A general operating state monitoring system and its application in air conditioning equipment

Abstract

The application relates to the technical field of control system and equipment state monitoring, and discloses a general operating state monitoring system and application thereof in air conditioning equipment, which comprises the following steps: synchronously collecting controller output instructions and sensor feedback sequences; the system extracts load agent signals based on the instructions to establish a dynamic baseline, and identifies power grid harmonics to purify the feedback sequences; then, the relationship characteristic spectrum generated based on the instructions and the purified feedback sequences is compared with the dynamic baseline to determine the operating state; the application clearly identifies non-instructional weak responses caused by early physical property degradation in the controlled object by taking the controller output as a dynamic baseline for evaluating feedback; meanwhile, the scheme can dynamically strip the interference caused by external load fluctuation, so that the final state determination can be more clearly attributed to the real change in the controlled object, and the applicability and accuracy of the monitoring mode under complex non-stationary working conditions are improved.

Description

Technical Field

[0001] This invention relates to a general-purpose operation status monitoring system and its application in air conditioning equipment, belonging to the field of control system and equipment status monitoring technology. Background Technology

[0002] Currently, closed-loop control systems with controllers at their core are a fundamental technology for ensuring the stable operation of production equipment. They continuously monitor the actual state of the controlled object and compare it with preset instructions, thereby dynamically adjusting the controller output to eliminate deviations. This approach has been widely used in maintaining the stability of the equipment's macroscopic operating indicators.

[0003] However, as industry demands higher standards for long-term equipment reliability and predictive maintenance, the physical components of equipment will experience gradual performance degradation during long-term service, such as mechanical wear or changes in media properties. In order to compensate for these internal disturbances and maintain the apparent stability of the final output, the controller's own adjustment behavior will inevitably become more frequent and drastic than when it is in a healthy state. This adjustment effort to maintain stability constitutes early information characterizing the degradation of the internal health of the controlled object, but under the current technological framework, this information is usually regarded as a normal part of the control process and is ignored.

[0004] To address this issue, those skilled in the art have explored related approaches, but insurmountable constraints remain. Specifically, existing technologies suffer from the following shortcomings: 1. Reliance on external dedicated sensors for monitoring, such as vibration sensors. While these can directly measure physical states, the hardware, wiring, and data acquisition systems are costly for large-scale equipment groups, lacking universality. 2. Threshold alarms based on macroscopic parameters of the controlled object, such as total current or temperature, only respond when performance degradation has progressed to a significant stage, resulting in significant and persistent anomalies in macroscopic parameters. This often misses the optimal maintenance window. Therefore, existing technologies present an irreconcilable contradiction between the economic efficiency of monitoring and the sensitivity of early warning. Thus, the technical problem this invention aims to solve is to find a universal monitoring method that does not require additional hardware costs. This method quantifies the dynamic adjustment efforts of the controller to maintain stability by interpreting existing data flows within the closed-loop control system, and uses this as a basis for judging early state changes of the controlled object. Summary of the Invention

[0005] This invention provides a general-purpose operation status monitoring system and its application in air conditioning equipment. Its main purpose is to solve the problem of how to effectively identify the early performance degradation of the controlled object by analyzing the dynamic adjustment behavior inside the closed-loop control system without increasing additional hardware costs.

[0006] To achieve the above objectives, the present invention provides a general-purpose operation status monitoring system, which includes:

[0007] A data acquisition module, specifically, synchronously acquires the controller output command sequence and sensor feedback sequence characterizing the state of the controlled object during the operation of the closed-loop control system.

[0008] A harmonic interference cancellation module is connected to the data acquisition module. Specifically, it performs spectrum analysis on the controller output command sequence to identify the harmonic interference characteristics originating from the external power grid, and generates a virtual harmonic response sequence based on a controlled object transfer function model. Then, it subtracts the virtual harmonic response sequence from the sensor feedback sequence to output a cleaned feedback sequence.

[0009] A baseline modeling module specifically extracts the low-frequency components of the controller output command sequence as a load proxy signal and establishes a functional baseline model that continuously maps the load proxy signal to a desired control response characteristic spectrum.

[0010] A state determination module, connected to a harmonic interference cancellation module and a baseline modeling module, specifically generates a real-time control response characteristic spectrum based on the controller output command sequence and the purification feedback sequence. It also obtains the expected control response characteristic spectrum corresponding to the current operating condition from the functional baseline model based on the real-time acquired load proxy signal. Then, by comparing the difference between the real-time control response characteristic spectrum and the expected control response characteristic spectrum, the operating state of the controlled object is determined.

[0011] Preferably, the state determination module specifically performs discrete Fourier transform on the controller output command sequence and the cleanup feedback sequence to obtain the command spectrum and the cleanup feedback spectrum; and calculates the real-time control response characteristic spectrum as a transfer function estimate of the cleanup feedback spectrum and the command spectrum.

[0012] Preferably, the baseline modeling module specifically comprises: during the learning phase when the controlled object is in the baseline operating state, continuously and synchronously recording the values ​​of load proxy signals acquired under different external loads and the corresponding control response characteristic spectra under the baseline state; and using a regression analysis algorithm to fit the correspondence between the recorded multiple sets of load proxy signal values ​​and the control response characteristic spectra under the baseline state to establish a functional baseline model. This functional baseline model can interpolate the corresponding expected control response characteristic spectra for load proxy signal values ​​that have not appeared in the learning phase.

[0013] Preferably, the general-purpose operation status monitoring system also includes a fault diagnosis module, which specifically: when the status determination module determines that the operation status of the controlled object has changed, calculates the difference spectrum between the real-time control response characteristic spectrum and the expected control response characteristic spectrum; and extracts a set of morphological feature vectors from the difference spectrum, the set of morphological feature vectors including the main peak frequency of the difference spectrum, harmonic structure, peak Q value and energy frequency band distribution; and matches the set of morphological feature vectors with a built-in fault knowledge base to determine a potential fault type corresponding to the change in operation status, wherein the fault knowledge base stores the mapping rules between various physical fault types and morphological feature vectors.

[0014] Preferably, the general-purpose operation status monitoring system also includes a diagnostic confidence assessment module, which is a downstream module of the fault diagnosis module. Specifically, within a time window, it caches a series of potential fault types and their corresponding morphological feature vectors continuously output by the fault diagnosis module; and based on the series of potential fault types cached within the time window, it calculates a confidence score characterizing the evolution trajectory of the current diagnostic result. ,in, ,in This represents the confidence score for the current diagnostic cycle. This is an indicator of the persistence of the same potential fault type within this time window. The stability index of the morphological feature vectors associated with this potential fault type. This serves as an indicator of the changing trend of the amplitude component in the morphological feature vector. , and Weighting coefficients set for the system initialization phase; and only if the confidence score Only when the value exceeds a certain confirmation threshold will the potential fault type be finally output as a valid diagnostic conclusion.

[0015] Preferably, the harmonic interference cancellation module specifically constructs the transfer function model of the controlled object into a second-order low-pass filter model, and the parameters of the second-order low-pass filter model are set according to the nameplate data of the controlled object.

[0016] Preferably, the data acquisition module specifically performs moving average filtering on the original controller output command sequence and sensor feedback sequence before synchronous acquisition, so as to filter out the slowly varying components caused by changes in the macroscopic operating point of the controlled object.

[0017] Preferably, the mapping rules in the fault knowledge base are a set of IF-THEN rules based on the dynamics principle of the controlled object. This set of rules points to different combinations of morphological feature vectors, which in turn point to one of the following: mass imbalance, rolling bearing defects, and gear meshing faults.

[0018] Preferably, the data acquisition module specifically acquires the controller output command sequence and sensor feedback sequence by reading-only accessing a standard industrial communication interface of the closed-loop control system controller. The state determination module specifically quantifies the difference between the real-time control response characteristic spectrum and the expected control response characteristic spectrum to generate a continuously changing system health index. Based on the rate of change of the system health index over a continuous period of time, it determines whether the controlled object is in a state of gradual performance degradation or has experienced a sudden failure.

[0019] Preferably, the closed-loop control system is a variable frequency compressor unit in an air conditioning unit; the controlled object is the compressor motor in the variable frequency compressor unit; the controller is the inverter in the variable frequency compressor unit; the controller output command sequence is the drive frequency command sequence output by the inverter to the compressor motor; and the sensor feedback sequence is the phase current feedback sequence of the compressor motor measured by the inverter.

[0020] Compared with the prior art, the beneficial effects of the present invention are:

[0021] 1. This invention establishes a direct method for measuring the dynamic response characteristics of a controlled object during closed-loop control by synchronously acquiring the controller output command sequence and sensor feedback sequence, and generating a control response characteristic spectrum based on both. Existing technologies typically focus on the absolute value of sensor feedback or its own changing trend when monitoring system operation, while ignoring the state information inherent in the dynamic adjustment process performed by the controller to maintain system stability. This invention uses the controller output command as the dynamic benchmark for evaluating the feedback sequence, enabling the identification of weak, non-command response components in the feedback sequence caused by early physical characteristic degradation within the controlled object. This approach avoids the limitations of relying on external dedicated sensors for state monitoring.

[0022] 2. Based on the generation of control response characteristic spectrum, this invention further introduces a load proxy signal based on the controller output command sequence, and dynamically adjusts the baseline characteristic spectrum used for comparison according to the proxy signal. In industrial environments where the external load on the controlled object varies randomly, it is difficult to distinguish whether the variation is due to normal fluctuations in the external load or a true degradation of the internal health status by simply analyzing the control response characteristic spectrum. However, this invention extracts the signal for high-frequency characteristic spectrum analysis and the load proxy signal for low-frequency operating condition characterization from the same controller output command, and correlates the two. This allows the system to dynamically consider the normal response caused by changes in external load from the evaluation benchmark when assessing the health status, so that the finally identified characteristic spectrum differences can be more clearly attributed to changes in the internal state of the controlled object, thus improving the applicability of the monitoring method in complex and non-stationary operating conditions.

[0023] 3. After identifying a change in operating status, this invention further performs morphological analysis on the difference between the real-time generated control response characteristic spectrum and the baseline characteristic spectrum, and comprehensively confirms this difference by combining its evolution trajectory over a continuous time period. When a difference appears in the control response characteristic spectrum, it may originate from a continuously developing physical fault or from a transient external disturbance. The spectral morphology of the two may be similar at a single time point. This invention not only analyzes the morphological characteristics of the difference spectrum, such as the main peak frequency and harmonic structure, to preliminarily identify the physical source of the potential fault, but also performs time-series analysis on the persistence and stability of the diagnostic result over time. A real physical fault has a continuous evolution process over time, while a transient artifact is isolated in time. Through this verification method that combines spatial morphology and time trajectory, the system can effectively distinguish between persistent faults and occasional disturbances, providing a higher confidence level for the final output operating status conclusion. Attached Figure Description

[0024] Fig. 1 This is a schematic diagram of the core processing flow of the monitoring method of the present invention;

[0025] Fig. 2 This is a schematic diagram illustrating the fault feature identification principle based on difference spectrum of the present invention;

[0026] Fig. 3 This is a logic diagram for classifying and determining progressive and sudden faults in this invention. Detailed Implementation

[0027] To make the objectives, technical solutions, and advantages of the present invention clearer, the technical solutions of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0028] This invention provides a general-purpose operational status monitoring system and its application in air conditioning equipment. Its overall architecture mainly includes a data acquisition module, a harmonic interference cancellation module, a baseline modeling module, and a status determination module. In some embodiments, it may also include a fault diagnosis module and a diagnostic confidence assessment module. These modules work together to achieve non-intrusive monitoring of early performance degradation of the controlled object through analysis of the data flow within the closed-loop control system. In specific deployments, this system can be applied to variable frequency compressor units in air conditioning equipment, where the inverter is the core controller and the compressor motor is the controlled object. The data acquisition module, as the system's data input, is configured to acquire basic data that characterizes the relationship between the controller's dynamic adjustment behavior and the actual response of the controlled object without disturbance. To achieve this, the data acquisition module performs read-only access to a standard industrial communication interface of the controller in the closed-loop control system, synchronously acquiring two sets of time series at a preset sampling frequency, for example, 1 kHz: the controller output command sequence... and sensor feedback sequence In the application scenario of air conditioning equipment, the controller output command sequence is specifically the drive frequency command sequence output by the frequency converter to the compressor motor, while the sensor feedback sequence is the compressor motor phase current feedback sequence measured inside the frequency converter. After the original sequence is acquired, in order to remove the DC or low-frequency components introduced by the change of the operating point of the controlled object, such as the slow change of the load over a long time scale, the data acquisition module is also configured to perform moving average filtering on the original controller output command sequence and sensor feedback sequence respectively. A specific filtering procedure is as follows: set a time window with a width of 100 sampling points, calculate the average value of the data within the window, and subtract the average value from the original sampling value corresponding to the center point of the window, thereby outputting a sequence containing only high-frequency dynamic information, providing input for subsequent feature spectrum analysis.

[0029] Given that the power grid environment in industrial sites often contains harmonic interference introduced by other nonlinear loads, these harmonics can contaminate the controller output command sequence. If left untreated, these signals will be treated as normal command responses in subsequent analysis, thus masking the characteristics of equipment degradation. Therefore, the system is equipped with a harmonic interference cancellation module to purify the sensor feedback sequence. This module first processes the acquired controller output command sequence. Spectral analysis is performed to identify spectral peaks whose frequencies are locked to integer multiples of the power frequency, such as the 5th and 7th harmonics, as characteristics of harmonic interference originating from the external power grid. Subsequently, the module generates a virtual harmonic response sequence in real time based on a controlled object transfer function model. This transfer function model is constructed as a second-order low-pass filter model, whose key parameters, such as natural frequency... With damping ratio This involves a one-time calibration setting based on the nameplate data of the controlled object, namely the compressor motor, such as rated power, speed, and moment of inertia; for example, when in A frequency of 250Hz and an amplitude of [value] were identified. When the 5th harmonic component is detected, the second-order low-pass filter model calculates the amplitude and phase of the theoretically expected current response of the command component, thus constructing a virtual harmonic response sequence. Finally, the module obtains the original sensor feedback sequence... Subtract this calculated virtual harmonic response sequence from the output to get a clean feedback sequence. In this way, interference introduced by the external power grid is suppressed at the feedback signal level, enabling subsequent state comparisons to more accurately focus on response differences caused by changes in the internal physical characteristics of the controlled object. To determine the specific failure mode of the controlled object based on the time series of the system health index, the state determination module further configures a failure mode determination procedure. This procedure has two built-in parallel analysis time windows: a short time window for identifying sudden faults, with a duration of 60 seconds, and a long time window for identifying gradual performance degradation, with a duration of 24 hours. Within the short time window, the module calculates the first-order difference of the system health index sequence. If the absolute value of this difference exceeds the sudden fault threshold at any time, for example, 50, it is determined that a sudden fault has occurred. Within the long time window, the module performs linear regression analysis on the system health index sequence. If the slope of the regression line is negative and its absolute value is greater than the gradual degradation threshold, for example, 0.1 points / hour, it is determined that the controlled object is in a state of gradual performance degradation.

[0030] To distinguish between internal state degradation and normal load response even under dynamically changing external load conditions, the system further includes a baseline modeling module. This module aims to establish a functional baseline model that dynamically maps the external load to the desired control response. Specifically, this module first outputs a sequence of commands from the controller. Extract its low-frequency components, for example, the components obtained through a low-pass filter with a cutoff frequency of 0.5Hz. This serves as a load proxy signal capable of representing the intensity of external loads borne by the controlled object in real time. Subsequently, during the learning phase when the controlled object is in its baseline operating state, the system continuously and synchronously records the load proxy signals acquired under different external loads. The module collects the numerical values ​​of the load surcharge signals and the corresponding control response characteristic spectrum generated by the state determination module at that time. After accumulating 500 sets of data points covering 20% ​​to 100% of the rated load, the module uses a multinomial regression analysis algorithm to fit the correspondence between the recorded multiple sets of load surcharge signal values ​​and the control response characteristic spectrum under the baseline state, thereby establishing a functional baseline model. This model can generate the corresponding expected control response characteristic spectrum for load surcharge signal values ​​that have not appeared in the learning phase through interpolation calculation, providing a dynamic and adaptive evaluation benchmark for state determination.

[0031] The status determination module is the core of the operational status assessment. It connects with the harmonic interference cancellation module and the baseline modeling module, receiving the purified feedback sequence and the dynamic expected response benchmark. Within a monitoring cycle, this module first determines the current controller output command sequence. With purification feedback sequence A real-time control response characteristic spectrum is generated; this process includes performing discrete Fourier transforms on the controller output command sequence and the cleanup feedback sequence respectively to obtain the command spectrum. and purification feedback spectrum Furthermore, the real-time control response characteristic spectrum is calculated as a transfer function estimate of the cleanup feedback spectrum and the command spectrum, i.e. Meanwhile, the module uses the load proxy signal acquired in real time. The desired control response feature spectrum corresponding to the current operating condition is obtained from the functional baseline model. Finally, by comparing the difference between the real-time control response characteristic spectrum and the expected control response characteristic spectrum, such as calculating the energy difference or correlation coefficient between the two in the critical frequency band, the operating state of the controlled object is determined. This difference is further quantified into a continuously changing system health index. The absolute value of the index represents the degree of deviation from the healthy state, while the rate of change of the index over a continuous period of time is used to determine whether the controlled object is in a state of gradual performance degradation or has experienced a sudden failure.

[0032] To provide deeper fault tracing capabilities, the system can also be configured with a fault diagnosis module. This module is activated when the state determination module determines that the operating state of the controlled object has changed. It first calculates the difference spectrum between the real-time control response characteristic spectrum and the desired control response characteristic spectrum. Then, it extracts a set of preset morphological feature vectors from this difference spectrum. These vectors specifically include the frequency corresponding to the strongest spectral peak in the difference spectrum (i.e., the main peak frequency), the existence of a series of harmonic structures with frequencies in integer multiples, the sharpness of the main peak (i.e., the peak Q-value), and the energy distribution ratio in different frequency bands. The system also has a built-in fault knowledge base. The library stores a set of IF-THEN rules based on the dynamics of the controlled object. These rules point to different combinations of morphological feature vectors, which in turn point to a specific physical fault type in unbalanced rolling bearing defects and gear meshing faults. For example, a rule can be defined as: IF (main peak frequency is approximately equal to motor rotation frequency) AND (spectral peak Q value is greater than 5) AND (harmonic structure is not obvious) THEN (potential fault type is determined to be unbalanced). By matching the real-time extracted morphological feature vectors with this fault knowledge base, the module can output a potential fault type corresponding to the change in the current operating state.

[0033] Considering that transient interference may lead to unreliable single diagnostic results, the system further configures a diagnostic confidence assessment module downstream of the fault diagnosis module to perform time-series verification of the validity of the diagnostic conclusions. This module caches a series of potential fault types and their corresponding morphological feature vectors continuously output by the fault diagnosis module within a settable time window, such as 300 seconds. Based on the cached data within this time window, the module calculates a confidence score characterizing the evolution trajectory of the current diagnostic result. Its calculation formula is ;in, This is a persistence indicator for the occurrence of the same potential fault type within this time window, and its value is the number of times the type occurs divided by the total number of diagnoses. The stability index of the morphological feature vector associated with this potential fault type can be specifically calculated as the reciprocal of the standard deviation of key features such as the main peak frequency within this time window. This is an indicator of the changing trend of the amplitude component in the morphological feature vector, characterized by the slope obtained through linear regression analysis of the amplitude sequence; weighting coefficients , and A constant summing to 1 is set for the system initialization phase, for example, 0.5, 0.3, and 0.2 respectively; a specific calculation example is: if 280 rolling bearing defects are diagnosed within a 300s window, then... The standard deviation of its main peak frequency is 0.2Hz. It can be normalized to 0.88; its amplitude component slope is positive, and after normalization... The final confidence score is 0.75. Only when the confidence score is... When the value exceeds a preset confirmation threshold, such as 0.8, the system will finally output the potential fault type as a valid diagnostic conclusion, thereby distinguishing between persistent physical faults and occasional interference artifacts and improving the confidence of the final output conclusion.

[0034] Example 1: In a cooling system of a large data center, hundreds of compressor motor units driven by frequency converters operate continuously. One unit's rolling bearing, due to long-term service, has developed a peeling point on its raceway surface, with a size on the order of micrometers. This defect, each time a roller passes over it, applies a high-frequency impact disturbance to the motor shaft with an amplitude lower than the detection threshold of a conventional vibration sensor. In the initial stage of the fault, this disturbance does not cause changes in the compressor motor unit's operating parameters, such as total current consumption, outlet fluid pressure, or machine body temperature, that could be detected by conventional monitoring system thresholds. This is because the frequency converter, as the core of the closed-loop control system, compensates for this internal disturbance by fine-tuning its output drive frequency and voltage, thereby maintaining the stability of the motor output speed and keeping the final cooling effect at the set value. However, this compensatory control behavior has caused dynamic changes in the frequency converter's output commands. The monitoring system of this invention, through its data acquisition module, synchronously acquires the controller output command sequence of the unit's frequency converter at a frequency of 1kHz. Sensor feedback sequence with motor phase current Because data centers contain a large number of other power electronic devices, grid harmonic interference is significant, and harmonic interference cancellation modules are crucial. After performing spectral analysis, the presence of 5th and 7th harmonics was identified. The module then generated a virtual harmonic response sequence based on a pre-defined second-order low-pass filter model for the motor, and extracted it from the original sensor feedback sequence. The value was deducted from the output, and a purification feedback sequence was output. The purification signal output in this step enables the subsequent state determination module to perform feature recognition under a high signal-to-noise ratio.

[0035] During this period, the data center's computing load fluctuated, causing the total cooling load to rise from 70% to 85% within minutes. The baseline modeling module extracted the controller's output command sequence. The low-frequency components were used to obtain the load proxy signal characterizing the load change. Accordingly, the state determination module, based on the real-time value of the load agent signal, queries and generates the expected control response characteristic spectrum corresponding to the current 85% load from the established functional baseline model. This expected spectrum reflects the response characteristics that the health device should exhibit when subjected to higher loads, thereby distinguishing the response caused by changes in external operating conditions from the response caused by changes in the internal state of the controlled object; furthermore, within the state determination module, the system bases its response on the controller output command sequence. With purification feedback sequence Calculate the real-time control response characteristic spectrum When this real-time spectrum is compared with the dynamically adjusted desired control response characteristic spectrum... During the comparison, a persistent energy concentration region, absent in the expected spectrum and present over time, appeared in a high-frequency band related to the characteristic frequency of bearing failure within the difference spectrum. The presence of this difference spectrum caused the system's health index to deviate from the normal baseline and exhibit a slow downward trend. This downward trend triggered the intervention of the fault diagnosis module and the diagnostic confidence assessment module. The fault diagnosis module performed morphological analysis on the difference spectrum, and the extracted feature vector, due to its energy concentration in the high-frequency band and broad spectral peaks, was initially identified as a potential rolling bearing defect by the built-in IF-THEN rule set. Subsequently, the diagnostic confidence assessment module performed time-series verification of this diagnostic conclusion within a 300-second time window. Because this feature exhibited high persistence over time, the verification was successful. With stability The final calculated confidence score If the confirmation threshold of 0.8 is exceeded, the system will output a high-confidence early bearing failure warning for the specific unit. Based on this warning, the maintenance personnel will inspect the unit during the next planned maintenance window, confirm the early stripping of the rolling bearing raceway, and replace it. After replacement, the system health index of the unit monitored by this system will return to the baseline level. The whole process avoids an unplanned downtime caused by bearing failure without affecting the data center cooling business.

[0036] Example 2: This example aims to verify the technical effectiveness of the monitoring system of the present invention under specific conditions. The specific conditions are that the controlled object, while experiencing external load fluctuations and power grid harmonic interference, also suffers from early-stage mechanical failures. To this end, a closed-loop control system test platform was built. This platform consists of a three-phase asynchronous motor general-purpose frequency converter and a centrifugal water pump load system that changes pipeline resistance by adjusting valves. All test data originates from this physical platform. Data acquisition is achieved by reading the frequency converter's communication port. Its functional specifications include supporting synchronous output of controller commands and motor phase current feedback at a frequency of not less than 1kHz. The acquired data... The data is fed into the monitoring system of this invention deployed in the edge computing unit for processing. The test set up two controlled objects: one was a healthy motor with a standard rolling bearing installed, and the other was a faulty motor with the bearing at the same position replaced with a scratch with a width of 0.1 mm and a depth of 0.05 mm machined on the outer raceway. During the test, the regulating valve of the load system was controlled by the program to make the motor run stably at three load points of 50%, 75% and 100% of its rated speed. At the same time, in order to simulate the industrial electromagnetic environment, a switching power supply was connected to the power supply bus of the test platform to introduce the 5th and 7th harmonics.

[0037] The experiment was divided into two phases. In the first phase, the monitoring system of this invention was run on a motor in a healthy state, and the learning process of the baseline modeling module was executed. The system ran continuously for 2 hours within a load range covering 20% ​​to 100%, establishing the load proxy signal. With the expected control response characteristic spectrum The associated functional baseline model; in the second stage, the faulty motor was replaced, and a sample group of the present invention and three control groups were set up for comparative testing. The sample group of the present invention enabled all functions, including harmonic interference cancellation and dynamic baseline. Control group A, while keeping other conditions unchanged, did not enable the harmonic interference cancellation module to isolate its effect. Control group B used a static baseline learned at a single load point of 75% to replace the dynamic baseline model to isolate the effect of the dynamic baseline. Control group C only monitored the effective value of the total motor current and the casing temperature. In the second stage of the experiment, each sample group ran stably for 10 minutes at three load points: 50%, 75%, and 100%. The system health index output by the system was normalized with the baseline modeled under healthy conditions set to 100. The current and temperature indicators monitored by control group C showed changes of less than 1.5% under all operating conditions of the faulty and healthy motors, failing to identify any state changes. Control group A, due to harmonic interference, had an output... The health index of the control group showed irregular fluctuations exceeding + / -25 under all operating conditions, making it impossible to form a stable judgment. The health index of the control group B was 87 at the 75% load point, but when the load was switched to 50% or 100%, its index dropped to 65 and 68 respectively, resulting in deviations due to baseline mismatch. In contrast, the system health index of the sample group of this invention showed consistency throughout the variable load test. At the three load points of 50%, 75%, and 100%, its index was stable at 86.5, 87.1, and 86.8 respectively, indicating that the system distinguished the impact of external load step changes on the state assessment through the load proxy signal and the functional baseline model. At the same time, compared with the health state baseline of 100, the index consistently and stably decreased by about 13 points at all load points. This stable deviation reflects the change in the dynamic response characteristics of the controlled object introduced by bearing scratches. See Table 1, which is a typical data record of the monitoring results of each sample group under the stable operating condition of 75% load point during the test.

[0038] Table 1: Comparison of monitoring results of each sample group at 75% load point.

[0039] ;

[0040] Experimental data shows that the monitoring system of this invention, through the combination of harmonic interference cancellation and dynamic baseline modeling, can identify changes in dynamic response characteristics caused by the degradation of the internal physical performance of the controlled object in an environment where external load changes and power grid harmonics coexist. In contrast, systems with missing functions or traditional monitoring methods fail to draw definite conclusions about the operating status under this non-stationary condition.

[0041] Example 3: This example combines Figs. 1 to 3 This document describes a general-purpose operational status monitoring system and its application in air conditioning equipment.Fig. 1 As shown, the data source of this system is a closed-loop control system, such as an air conditioning compressor unit. The system synchronously acquires controller output command sequences and sensor feedback sequences from this source. These two sequences first enter the data acquisition module for synchronous acquisition and filtering. The processed sequences are divided into two paths. The controller output command sequence is sent to the baseline modeling module to extract load proxy signals and establish a dynamic functional baseline, which is output in the form of the desired control response characteristic spectrum. Simultaneously, the filtered dual sequences are sent to the harmonic interference cancellation module. This module identifies and eliminates the interference of power grid harmonics on the feedback sequence based on the command sequence, outputting a purified feedback sequence. Subsequently, the state determination module receives the command sequence, the purified feedback sequence, and the dynamic baseline provided by the baseline modeling module. It determines the operating state by comparing the real-time response with the desired response. If the operating state is determined to have changed, the fault diagnosis module is triggered. This module identifies potential fault types by analyzing the difference spectrum and matching it with a fault knowledge base. Finally, the system outputs an early fault warning after high-confidence state determination.

[0042] like Fig. 2 As shown in the figure, the horizontal axis represents frequency in Hertz (Hz), and the vertical axis represents amplitude in Decibels (dB). The healthy state curve, represented by the solid line in the figure, exhibits a characteristic spectral morphology where the amplitude smoothly decreases as the frequency increases. The fault state curve, represented by the dashed line in the figure, shows a similar overall trend to the healthy state, but an abnormal spectral peak appears at a specific frequency, around 200 Hz. The difference spectral curve, represented by the dotted line in the figure, is obtained by subtracting the healthy state spectrum from the fault state spectrum. This difference spectrum clearly highlights the frequency position and energy amplitude of the abnormal spectral peak introduced by the fault, thus providing a clear basis for subsequent state determination and fault diagnosis.

[0043] like Fig. 3 As shown, the complete decision path of the system from health monitoring to fault confirmation is depicted. After system initialization, it enters the healthy operation state. When the system health index is detected to be lower than the normal baseline, the state transitions to the state deviation state. In this state, the system analyzes the rate of change of the health index through parallel time windows. If the rate of change exceeds the sudden failure threshold within a short time window, it enters the sudden failure decision branch. If the rate of change exceeds the progressive degradation threshold within a long time window, it enters the progressive performance degradation decision branch. Whether it is a sudden failure or progressive performance degradation, the corresponding diagnostic confidence score must exceed the confirmation threshold before it can finally enter the fault confirmed state. When the system is in the state deviation state, if the health index recovers to the baseline, the system returns to the healthy operation state. When the system is in the fault confirmed state, it needs to be manually intervened or reset after maintenance to return to the healthy operation state.

[0044] Example 4: This example discloses a specific process for on-site calibration and configuration of the internal model and parameters of the aforementioned monitoring system when deployed on a specific controlled object, namely an air conditioner inverter compressor unit. Before the compressor unit is put into operation for the first time, or after confirming that it is in good working order during a planned overhaul, the following offline calibration process is executed: First, the parameters of the transfer function model of the controlled object required in the harmonic interference cancellation module are determined. This model is constructed as a second-order low-pass filter model, whose parameters are the natural frequency... With damping ratio The rated power is determined from the compressor motor's nameplate data through the following steps: read the rated power from the motor nameplate. Rated speed The rotor moment of inertia was obtained from the technical manual of the compressor unit. Based on these data, the mechanical time constant of the motor was calculated. With electromagnetic time constant Furthermore, Set as and will Set as .

[0045] Secondly, the functional baseline model in the baseline modeling module is learned and constructed. After confirming that the compressor unit is in a healthy state, the host computer control system executes a preset operating cycle. This cycle includes a process of slowly and linearly climbing from 20% rated load to 100% rated load for 30 minutes, followed by a step process of stabilizing at three load points of 40%, 60%, and 80% for 10 minutes each. During this period, the monitoring system continuously records the sequence of commands output by the controller. The load surrogate signal obtained after low-pass filtering The instantaneous values ​​of the input and the control response feature spectrum generated synchronously under healthy conditions are used. After data acquisition, the baseline modeling module uses a third-order polynomial regression algorithm to fit all acquired data points, thereby obtaining a set of input values. Numerical mapping to a desired control response characteristic spectrum The polynomial coefficients constitute the functional baseline model of this compressor unit; finally, the weighting coefficients in the diagnostic confidence assessment module are... , and Configuration is performed; given that the expected failure mode of rotating machinery such as air conditioning compressors is primarily progressive wear, their failure characteristics typically exhibit high persistence and a slow growth trend over time, while the morphology of the characteristics themselves is relatively stable; to ensure that confidence assessment focuses on this evolutionary characteristic, the weight representing the persistence of diagnostic results is increased. Setting it to 0.5 will assign a weight to the characteristic of evolving trends. The weight representing the stability of the feature is set to 0.3. Set to 0.2; by executing the calibration procedure of the above three steps, the internal model and parameters of the monitoring system are adapted to the specific controlled object.

[0046] Example 5: To apply the monitoring system of the present invention to a new model of planetary gearbox compressor unit and enable its fault diagnosis module to identify specific fault types, an offline fault feature injection and knowledge base calibration test needs to be performed in advance. This test is conducted on a test platform similar to the one described above. First, a planetary gearbox compressor unit confirmed to be in good condition is installed on the platform, and its data under different operating conditions are collected to establish a complete functional baseline model. Subsequently, three sets of test components with pre-implanted single-type early faults are installed in sequence. The first set is a rotor with a known unbalanced mass block on the input shaft, the second set is an etched pit of a known size on the outer ring of the rolling bearing of one of the planetary gears, and the third set is a peeling defect of a known size on the surface of the teeth of a sun gear.

[0047] For each group of faulty components, a variable load test procedure was repeatedly executed on the test platform, and the monitoring system continuously calculated the difference spectrum between the real-time control response characteristic spectrum and the expected control response characteristic spectrum. Statistical analysis of the difference spectrum data was performed to identify the corresponding morphological feature vector with stable statistical characteristics for each known physical fault. The morphological feature vector corresponding to the mass imbalance fault had its main peak frequency component strictly equal to one harmonic of the motor rotation frequency, with insignificant harmonic components and a peak Q value greater than 5. The morphological feature vector corresponding to the tooth surface spalling fault was identified as having significant sidebands at the gear meshing frequency and its harmonic frequencies, with energy mainly distributed in the mid-to-high frequency band. These identified morphological features associated with specific physical faults were further analyzed. The feature vector combination is ultimately solidified into IF-THEN rules and loaded into the fault knowledge base. After completing the above calibration, to verify the effectiveness of the knowledge base, a compressor unit with two mixed faults—slight dynamic imbalance and early bearing wear—was installed on a test bench for testing. During operation, the fault diagnosis module extracts two separate morphological feature vectors from the difference spectrum and, by matching them with the calibrated knowledge base rules, outputs two potential fault types: suspected mass imbalance and suspected bearing fault. This offline calibration procedure provides a systematic method for constructing and verifying the internal fault knowledge base of the monitoring system when it is extended to new or differentiated controlled objects, establishing the logical judgment basis for fault diagnosis on reproducible experimental data.

[0048] Example 6: Before the monitoring system completes all offline calibrations and is about to be put into continuous online monitoring, a sensor channel integrity self-test procedure needs to be executed to confirm the status of the sensor channel. This procedure is executed under the condition that the controlled object, namely the air conditioning compressor unit, is operating stably at 50% rated load. The system collects a 60-second sensor feedback sequence. The system calculates the variance, kurtosis, and kurtosis of the time series signal as its statistical characteristics. Then, it compares the statistical characteristics calculated in real time with the reference statistical characteristics corresponding to the current sensor model under healthy working conditions that are stored in the system. If the relative deviation of any characteristic exceeds 15%, the system determines that there is an anomaly in the current sensor channel and interrupts subsequent operations and outputs a self-test alarm.

[0049] After a self-check of sensor channel integrity, confirmation thresholds are set for different potential fault types. The system then calls up the data collected and stored in the offline fault feature injection test in Example 4. For the three types of fault data—mass imbalance, rolling bearing defects, and gear meshing faults—the system backtracks and calculates a series of confidence scores output by the diagnostic confidence assessment module under the fault state. The statistical distribution, and similarly calculate the corresponding baseline data collected in a healthy state. Score distribution; Based on these two sets of data distributions, the system performs receiver operation characteristic curve analysis for each fault type, and sets the acknowledgment threshold corresponding to that fault type at a minimum level that ensures the actual alarm rate is greater than or equal to 99.5%. The numerical values; according to this procedure, the confirmation threshold for rolling bearing defects is set at 0.85, and the confirmation threshold for mass imbalance is set at 0.80; after completing the sensor self-test and threshold setting, the pre-configuration process required for the monitoring system is completed.

[0050] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0051] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A general-purpose operating condition monitoring system, characterized by, General-purpose operation status monitoring systems include: A data acquisition module, specifically, synchronously acquires the controller output command sequence and sensor feedback sequence characterizing the state of the controlled object during the operation of the closed-loop control system. A harmonic interference cancellation module is connected to the data acquisition module. Specifically, it performs spectrum analysis on the controller output command sequence to identify the harmonic interference characteristics originating from the external power grid, and generates a virtual harmonic response sequence based on a controlled object transfer function model. Then, it subtracts the virtual harmonic response sequence from the sensor feedback sequence to output a cleaned feedback sequence. A baseline modeling module specifically extracts the low-frequency components of the controller output command sequence as a load proxy signal and establishes a functional baseline model that continuously maps the load proxy signal to a desired control response characteristic spectrum. A state determination module, connected to a harmonic interference cancellation module and a baseline modeling module, specifically generates a real-time control response characteristic spectrum based on the controller output command sequence and the purification feedback sequence. It also obtains the expected control response characteristic spectrum corresponding to the current operating condition from the functional baseline model based on the real-time acquired load proxy signal. Then, by comparing the difference between the real-time control response characteristic spectrum and the expected control response characteristic spectrum, the operating state of the controlled object is determined.

2. A general-purpose operating state monitoring system according to claim 1, characterized in that, The state determination module specifically performs discrete Fourier transforms on the controller output command sequence and the cleanup feedback sequence to obtain the command spectrum and the cleanup feedback spectrum, and calculates the real-time control response characteristic spectrum as a transfer function estimate of the cleanup feedback spectrum and the command spectrum.

3. The universal operating condition monitoring system of claim 1, wherein, The baseline modeling module specifically involves: during the learning phase when the controlled object is in a baseline operating state, continuously and synchronously recording the values ​​of load proxy signals acquired under different external loads and the corresponding control response characteristic spectra under the baseline state; and using a regression analysis algorithm to fit the correspondence between the recorded multiple sets of load proxy signal values ​​and the control response characteristic spectra under the baseline state to establish a functional baseline model. This functional baseline model can interpolate the corresponding expected control response characteristic spectra for load proxy signal values ​​that have not appeared in the learning phase.

4. The universal operating condition monitoring system of claim 1, wherein, The general-purpose operation status monitoring system also includes a fault diagnosis module, which specifically: when the status determination module determines that the operation status of the controlled object has changed, it calculates the difference spectrum between the real-time control response characteristic spectrum and the expected control response characteristic spectrum; and extracts a set of morphological feature vectors from the difference spectrum, which includes the main peak frequency, harmonic structure, peak Q value, and energy frequency band distribution of the difference spectrum; and matches the set of morphological feature vectors with a built-in fault knowledge base to determine a potential fault type corresponding to the change in operation status, wherein the fault knowledge base stores the mapping rules between various physical fault types and morphological feature vectors.

5. A general-purpose operating condition monitoring system according to claim 4, characterized in that, The general operating state monitoring system further comprises a diagnosis confidence evaluation module as a downstream module of the fault diagnosis module, which is specifically: in a time window, a series of potential fault types and corresponding morphological feature vectors continuously output by the fault diagnosis module are cached; and based on the series of potential fault types cached in the time window, a confidence score representing the evolution track of the current diagnosis result is calculated wherein, wherein, is the confidence score of the current diagnosis cycle, is a persistence index of the same potential fault type appearing in the time window, is a stability index of the morphological feature vector associated with the potential fault type, is a change trend index of the amplitude component in the morphological feature vector, and is a weight coefficient set in the system initialization stage; and only when the confidence score exceeds a confirmation threshold, the potential fault type is finally output as an effective diagnosis conclusion.​ 6. The universal operating condition monitoring system of claim 1, wherein, The harmonic interference cancellation module specifically constructs the transfer function model of the controlled object into a second-order low-pass filter model. The parameters of the second-order low-pass filter model are set according to the nameplate data of the controlled object.

7. The universal operating condition monitoring system of claim 1, wherein, The data acquisition module is specifically configured to perform moving average filtering on the original controller output instruction sequence and the sensor feedback sequence respectively before synchronous acquisition.

8. The universal operating condition monitoring system of claim 4, wherein, The mapping rules in the fault knowledge base are a set of IF-THEN rules established based on the dynamics of the controlled object, and the rules point different combination modes of the morphological feature vector to one of the mass imbalance, rolling bearing defect and gear meshing fault.

9. The universal operating condition monitoring system of claim 1, wherein, The data acquisition module is specifically configured to obtain the controller output instruction sequence and the sensor feedback sequence by performing read-only access to a standard industrial communication interface of a controller of the closed-loop control system.

10. The general-purpose operating condition monitoring system according to claim 1, wherein The closed-loop control system is a variable frequency compressor unit in an air conditioning device; the controlled object is a compressor motor in the variable frequency compressor unit; the controller is a frequency converter in the variable frequency compressor unit; the controller output instruction sequence is a drive frequency instruction sequence output by the frequency converter to the compressor motor; and the sensor feedback sequence is a phase current feedback sequence of the compressor motor measured by the frequency converter.

Images