High-pressure sterilizer operation state fault prediction system based on digital twinning

Abstract

The invention relates to the technical field of fault prediction, in particular to a high-pressure sterilizer operation state fault prediction system based on digital twinning, which comprises a stator current vector acquisition module used for acquiring three-phase stator current signals of a vacuum pump motor of a high-pressure sterilizer, mapping the three-phase stator current signals to a synchronous rotating coordinate system, and outputting a synchronous rotating coordinate system; and calculating a d-axis component value and a q-axis component value, generating stator current Park vector component data, and extracting current module value envelope data according to the stator current Park vector component data. According to the method, the deviation amplitude and the fluctuation intensity are measured at the same time through integral operation, the transient instability phenomenon of the inductive load in the vacuum extraction or pressure maintaining stage is sensed, the electromechanical energy conversion mapping index and the power factor dynamic distortion rate are integrated for multi-dimensional state evaluation, and electrical insulation faults and mechanical resistance abnormity are distinguished. And the early warning capability and the diagnosis accuracy of hidden faults of the high-pressure sterilization equipment are improved.

Description

Technical Field

[0001] This invention relates to the field of fault prediction technology, and in particular to a fault prediction system for the operating status of autoclaves based on digital twins. Background Technology

[0002] Fault prediction utilizes sensing technology, signal processing algorithms, and data analysis models to perform real-time monitoring and in-depth analysis of the operating parameters of industrial equipment or complex systems throughout their entire lifecycle.

[0003] Existing fault prediction technologies rely on simple threshold monitoring of overall equipment operating parameters or general time-domain statistical analysis. These are ill-suited to complex electromechanical systems like autoclaves, which exhibit multi-stage and variable load characteristics. In actual operation, monitoring only the effective value fluctuations of current or voltage is insufficient. Because the load characteristics of the autoclave vary significantly across different process stages such as vacuuming, heating, pressure holding, and drying, using fixed thresholds or single-dimensional trend analysis can easily lead to false alarms or missed alarms. For example, during high-load operation of the vacuum pump, the weak current vector distortion caused by early stator coil insulation aging is often masked by normal load current fluctuations, making it difficult for the system to promptly identify insulation degradation. Therefore, improvements are needed. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of the existing technology and propose a fault prediction system for the operating status of autoclaves based on digital twins.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: a digital twin-based autoclave operation status fault prediction system includes: The stator current vector acquisition module is used to acquire the three-phase stator current signal of the vacuum pump motor of the autoclave, map the three-phase stator current signal to the synchronous rotating coordinate system, calculate the d-axis component value and q-axis component value, generate stator current Park vector component data, extract the current magnitude envelope data based on the stator current Park vector component data, and generate current vector trajectory envelope feature data. The load mapping fault analysis module is used to call the current vector trajectory envelope feature data, calculate and generate vector trajectory geometric distortion parameters, and match the vector trajectory geometric distortion parameters with the mechanical resistance mapping table to generate vacuum pump electromechanical energy conversion mapping index. The power factor distortion assessment module is used to collect the bus voltage and total current of the autoclave, calculate the cosine value of the instantaneous phase difference between voltage and current, generate a real-time power factor value sequence, call the standard operating cycle reference curve data, align the real-time power factor value sequence with the reference curve data on the time axis, and calculate and generate the dynamic distortion rate of the power factor of the inductive load. The operating status fault prediction module is used to call the vacuum pump's electro-electrical energy conversion mapping index and the inductive load's power factor dynamic distortion rate, compare them with preset alarm thresholds, generate multi-dimensional over-limit status data of electrical parameters, retrieve the fault type database based on the multi-dimensional over-limit status data of electrical parameters, output the corresponding stator coil insulation or pump body mechanical jamming status, and generate the operating status fault prediction result of the high-pressure sterilizer.

[0006] Preferably, the step of obtaining the current vector trajectory envelope feature data is as follows: Collect the three-phase stator current signal and timestamp of the vacuum pump motor of the autoclave, resample at a uniform sampling interval and verify the phase sequence, remove the zero sequence component and calculate the angle sequence of the synchronous rotating coordinate system based on the mechanical angular velocity sequence, map the three-phase stator current signal to the synchronous rotating coordinate system and calculate the d-axis component value and q-axis component value to generate stator current Park vector component data; Based on the stator current Park vector component data, the current magnitude sequence is obtained by amplitude synthesis. Abnormal segments are removed by time stamp continuity review and boundary interpolation is performed to complete the current magnitude envelope data. Based on the current magnitude envelope data, time axis alignment is completed according to the phase index in the stator current Park vector component data, and the envelope point coordinate sequence is reconstructed by interpolation according to a fixed angular resolution. The geometric center coordinates, axial contour parameters, and trajectory boundary density are statistically analyzed and a unified format identifier is output to generate current vector trajectory envelope feature data.

[0007] Preferably, the steps for obtaining the geometric distortion parameters of the vector trajectory are as follows: The current vector trajectory envelope feature data is called, the envelope point coordinates are sorted according to the sampling point index and duplicate coordinates are deleted, the mean values ​​of the envelope point coordinates in the x-axis direction and y-axis direction are calculated, and the mean values ​​are defined as the x-component and y-component of the trajectory geometric center coordinates, respectively. The standard origin coordinates are set to zero, and the trajectory geometric center coordinates and standard origin coordinates are generated. Calculate the vector trajectory center offset based on the geometric center coordinates of the trajectory and the standard origin coordinates; Based on the vector trajectory center offset, the variance matrix of the envelope point coordinates relative to the geometric center coordinates of the trajectory in the current vector trajectory envelope feature data is statistically analyzed, and the principal direction variance is extracted. The major axis value and minor axis value are calculated separately. The ratio of the major axis value to the minor axis value and the vector trajectory center offset are combined into an ordered binary tuple to generate the vector trajectory geometric distortion parameters.

[0008] Preferably, the step of obtaining the electromechanical energy conversion mapping index of the vacuum pump is as follows: Based on the geometric distortion parameters of the vector trajectory, the distortion dimension key value range and mechanical resistance output field of the mechanical resistance mapping table are read. The intervals are located item by item according to the vector trajectory center offset and the ratio of the major axis value to the minor axis value. When the interval is at the boundary, a closed interval strategy is adopted. When the interval falls between adjacent intervals, linear interpolation is performed according to the ratio of the upper and lower boundaries. When multiple candidate records appear, the minimum absolute value of the parameter difference is used as the standard. The mechanical resistance value is extracted to generate the vacuum pump electromechanical energy conversion mapping index.

[0009] Preferably, the step of obtaining the real-time power factor numerical sequence is as follows: The bus voltage and total current of the autoclave are collected, the time base is unified and the synchronization accuracy is verified, DC bias and abnormal sudden change segments are removed, the sampling waveform is analyzed to calculate the voltage phase and current phase difference of each sampling point, and the cosine value of the phase difference is taken to obtain the instantaneous power factor value, forming a real-time power factor value sequence.

[0010] Preferably, the step of obtaining the dynamic distortion rate of the inductive load power factor is as follows: Based on the real-time power factor numerical sequence, the standard operating cycle reference curve data is called, the starting point identifier of the reference curve is extracted and time-aligned, and linear interpolation is performed according to the sampling interval and cycle length to generate the real-time power factor numerical sequence and the standard operating cycle reference curve data after time axis alignment. The dynamic distortion rate of the power factor of the inductive load is calculated based on the real-time power factor numerical sequence after time axis alignment and the standard operating cycle reference curve data.

[0011] Preferably, the steps for obtaining the multidimensional out-of-limit state data of the electrical parameters are as follows: The system calls upon the vacuum pump electromechanical energy conversion mapping index and the inductive load power factor dynamic distortion rate, reads their respective preset alarm threshold ranges and unifies their dimensions, aligns the two indexes by time index, compares the index values ​​with the corresponding preset alarm thresholds for each time period, marks the over-limit identifiers and records the start and end times, duration, over-limit amplitude and over-limit direction, and summarizes them into multi-dimensional over-limit status data of electrical parameters with complete fields.

[0012] Preferably, the steps for obtaining the fault prediction results of the autoclave operating status are as follows: Based on the multidimensional over-limit status data of electrical parameters, the field definitions and index keys of the fault type database are read, and the search keys are constructed according to the over-limit identifier combination, duration interval, over-limit amplitude level and over-limit direction order. First, a full key match is performed. If no match is found, the matching is backed up according to the field descending order. When multiple candidates appear, the rule with the closest duration is used to select a unique record to obtain the stator coil insulation or pump body mechanical jamming status. Based on the insulation of the stator coil or the mechanical jamming of the pump body, the time index and equipment identifier are merged to generate a summary of the start and end times, affected parts, trigger sources and judgment criteria for the same operating cycle. The fault level and recommended handling actions are supplemented with coded identifiers, and structured text and status labels are output to generate the fault prediction results of the autoclave operating status.

[0013] Compared with the prior art, the advantages and positive effects of the present invention are as follows: In this invention, the three-phase stator current signal of the vacuum pump motor in a high-pressure sterilizer is acquired and mapped to a synchronous rotating coordinate system. The d-axis and q-axis components are decoupled to reconstruct the Park vector trajectory. Vector synthesis and Hilbert transform are used to extract the magnitude envelope data, which can filter out high-frequency noise interference and generate the current vector trajectory envelope characteristics. By calculating geometric distortion parameters such as the trajectory geometric center coordinate offset and the ratio of major and minor axes, weak electromagnetic asymmetry characteristics caused by inter-turn short circuits or air gap eccentricity of the stator coils are captured. The geometric distortion parameters of the vector trajectory are matched with a mechanical resistance mapping table to generate electromechanical energy conversion mapping indicators. This allows for non-invasive treatment of the pump body. The quantitative assessment of mechanical jamming states combines the collection of bus voltage and total current to calculate the real-time power factor sequence. It calls the standard operating cycle reference curve and performs high-precision time axis alignment to eliminate timing errors caused by the fluctuation of the sterilization process cycle. Based on the alignment data, it calculates the dynamic distortion rate of the power factor of the inductive load. Through integral calculation, it simultaneously measures the deviation amplitude and the severity of fluctuations, and senses the transient instability of the inductive load during the vacuum extraction or pressure holding stage. It conducts multi-dimensional state assessment by integrating electromechanical energy conversion mapping indicators and power factor dynamic distortion rate, distinguishes between electrical insulation faults and abnormal mechanical resistance, and improves the early warning capability and diagnostic accuracy of hidden faults in high-pressure sterilization equipment. Attached Figure Description

[0014] Figure 1 This is a system flowchart of the present invention. Detailed Implementation

[0015] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0016] Please see Figure 1 The present invention provides a technical solution: a fault prediction system for the operating status of a high-pressure sterilizer based on digital twins, comprising: The stator current vector acquisition module is used to acquire the three-phase stator current signal of the vacuum pump motor of the autoclave, map the three-phase stator current signal to the synchronous rotating coordinate system, calculate the d-axis component value and q-axis component value, generate stator current Park vector component data, extract the current magnitude envelope data based on the stator current Park vector component data, and generate current vector trajectory envelope feature data. The load mapping fault analysis module is used to call the current vector trajectory envelope feature data, calculate and generate vector trajectory geometric distortion parameters, and match the vector trajectory geometric distortion parameters with the mechanical resistance mapping table to generate the vacuum pump electromechanical energy conversion mapping index. The power factor distortion assessment module is used to collect the bus voltage and total current of the autoclave, calculate the cosine value of the instantaneous phase difference between voltage and current, generate a real-time power factor value sequence, call the standard operating cycle reference curve data, align the real-time power factor value sequence with the reference curve data on the time axis, and calculate and generate the dynamic distortion rate of the power factor of the inductive load. The operational status fault prediction module is used to call the vacuum pump's electro-electrical energy conversion mapping index and the dynamic distortion rate of the inductive load power factor, compare them with preset alarm thresholds, generate multi-dimensional over-limit status data of electrical parameters, search the fault type database based on the multi-dimensional over-limit status data of electrical parameters, output the corresponding stator coil insulation or pump body mechanical jamming status, and generate the operational status fault prediction result of the autoclave.

[0017] The steps for obtaining the current vector trajectory envelope feature data are as follows: Collect the three-phase stator current signal and timestamp of the vacuum pump motor of the autoclave, resample at a uniform sampling interval and verify the phase sequence, remove the zero sequence component and calculate the angle sequence of the synchronous rotating coordinate system based on the mechanical angular velocity sequence, map the three-phase stator current signal to the synchronous rotating coordinate system and calculate the d-axis component value and q-axis component value to generate stator current Park vector component data; Based on the stator current Park vector component data, the current magnitude sequence is obtained by amplitude synthesis. Abnormal segments are removed by time stamp continuity review and boundary interpolation is performed to complete the current magnitude envelope data. Based on the current magnitude envelope data, the time axis is aligned according to the phase index in the stator current Park vector component data, and the envelope point coordinate sequence is reconstructed by interpolation with a fixed angular resolution. The geometric center coordinates, axial contour parameters, and trajectory boundary density are statistically analyzed and a unified format identifier is output to generate current vector trajectory envelope feature data.

[0018] Specifically, the three-phase stator current signal and timestamps of the vacuum pump motor in the autoclave were collected. Three Hall effect current sensors installed at the stator winding input were used for physical signal capture. The sampling frequency was set to 10kHz to cover the high-order harmonic components of the motor operation. The instantaneous current values ​​of phases A, B, and C were recorded and simultaneously stamped with nanosecond-level precision UNIX timestamps. To address noise interference during the acquisition process, a Butterworth low-pass filter with a cutoff frequency five times the motor's fundamental frequency was used for preprocessing. Pre-stored motor rated parameters and pole pair configurations were read, and the timing of the three-phase current zero-crossing points was checked. If phase B's zero-crossing point lags behind phase A's zero-crossing point by approximately 120 electrical degrees, and phase C lags behind phase B by approximately 120 electrical degrees, then the phase sequence is considered correct. If not, logical correction is performed by exchanging data channel indexes. Subsequently, the Clarke transform principle is used to convert the current in the three-phase stationary coordinate system into the current in the two-phase stationary coordinate system. During this process, the influence of unbalanced current is eliminated by setting the zero-sequence component coefficient to zero. Simultaneously, the mechanical angular velocity sequence synchronously acquired with the current signal is read, and the trapezoidal integral method is used to integrate the angular velocity. The initial position angle is set to zero, and the angle increment for each sampling period is accumulated. The calculation formula is as follows: ,in, For the current moment electrical angle, For the previous moment electrical angle, Let be the electric angular velocity at the current moment. The sampling period is used to obtain a synchronous rotating coordinate system angle sequence corresponding to each current sampling point. This angle sequence is then used to construct a rotation transformation matrix, projecting the current components from the stationary coordinate system to the synchronous rotating coordinate system. The d-axis and q-axis component values ​​are then calculated using the following formula: ,in, The d-axis current component represents the excitation component. The q-axis current component represents the torque component. The electrical angle at the current sampling time. These are the instantaneous values ​​of the stator current for phases A, B, and C, respectively. The constant power conversion coefficients are used to calculate the data conversion for the entire time period by substituting the values ​​point by point, thereby generating the stator current Park vector component data.

[0019] Based on the stator current Park vector component data, the d-axis and q-axis component values ​​are read. The Pythagorean theorem is used to calculate the composite vector length, i.e., the current magnitude, at each sampling point, resulting in the original current magnitude sequence. Subsequently, the integrity of the data timestamps is checked, and the difference between two adjacent timestamps is calculated. To define the criteria for abnormal interruptions, a period of historical timestamp data with 1000 sampling points was selected when the equipment was in a stable operating state, and the average sampling interval of this historical data was calculated. and sampling interval standard deviation Set a time threshold for continuity determination In this formula The average of the baseline sampling intervals, The standard deviation of sampling jitter is 4, and the confidence coefficient is 4. If the difference between the current adjacent timestamps... If an abnormal interruption is detected at a given location, five normal modulus data points before and after the interruption are extracted as reference nodes. A smooth transition curve is constructed using a cubic spline interpolation algorithm. The modulus estimation results corresponding to the missing moments during the interruption are calculated and filled into the sequence gaps to maintain the continuity of the data in the time domain. For the beginning and end boundaries of the data sequence, if there are incomplete periods, mirror extension is performed based on the waveform characteristics of adjacent complete periods to ensure that the boundary effect is minimized in subsequent analysis. After the above cleaning and completion operations, all the obtained modulus data at all times are summarized and organized to form the current modulus envelope data.

[0020] Based on the current magnitude envelope data, the phase angle information of the stator current Park vector component is extracted. The current magnitude sequence in the time domain is mapped to the angle domain, with an angle resolution parameter set to 0.5 degrees. This parameter is preset based on the motor control accuracy requirements and the subtlety of fault characteristics. Within a complete electrical angle cycle from 0 degrees to 360 degrees, a standard angle sequence with equal intervals is generated according to the set resolution. For each standard angle, two adjacent phase points are found in the original data, and the current magnitude corresponding to that standard angle is calculated using linear interpolation. This reconstructs a uniformly distributed envelope point coordinate sequence. The magnitude and angle in the polar coordinate system are converted to x and y coordinates in the rectangular coordinate system, and the geometric center coordinates of all envelope points are calculated. Then, a covariance matrix describing the distribution of trajectory shape is constructed, calculated using the following formula: ,in, It is a 2x2 covariance matrix. The total number of envelope points. For the first The coordinate vector of the envelope points , Geometric center coordinate vector , This represents the matrix transpose, obtained by solving for the eigenvalues ​​of the covariance matrix. and (For example ), calculate the length of the major axis and minor axis length Using this as the axial contour parameter, the envelope region is divided into several tiny grid units. The number of original sampling points falling into each grid unit is counted, and the ratio of the number of points to the grid area is calculated as the trajectory boundary density. The center coordinates, major and minor axis parameters, and density data obtained above are encapsulated according to a predefined binary structure to generate current vector trajectory envelope feature data.

[0021] The steps for obtaining the geometric distortion parameters of a vector trajectory are as follows: Call the current vector trajectory envelope feature data, organize the envelope point coordinates according to the sampling point index and delete duplicate coordinates, calculate the mean values ​​of the envelope point coordinates in the x-axis and y-axis directions, define the mean values ​​as the x-component and y-component of the trajectory geometric center coordinates respectively, set the standard origin coordinates as zero, and generate the trajectory geometric center coordinates and standard origin coordinates; Based on the coordinates of the geometric center of the trajectory and the coordinates of the standard origin, the offset of the vector trajectory center is calculated using the following formula: ; in, This represents the offset of the vector trajectory center. Let be the x-component of the coordinates of the geometric center of the trajectory, and be the mean x-coordinate of all envelope points. Let be the y-component of the geometric center coordinates of the trajectory, and be the mean y-coordinate of all envelope points. To prevent extremely small positive numbers with a denominator of zero; Based on the vector trajectory center offset, the variance matrix of the envelope point coordinates relative to the geometric center coordinates of the trajectory in the current vector trajectory envelope feature data is statistically analyzed, and the principal direction variance is extracted. The major axis value and minor axis value are calculated separately. The ratio of the major axis value to the minor axis value and the vector trajectory center offset are combined into an ordered binary tuple to generate the vector trajectory geometric distortion parameters.

[0022] Specifically, the process involves calling the current vector trajectory envelope feature data, reading the d-axis and q-axis current coordinate values ​​in the sampling point index order, constructing an original coordinate point set, setting a minimum resolution threshold for coordinate deduplication (e.g., 0.001A), traversing the coordinate point set to calculate the Euclidean distance between adjacent points, and discarding any point with a distance less than the threshold to retain unique coordinates and eliminate sampling redundancy. For the cleaned envelope point set, the x-axis and y-axis coordinate values ​​of all points are extracted, and an accumulation and summation algorithm is used to calculate the total values ​​in the x-axis and y-axis directions, obtaining the total number of valid points, and then setting the x-axis direction... The total value is divided by the total number of valid points to obtain the average value in the x-axis direction. The total value in the y-axis direction is divided by the total number of valid points to obtain the average value in the y-axis direction. These two average values ​​represent the DC component offset of the current vector in the d-axis and q-axis, respectively. They are defined as the x-component and y-component of the trajectory geometric center coordinates. At the same time, a standard origin object is initialized in the system memory, and its x and y coordinates are explicitly set to 0.0, representing the motor running center without bias in the ideal state. The calculated actual trajectory geometric center coordinates are encapsulated with the standard origin coordinates to generate the trajectory geometric center coordinates and the standard origin coordinates.

[0023] In the formula for calculating the center offset of a vector trajectory, the numerator aggregates the energy deviation between the x and y axes using the square term of the L2 norm, with the dimension being the square of the current. The denominator linearly aggregates the current amplitude using the L1 norm, with the dimension being current. Dividing the two gives the final offset. It retains the first-order dimension of current (Ampere), which can directly and physically reflect the combined distance of the current vector from the origin, while the regularization term in the denominator avoids numerical divergence near zero current.

[0024] The steps for obtaining the parameter are as follows: This parameter represents the x-component of the trajectory geometric center coordinates, i.e., the average DC bias current in the d-axis direction. Its acquisition requires extracting the x-coordinate data of all envelope points from the deduplicated envelope point set prepared in the previous steps. Using monitoring data from the vacuum pump motor of the actual autoclave under steady-state operation, 3600 sampling points are collected within a complete cycle. The d-axis current values ​​are summed to obtain a total value of 1800A. This value is then divided by the number of sampling points (3600). The calculation formula is as follows: ,get .

[0025] The steps for obtaining the parameter are as follows: This parameter represents the y-component of the trajectory geometric center coordinates, i.e., the average DC bias value of the current in the q-axis direction. Similarly, extract the y-coordinate data of all envelope points. Based on the monitoring data of the same operating cycle, sum the q-axis current values ​​to obtain a total value of -720A. Divide this by the number of sampling points (3600), and the calculation formula is as follows: ,get .

[0026] The steps for obtaining the parameter are as follows: To prevent extremely small positive numbers with a denominator of zero, and to ensure the consistency of the physical dimensions of the denominator, this parameter is assigned the unit of current (ampere). Its value is set based on the least significant bit (LSB) or the lower limit of the calculation precision of the current sampling system. The floating-point arithmetic specifications of the control system's microprocessor are consulted, and a safe lower limit value for single-precision floating-point precision is selected and set as follows. ,Right now .

[0027] Calculations based on parameters: Substitute the obtained parameter values ​​into the formula for calculation: 1. Calculate the numerator (dimensions: 1). ): ; ; molecular ; 2. Calculate the denominator (dimensions: 1) ): ; ; denominator ; 3. Calculate the final result (dimensions: ). ): ; The result indicates that there is a geometric center offset of approximately 0.414A in the current vector trajectory of the motor. This value quantifies the overall degree of three-phase imbalance in the stator winding or zero-point drift in the drive circuit, serving as a key quantitative indicator for subsequent judgment of whether an electrical fault exists.

[0028] Based on the offset of the vector trajectory center, the coordinates of all envelope points in the current vector trajectory envelope feature data are read. A 2x2 covariance matrix is ​​constructed using statistical methods. The calculation formula involves the mean of the product of the differences between each point coordinate and the geometric center of the trajectory. Eigenvalue decomposition is performed on this covariance matrix to obtain two eigenvalues. The square root of the larger eigenvalue is multiplied by 2 to obtain the major axis value, and the square root of the smaller eigenvalue is multiplied by 2 to obtain the minor axis value. These two values ​​represent the major and minor axis diameters of the fitted ellipse, respectively. For example, if the calculated major axis value is 4.0A and the minor axis value is 3.8A, then the ratio of the major axis value to the minor axis value is calculated. This ratio reflects the circularity distortion of the trajectory. Finally, the ratio value of 1.05 and the previously calculated vector trajectory center offset of 0.414285A are stored in a data pair in sequence. This data pair can simultaneously characterize the shape deformation and positional offset of the current trajectory, and serve as a joint feature vector for fault diagnosis to generate vector trajectory geometric distortion parameters.

[0029] The steps for obtaining the electromechanical energy conversion mapping index of a vacuum pump are as follows: Based on the geometric distortion parameters of the vector trajectory, the distortion dimension key value range and mechanical resistance output field of the mechanical resistance mapping table are read. The intervals are located item by item according to the vector trajectory center offset and the ratio of major axis value to minor axis value. When it is at the boundary, a closed interval strategy is adopted. When it falls between adjacent intervals, linear interpolation is performed according to the ratio of upper and lower boundaries. When multiple candidate records appear, the minimum absolute value of parameter difference is used as the standard. The mechanical resistance value is extracted to generate the vacuum pump electromechanical energy conversion mapping index.

[0030] Specifically, based on the geometric distortion parameters of the vector trajectory, a mechanical resistance mapping table pre-stored in non-volatile memory is loaded. This mapping table was constructed by testing the same type of vacuum pump motor under different levels of braking torque in a laboratory environment, covering the entire operating range from no-load to stall. The table defines a two-dimensional index plane with the vector trajectory center offset as the X-axis and the ratio of the major axis value to the minor axis value as the Y-axis, and the corresponding mechanical resistance value as the Z-axis output. The key value range definition of this mapping table is read, for example, the center offset range is 0A to 2.0A with a step size of 0.1A, and the major-minor axis ratio range is 1.0 to 2.0 with a step size of 0.1. The input real-time vector trajectory geometric distortion parameters are compared with the key value nodes in the table. If the input center offset is 0.414A and the major-minor axis ratio is 1.05, then the interval between the center offset of 0.4A and 0.5A and the major-minor axis ratio of 1 are first located. The input value falls within the range of .0 and 1.1. If it does not directly hit a node, bilinear interpolation logic is initiated. First, in the dimension where the ratio of major to minor axis is fixed at 1.0, the resistance interpolation value corresponding to 0.414A between 0.4A and 0.5A is calculated. Then, the same calculation is performed in the dimension where the ratio of major to minor axis is fixed at 1.1. Finally, a weighted average is taken between the interpolation results of the two dimensions, combined with the real-time ratio of 1.05. If the input parameter exceeds the boundary range defined by the mapping table, such as an offset greater than 2.0A, the maximum mechanical resistance data corresponding to the boundary value is directly locked to prevent extrapolation error. During the multidimensional index matching process, if the input point falls into the fuzzy zone of multiple adjacent grids due to data jitter, the Euclidean distance between the input point and the center point of each candidate grid is calculated, and the grid parameter with the closest distance is selected as the benchmark. Finally, the uniquely determined quantized resistance value is parsed from the mapping table to generate the vacuum pump electromechanical energy conversion mapping index.

[0031] The steps for obtaining the real-time power factor numerical sequence are as follows: The bus voltage and total current of the autoclave are collected, the time base is unified and the synchronization accuracy is verified, DC bias and abnormal sudden change segments are removed, the sampling waveform is analyzed to calculate the voltage phase and current phase difference of each sampling point, and the cosine value of the phase difference is taken to obtain the instantaneous power factor value, forming a real-time power factor value sequence.

[0032] Specifically, the bus voltage and total current of the autoclave are collected. High-precision voltage transformers and Hall effect current sensors are connected to the main power supply line of the equipment. The sampling frequency of the analog-to-digital converter is set to 10kHz to ensure the capture of subtle waveform changes. The raw waveform data streams of voltage and current are read. Clock synchronization checks are performed on the data from the two channels to ensure that the time deviation corresponding to the same sampling index is less than 1 microsecond. The collected waveform sequence is preprocessed, and the average value within a sliding window of 0.5 seconds is calculated. This average value is subtracted from the raw waveform to eliminate the DC bias component caused by sensor zero-point drift. A sudden change detection threshold is set, for example, 1.2 times the peak value of the rated voltage. The data sequence is iterated, and spikes exceeding this threshold are detected. Peak pulses are considered as grid interference and are either eliminated or replaced with smoothed adjacent points. Subsequently, the cleaned voltage and current waveforms are analyzed, and the instantaneous analytical signals of the voltage and current signals are extracted using the Hilbert transform algorithm. The instantaneous phase angle of the voltage analytical signal and the instantaneous phase angle of the current analytical signal at each sampling point are calculated, and the two are subtracted to obtain the instantaneous phase difference sequence. To address the potential 2π jump problem in the phase difference, unwrapping processing is performed to ensure the continuity of the phase difference. Finally, the cosine value of each value in the phase difference sequence is calculated to obtain the power factor sequence reflecting the real-time load characteristics. Since the vacuum pump load fluctuates rapidly, a 50-point moving average filter is applied to the sequence to smooth out glitches, forming a real-time power factor numerical sequence.

[0033] The steps for obtaining the dynamic distortion rate of the power factor of an inductive load are as follows: Based on the real-time power factor numerical sequence, the standard operating cycle reference curve data is called, the starting point identifier of the reference curve is extracted and time-aligned, and linear interpolation is performed according to the sampling interval and cycle length to generate the real-time power factor numerical sequence and the standard operating cycle reference curve data after time axis alignment. Based on the real-time power factor numerical sequence aligned with the time axis and the baseline curve data of the standard operating cycle, the dynamic distortion rate of the inductive load power factor is calculated using the following formula: ; in, The dynamic distortion rate of the power factor for inductive loads. The complete operating cycle duration is defined as the total time length of the power factor sequence after time axis alignment. For time variables, corresponding to the power factor sampling time, For the real-time power factor numerical sequence at time 10:00 The power factor value, The standard operating cycle baseline curve data at time The power factor value, It is the characteristic time constant, used to balance the dimensions of the squared term of the deviation and the squared term of the rate of change of deviation.

[0034] Specifically, based on the real-time power factor numerical sequence, the standard operating cycle baseline curve data of this model of autoclave under standard full-load operating conditions is retrieved from the database. This baseline curve records the ideal power factor change trajectory throughout the entire process from startup, vacuuming, sterilization holding to drying completion. The moment when the power factor jumps from a static 0.0 to above 0.3 in the real-time sequence is identified as the equipment startup moment. Simultaneously, the corresponding startup characteristic point in the baseline curve is searched, and the deviation between the two on the time axis is calculated. By shifting the timestamps of the real-time sequence, the startup moments of the two are made to coincide, achieving a rough alignment. Considering that the actual operating cycle may differ from the baseline due to different load weights, the process is further refined. Since there are slight differences in the quasi-period, the ratio of the actual vacuuming phase duration to the corresponding phase duration in the reference curve is calculated. This ratio is then used to perform time-axis scaling transformation on the real-time sequence, i.e., resampling. For example, if the actual duration is 1.1 times the reference duration, the real-time data is downsampled at a ratio of 1.1:1 or the reference data is interpolated and encrypted to ensure that the number of data points in the critical phases remains consistent. Non-overlapping areas at the end of the data are truncated or padded with zeros to ensure that the two curves correspond strictly in the time dimension, generating a time-axis aligned real-time power factor numerical sequence and standard operating cycle reference curve data.

[0035] In the formula for calculating the dynamic distortion rate of the power factor of inductive loads, a comprehensive distortion measurement model integrating amplitude deviation and rate of change deviation is constructed. The first term within the integral sign... Used to capture the degree to which the real-time power factor deviates numerically from a reference, reflecting static anomalies in load resistance, the second term The first derivative of the deviation is introduced to capture anomalies in the rate of power factor fluctuations, i.e., deviations from the dynamic trend. This is particularly effective in identifying minute jitters in the early stages of mechanical jamming. Both parameters are expressed through a characteristic time constant. Weighted fusion was performed to achieve full-band coverage of fault characteristics.

[0036] The steps for obtaining the parameter are as follows: This parameter represents the complete operating cycle duration, defined as the total time length of the power factor sequence after time axis alignment. Its acquisition requires the data sequence after alignment completed in the previous steps. The difference between the maximum and minimum timestamp values ​​of this sequence is read. For example, a complete standard sterilization cycle of an autoclave includes pre-vacuuming, heating, sterilization, exhaust, and drying, with a total duration typically of 3600 seconds. Therefore, the parameter is set... .

[0037] The steps for obtaining the parameter are as follows: This parameter is a time integration variable, corresponding to the power factor sampling time. In the discretization calculation, it is represented as the time index of the sampling point, and its step size is determined by the sampling frequency, such as the sampling interval. ,but Starting from 0, incrementing by 0.1s until... .

[0038] The steps for obtaining the parameter are as follows: This parameter is a real-time power factor numerical sequence at time [time value missing]. The power factor value, ranging from 0 to 1, is directly retrieved from the real-time power factor sequence calculated in the preceding steps, for example, at a certain moment during the vacuuming phase. The real-time power factor obtained from the data collection and calculation is 0.75.

[0039] The steps for obtaining the parameter are as follows: This parameter is the standard operating cycle baseline curve data at time [time value missing]. The power factor value is also between 0 and 1, obtained by consulting the standard curve database set by the equipment manufacturer, corresponding to the aforementioned time. The ideal power factor recorded by the standard curve is 0.80.

[0040] The steps for obtaining the parameter are as follows: This parameter is the characteristic time constant, used to balance the square term of the deviation (dimensionless) and the square term of the rate of change of deviation (dimensionless). The dimensions of the power factor deviation are determined to allow for physical addition and to adjust the sensitivity to dynamic changes. The setting is based on the system's response time constant to sudden load changes. By analyzing historical data from multiple faulty devices, the average energy level of the power factor deviation change rate is calculated, ensuring it is on the same order of magnitude as the average energy level of the deviation amplitude. It is 10 seconds, that is .

[0041] Calculations based on parameters: To demonstrate the calculation process, a small integration interval is selected. Estimate the single-point contribution within, for example, in time: 1. Calculate the amplitude deviation term: Real-time value ; benchmark value ; deviation ; Squared terms ; 2. Calculate the rate of change deviation term: For example, the previous moment The deviation is -0.048, then the change in deviation is ; Deviation change rate ; square of rate of change ; Weighted Items ; 3. Compose the integrand values: ; 4. For example, the entire cycle The inner mean integrand value remains at the 0.0425 level: Points ; Divide by period ; square root ; The results indicate that the dynamic distortion rate of the inductive load power factor during the current operating cycle is approximately 0.206, which is significantly higher than the baseline noise level during normal operation (typically less than 0.05). This suggests that the vacuum pump not only experiences a low power factor during operation (contributed by the amplitude term) but also suffers from severe dynamic fluctuations (contributed by the rate of change term), indicating a potential risk of jamming due to increased mechanical clearance or lubrication failure within the vacuum pump.

[0042] The steps for obtaining multidimensional out-of-limit state data of electrical parameters are as follows: The system calls upon the vacuum pump's electromechanical energy conversion mapping index and the inductive load's power factor dynamic distortion rate, reads their respective preset alarm threshold ranges and unifies their dimensions, aligns the two indices by time index, compares the index values ​​with the corresponding preset alarm thresholds for each time period, marks the over-limit indicators and records the start and end times, duration, over-limit magnitude and over-limit direction, and summarizes them into multi-dimensional over-limit status data of electrical parameters with complete fields.

[0043] Specifically, the system calls upon the vacuum pump's electromechanical energy conversion mapping index and the inductive load's power factor dynamic distortion rate. It loads the corresponding threshold configuration files from a pre-set equipment safety parameter library. These configuration files define the normal operating ranges set by the equipment manufacturer based on ISO safety standards and motor durability test data. For example, the alarm upper limit for the electromechanical energy conversion mapping index (representing mechanical resistance) is set to 20 N·m, and the alarm upper limit for the inductive load's power factor dynamic distortion rate is set to 0.15. To eliminate the difficulty of comparison caused by dimensional differences, the two indices and their corresponding thresholds are normalized to their maximum values. The real-time index values ​​are divided by their respective theoretical maximum possible values ​​(e.g., maximum torque 50 N·m, maximum distortion rate 1.0), mapping them to a dimensionless range of 0 to 1. Using a unified time index, the aligned index sequence and threshold sequence are scanned second by second to determine whether the over-limit condition is met at each time point. For example, if the normalized resistance index is 0.45 (corresponding to 22.5 N·m) at a certain time, which is greater than the normalized threshold of 0.4, it is determined that the mechanical resistance is over-limit. The time stamps of the start and end of the over-limit are recorded, and the difference between the two is calculated to obtain the duration. The difference between the over-limit value and the threshold is calculated as the over-limit amplitude. At the same time, the over-limit direction is marked as "positive" or "negative" (e.g., excessive resistance is positive). This information is combined into an event record. The entire operation cycle is traversed, and all identified event fragments are summarized to construct a structured list containing event ID, start and end time, type (resistance or power factor), amplitude, and direction, generating multi-dimensional over-limit state data of electrical parameters.

[0044] The steps for obtaining the fault prediction results of the autoclave operating status are as follows: Based on the multidimensional over-limit status data of electrical parameters, the field definitions and index keys of the fault type database are read. The search keys are constructed according to the combination of over-limit identifiers, duration intervals, over-limit amplitude levels, and over-limit direction order. First, a full key match is performed. If no match is found, the matching is backed up by field in descending order. When multiple candidates appear, the rule with the closest duration is used to select the unique record to obtain the stator coil insulation or pump body mechanical jamming status. Based on the insulation of the stator coil or the mechanical jamming of the pump body, the time index and equipment identifier are merged to generate a summary of the start and end times, affected parts, trigger sources and judgment criteria for the same operating cycle. The fault level and recommended handling actions are supplemented with coded identifiers, and structured text and status labels are output to generate the fault prediction results of the autoclave operating status.

[0045] Specifically, based on the multidimensional over-limit status data of electrical parameters, a fault type database is loaded. This database predefines the characteristic fingerprints of "stator coil insulation fault" and "pump body mechanical jamming fault." For example, the characteristic definition of "stator coil insulation fault" is: the power factor distortion rate continuously exceeds the limit while the mechanical resistance index is normal or fluctuates slightly, and the direction of the over-limit is negative (power factor decreases). The characteristic definition of "pump body mechanical jamming fault" is: the mechanical resistance index significantly exceeds the limit while the power factor distortion rate fluctuates positively, and the duration exceeds 5 seconds. The input over-limit status data is parsed, and the over-limit identifier combination (such as "resistance over-limit + distortion rate over-limit"), duration (such as "8 seconds"), and over-limit magnitude (such as "8 seconds") are extracted. A composite retrieval key is constructed using the "high-level" and "over-limit direction" (e.g., "forward + forward"). This key value is then used to perform a full match search with the feature fingerprints in the database. If a completely matching record is found, the corresponding fault type is directly output. If no match is found, a step-by-step dimensionality reduction strategy is adopted, prioritizing the matching of over-limit identifier combinations and over-limit directions, while relaxing the matching precision for duration and amplitude. For example, the duration error is allowed to be within ±2 seconds. If multiple possible fault candidates still exist, the Euclidean distance between the duration of the input data and the typical duration defined for each candidate fault is calculated, and the closest distance is selected as the best matching result. Finally, a unique fault diagnosis conclusion is determined, resulting in the stator coil insulation or pump body mechanical jamming state.

[0046] Based on the stator coil insulation or pump body mechanical jamming status, extract the metadata corresponding to the fault status, including the start and end times of the fault determination, and the names of the affected components (e.g., "vacuum pump motor stator" or "pump head bearing"). Backtrack the fault determination logic, integrate the original indicator sources that triggered the fault determination (e.g., "abnormal electromechanical energy conversion indicators") and specific determination rule descriptions (e.g., "resistance value exceeds threshold 20% for 10 consecutive seconds"). Consult the fault classification standards in the equipment maintenance manual, assign a level code to the current fault (e.g., "L1-General Warning" or "L2-Emergency Stop"), and associate it with the corresponding suggested handling measure code (e.g., "Check lubricating oil level" or "Measure winding insulation resistance"). Encapsulate all the above information according to a predefined JSON format to generate a structured text object containing a timestamp, equipment ID, fault name, level, location, cause summary, and suggested action. Simultaneously, generate a short status label for display on the human-machine interface (e.g., "Err-01: Mechanical Jamming"), and generate the autoclave operating status fault prediction result.

Claims

1. A fault prediction system for the operating status of a high-pressure sterilizer based on digital twins, characterized in that, The system includes: The stator current vector acquisition module is used to acquire the three-phase stator current signal of the vacuum pump motor of the autoclave, map the three-phase stator current signal to the synchronous rotating coordinate system, calculate the d-axis component value and q-axis component value, generate stator current Park vector component data, extract the current magnitude envelope data based on the stator current Park vector component data, and generate current vector trajectory envelope feature data. The load mapping fault analysis module is used to call the current vector trajectory envelope feature data, calculate and generate vector trajectory geometric distortion parameters, and match the vector trajectory geometric distortion parameters with the mechanical resistance mapping table to generate vacuum pump electromechanical energy conversion mapping index. The power factor distortion assessment module is used to collect the bus voltage and total current of the autoclave, calculate the cosine value of the instantaneous phase difference between voltage and current, generate a real-time power factor value sequence, call the standard operating cycle reference curve data, align the real-time power factor value sequence with the reference curve data on the time axis, and calculate and generate the dynamic distortion rate of the power factor of the inductive load. The operating status fault prediction module is used to call the vacuum pump's electro-electrical energy conversion mapping index and the inductive load's power factor dynamic distortion rate, compare them with preset alarm thresholds, generate multi-dimensional over-limit status data of electrical parameters, retrieve the fault type database based on the multi-dimensional over-limit status data of electrical parameters, output the corresponding stator coil insulation or pump body mechanical jamming status, and generate the operating status fault prediction result of the high-pressure sterilizer.

2. The high-pressure sterilizer operation status fault prediction system based on digital twin according to claim 1, characterized in that, The steps for obtaining the current vector trajectory envelope feature data are as follows: Collect the three-phase stator current signal and timestamp of the vacuum pump motor of the autoclave, resample at a uniform sampling interval and verify the phase sequence, remove the zero sequence component and calculate the angle sequence of the synchronous rotating coordinate system based on the mechanical angular velocity sequence, map the three-phase stator current signal to the synchronous rotating coordinate system and calculate the d-axis component value and q-axis component value to generate stator current Park vector component data; Based on the stator current Park vector component data, the current magnitude sequence is obtained by amplitude synthesis. Abnormal segments are removed by time stamp continuity review and boundary interpolation is performed to complete the current magnitude envelope data. Based on the current magnitude envelope data, time axis alignment is completed according to the phase index in the stator current Park vector component data, and the envelope point coordinate sequence is reconstructed by interpolation according to a fixed angular resolution. The geometric center coordinates, axial contour parameters, and trajectory boundary density are statistically analyzed and a unified format identifier is output to generate current vector trajectory envelope feature data.

3. The high-pressure sterilizer operation status fault prediction system based on digital twin according to claim 1, characterized in that, The steps for obtaining the geometric distortion parameters of the vector trajectory are as follows: The current vector trajectory envelope feature data is called, the envelope point coordinates are sorted according to the sampling point index and duplicate coordinates are deleted, the mean values ​​of the envelope point coordinates in the x-axis direction and y-axis direction are calculated, and the mean values ​​are defined as the x-component and y-component of the trajectory geometric center coordinates, respectively. The standard origin coordinates are set to zero, and the trajectory geometric center coordinates and standard origin coordinates are generated. Calculate the vector trajectory center offset based on the geometric center coordinates of the trajectory and the standard origin coordinates; Based on the vector trajectory center offset, the variance matrix of the envelope point coordinates relative to the geometric center coordinates of the trajectory in the current vector trajectory envelope feature data is statistically analyzed, and the principal direction variance is extracted. The major axis value and minor axis value are calculated separately. The ratio of the major axis value to the minor axis value and the vector trajectory center offset are combined into an ordered binary tuple to generate the vector trajectory geometric distortion parameters.

4. The high-pressure sterilizer operation status fault prediction system based on digital twin according to claim 1, characterized in that, The steps for obtaining the electromechanical energy conversion mapping index of the vacuum pump are as follows: Based on the geometric distortion parameters of the vector trajectory, the distortion dimension key value range and mechanical resistance output field of the mechanical resistance mapping table are read. The intervals are located item by item according to the vector trajectory center offset and the ratio of the major axis value to the minor axis value. When the interval is at the boundary, a closed interval strategy is adopted. When the interval falls between adjacent intervals, linear interpolation is performed according to the ratio of the upper and lower boundaries. When multiple candidate records appear, the minimum absolute value of the parameter difference is used as the standard. The mechanical resistance value is extracted to generate the vacuum pump electromechanical energy conversion mapping index.

5. The high-pressure sterilizer operation status fault prediction system based on digital twin according to claim 1, characterized in that, The steps for obtaining the real-time power factor numerical sequence are as follows: The bus voltage and total current of the autoclave are collected, the time base is unified and the synchronization accuracy is verified, DC bias and abnormal sudden change segments are removed, the sampling waveform is analyzed to calculate the voltage phase and current phase difference of each sampling point, and the cosine value of the phase difference is taken to obtain the instantaneous power factor value, forming a real-time power factor value sequence.

6. The high-pressure sterilizer operation status fault prediction system based on digital twin according to claim 1, characterized in that, The steps for obtaining the dynamic distortion rate of the power factor of the inductive load are as follows: Based on the real-time power factor numerical sequence, the standard operating cycle reference curve data is called, the starting point identifier of the reference curve is extracted and time-aligned, and linear interpolation is performed according to the sampling interval and cycle length to generate the real-time power factor numerical sequence and the standard operating cycle reference curve data after time axis alignment. The dynamic distortion rate of the power factor of the inductive load is calculated based on the real-time power factor numerical sequence after time axis alignment and the standard operating cycle reference curve data.

7. The high-pressure sterilizer operation status fault prediction system based on digital twin according to claim 1, characterized in that, The steps for obtaining the multidimensional out-of-limit state data of the electrical parameters are as follows: The system calls upon the vacuum pump electromechanical energy conversion mapping index and the inductive load power factor dynamic distortion rate, reads their respective preset alarm threshold ranges and unifies their dimensions, aligns the two indexes by time index, compares the index values ​​with the corresponding preset alarm thresholds for each time period, marks the over-limit identifiers and records the start and end times, duration, over-limit amplitude and over-limit direction, and summarizes them into multi-dimensional over-limit status data of electrical parameters with complete fields.

8. The high-pressure sterilizer operation status fault prediction system based on digital twin according to claim 1, characterized in that, The steps for obtaining the fault prediction results of the autoclave operating status are as follows: Based on the multidimensional over-limit status data of electrical parameters, the field definitions and index keys of the fault type database are read, and the search keys are constructed according to the over-limit identifier combination, duration interval, over-limit amplitude level and over-limit direction order. First, a full key match is performed. If no match is found, the matching is backed up according to the field descending order. When multiple candidates appear, the rule with the closest duration is used to select a unique record to obtain the stator coil insulation or pump body mechanical jamming status. Based on the insulation of the stator coil or the mechanical jamming of the pump body, the time index and equipment identifier are merged to generate a summary of the start and end times, affected parts, trigger sources and judgment criteria for the same operating cycle. The fault level and recommended handling actions are supplemented with coded identifiers, and structured text and status labels are output to generate the fault prediction results of the autoclave operating status.

Images