following steps: During the inference process of the prediction model, some neurons are randomly discarded and the inference is repeated 100 times; Calculate the mean and standard deviation of the predicted values ​​and obtain the confidence interval as the confidence level.

5. The method for calibrating the offset of a scale body based on the hoist current according to claim 1, characterized in that: The method of automatically calibrating the zero point or range parameter of the scale body by the progressive parameter adjustment strategy specifically includes the following steps: Zero point calibration: new zero point = original zero point + (predicted value - measured value) × 0.8; Range calibration: New range coefficient = original range coefficient × (measured value / predicted value); The single adjustment shall not exceed 0.2%, and the cumulative adjustment shall not exceed 5%.

6. A device for calibrating the offset of a scale body based on the hoist current, characterized in that: It includes data set acquisition module, model building module, prediction module and calibration module; The data set acquisition module is used to collect the current signal of the hoist motor, the vibration acceleration signal, the ambient temperature and humidity signal and the weight signal of the scale body in real time, and form a data set after preprocessing; The model building module is used to build a prediction model based on multimodal adversarial learning, the prediction model includes a physical constraint flow and a process knowledge flow, and realizes feature extraction independent of working conditions through a gradient reversal layer, and the prediction model is trained through the data set; The prediction module is used to compare the weight predicted by the prediction model with the weight signal of the scale in real time, and when the deviation exceeds the first preset value for multiple consecutive times and the confidence level is greater than the second preset value, it is determined that the scale is offset; The calibration module is used to automatically calibrate the zero point or range parameters of the scale body according to the offset type through a progressive parameter adjustment strategy, and record a calibration log.

7. The device for calibrating the offset of a scale body based on the hoisting machine current as claimed in claim 6, characterized in that: The data set acquisition module includes a current signal acquisition unit, a vibration acceleration signal acquisition unit, an ambient temperature and humidity signal acquisition unit and a data set forming unit; The current signal acquisition unit is used to collect the current signal through the Hall sensor, perform wavelet packet decomposition on the current signal, and extract the ripple energy entropy; The vibration acceleration signal acquisition unit is used to acquire the vibration acceleration signal through the MEMS sensor, perform spectrum analysis on the vibration acceleration signal, and extract the main frequency amplitude ratio; The environmental temperature and humidity signal acquisition unit is used to acquire environmental temperature and humidity signals through an integrated sensor; The data set forming unit is used to form a data set from the preprocessed data.

8. The device for calibrating the offset of a scale body based on the hoisting machine current as claimed in claim 6 or 7, characterized in that: The model building module includes a high-dimensional feature extraction unit, a process extraction unit and an irrelevant feature extraction unit; The high-dimensional feature extraction unit is used to input the ripple energy entropy and the main frequency amplitude ratio into the physical constraint flow, and extract high-dimensional features through a residual convolution network; The process extraction unit is used to input the humidity compensation coefficient and the material flow setting value into the process knowledge flow, and obtain the process characteristics through a fully connected network; The irrelevant feature extraction unit is used for the gradient inversion layer to realize feature extraction irrelevant to working conditions.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the method for calibrating the offset of the scale body based on the hoist current as described in any one of claims 1 to 5 is implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method for calibrating the offset of a scale body based on the hoist current as claimed in any one of claims 1 to 5 are implemented.

Images