Permanent magnet propulsion motor health state on-line monitoring device and method

Abstract

The invention discloses a permanent magnet propulsion motor health state on-line monitoring device which comprises a core calculation circuit, a mass memory and a communication circuit, the mass memory and the communication circuit are connected with the core calculation circuit, and the communication circuit is connected with a controller of a permanent magnet propulsion motor for communication to obtain real-time operation data of the permanent magnet propulsion motor; the invention further discloses a monitoring method. According to the method, the operation data of the permanent magnet propulsion motor are collected to form a sample set, a fault diagnosis model is designed, normal distribution of normal operation parameters of various working conditions of the permanent magnet propulsion motor is formed through learning and training, and possible hidden dangers are identified on line based on the six-sigma principle. And reasons for possible faults are analyzed and diagnosed in the fault diagnosis model, and preventive maintenance suggestions are given.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of ship electric propulsion, and relates to a permanent magnet propulsion motor, in particular to a permanent magnet propulsion motor health state online monitoring device and a monitoring method thereof. BACKGROUND

[0002] Permanent magnet propulsion motors are widely used in the fields of ships, aerospace, electric vehicles, etc. due to their advantages of high efficiency, energy saving, large power density, etc. In the field of ships, permanent magnet propulsion motors can improve the power performance, energy utilization efficiency of ships and have the characteristics of low vibration and noise; in electric vehicles, they can improve the endurance mileage and power output.

[0003] Traditional offline monitoring technology of permanent magnet propulsion motors needs to inject current to identify parameters when the motor is stationary, which is complex, time-consuming and low in efficiency, and the offline identification results cannot accurately reflect the parameters under actual operating conditions of the motor. The existing online identification method has the problem of parameter rank loss, which leads to low accuracy of parameter identification results and cannot meet the requirements. At the same time, the artificial periodic inspection method is not only low in efficiency and strong in subjectivity, but also cannot obtain the running state of the motor in real time, and cannot give early warning for some potential fault hidden dangers and cannot discover and handle some sudden faults in time.

[0004] With the development of big data, artificial intelligence and other technologies, the intelligentization and automation of equipment in various industries are continuously improved. For permanent magnet propulsion motors, it is necessary to monitor their health state in real time and accurately, to realize early warning of hidden dangers and accurate diagnosis of faults, so as to take timely maintenance measures, reduce maintenance cost and downtime, and improve the task reliability and operating efficiency of permanent magnet propulsion motors. In the ship industry, real-time monitoring of the health state of permanent magnet propulsion motors can avoid sailing accidents caused by permanent magnet propulsion motor failures; in the electric vehicle industry, it can improve the safety and reliability of vehicles and enhance user experience.

[0005] The technical features of CN117491869A, a permanent magnet motor health state monitoring control method and system based on digital twinning, are to construct a digital twinning motor model, randomly generate initial values of motor parameters, compare predicted current with actual current to obtain optimal motor parameters, input them into a zero-error prediction motor model to obtain the required reference voltage to realize control of the permanent magnet motor. Its essence is online monitoring of permanent magnet motor parameters. The technical features of CN117267066A, an online monitoring system for a wind turbine, are to use a temperature detection module to detect the temperature of the generator bearing, judge the temperature, and control the cooling module to act when the temperature exceeds the upper limit of the set value. Its beneficial effect is to realize real-time detection and automatic cooling of the temperature of the wind turbine bearing, and to avoid bearing scrap due to high temperature.

[0006] CN116298868A A technical solution of a method for detecting static eccentricity fault of a wound brushless doubly-fed motor, the technical solution is a marine predator algorithm optimized random forest diagnostic model. CN114935483A A technical feature of a sugar coating removal equipment for drug testing based on a crankshaft adjustment, the method adopts fuzzy mathematics membership to obtain fault characteristics from online monitoring data of the drive motor, and the possibility of the fault mode of the drive motor is quickly obtained. CN110554316A A motor fault diagnosis method is similar to the fault diagnosis method of CN114935483A a sugar coating removal equipment for drug testing based on a crankshaft adjustment.

[0007] CN110133500A A technical solution of an online monitoring and fault precursor diagnosis system and method for an electric motor based on a multi-layer architecture, the technical solution is to construct a multi-level diagnostic service layer, and use hierarchical screening and step-by-step analysis and diagnosis to identify fault precursors of the motor operating state, the diagnostic result accuracy is higher, and the diagnostic efficiency is greatly improved. This technical solution is not suitable for the ship and automobile industries. CN109765484A An online monitoring and fault diagnosis method based on a "correct tree" model collects voltage and current data of the motor in real time, uses machine learning algorithms to learn the operating mode of the motor, establishes a "correct tree" model of the motor operating in normal working conditions, compares with the actual operating conditions of the motor, and discovers mechanical and electrical faults of the motor in the early stage and alarms, reduces unplanned downtime of equipment, and improves productivity. SUMMARY

[0008] In view of the above deficiencies in the prior art, one of the purposes of the present application is to provide an online monitoring device for the health status of a permanent magnet propulsion motor.

[0009] The technical solution adopted by the present application to solve its technical problems is: an online monitoring device for the health status of a permanent magnet propulsion motor for monitoring a permanent magnet propulsion motor composed of a permanent magnet motor and a frequency converter, mainly including a core computing circuit, a large-capacity memory connected with the core computing circuit, and a communication circuit, the communication circuit is connected with the controller of the permanent magnet propulsion motor for communication to obtain real-time operating data of the permanent magnet propulsion motor.

[0010] Further, the core computing circuit adopts JM9271 GPU chip, the large-capacity memory adopts a solid state disk array based on a master control device, NAND flash memory chip and DRAM cache chip, and the communication circuit adopts Ethernet communication and is connected to the gateway of the controller.

[0011] The second object of the present application is to provide a permanent magnet propulsion motor health state online monitoring method, which can monitor the health state of the permanent magnet propulsion motor online, identify abnormal conditions in the operation process, and provide preventive maintenance opportunity and suggestions.

[0012] The technical solution adopted by the present application to solve its technical problems is: a permanent magnet propulsion motor health state online monitoring method, comprising the following steps:

[0013] S1, the core computing circuit stores the real-time operation data obtained by the communication circuit in the mass storage according to the sample set format, the core computing circuit continuously learns the sample set data, and combines the fault threshold value stored in the mass storage to form a normal distribution function of the normal operation parameters of the permanent magnet propulsion motor under each working condition , wherein x is the point for which the probability is to be calculated, t is used to traverse all possible values from -∞ to x, μ is the mean, σ is the standard deviation, σ>0, and the normalization constant ensures that the integral of the entire probability density function from -∞ to +∞ is equal to 1;

[0014] S2, the core computing circuit stores the normal distribution function parameters in the mass storage, and the core computing circuit compares the real-time operation data with the normal distribution function to identify abnormal data online;

[0015] S3, a fault diagnosis model is designed: assuming that the sample set conforms to the normal distribution X~N(μ,σ 2 ), wherein μ is the mathematical expectation of the normal distribution, σ 2 is the variance of the normal distribution, the core computing circuit estimates the μ and σ 2 parameters based on the collected sample set data according to the least square method, the core computing circuit determines whether the real-time operation data is between the data points with six standard deviations from the mean value, i.e. in the interval (μ-6σ, μ+6σ) according to the six sigma principle, and simultaneously requires that μ-6σ and μ+6σ do not reach the alarm threshold and the fault threshold set by the permanent magnet propulsion motor, and the running data is abnormal if it exceeds the six sigma interval;

[0016] S4, after the core computing circuit identifies the possible hidden danger online based on the six sigma principle, the core computing circuit analyzes and diagnoses the possible failure causes in the fault diagnosis model, gives preventive maintenance suggestions, and evaluates to form a preventive maintenance report;

[0017] S5, the core computing circuit identifies the abnormal data, stores it in the mass storage together with the possible hidden danger and the preventive maintenance report, and simultaneously transmits the hidden danger information and the preventive maintenance report to the permanent magnet propulsion motor controller through the communication circuit.

[0018] Further, the single fault phenomenon in S4 is taken as a top event, a plurality of possible causes are set as bottom events by a plurality of fault trees designed, and a preventive maintenance report is formed according to abnormal operation data.

[0019] Further, when the heat sink temperature of the A-phase power unit in the frequency converter is taken as operation data by the core calculation circuit through the communication circuit, the online monitoring method comprises the following steps:

[0020] 1) reading the heat sink temperature t1, the water inlet temperature t2, the water inlet pressure p1 and the water outlet pressure p2, and storing the above data into a mass storage;

[0021] 2) the core calculation circuit analyzes the t1 data in the sample set data, calculates the equidistant normal distribution function P(X=k)=C(n,k)p k (1-p) n-k , estimates the mean value mu t1 and the variance sigma t1 ;

[0022] 3) judging whether the current heat sink temperature t1 exceeds the alarm threshold AT t1 and the fault threshold FT t1 , if the fault threshold is exceeded, it is determined as a fault, if the alarm threshold is exceeded, it is determined as an alarm;

[0023] 4) for the data without alarm and fault, according to the six sigma principle, it is judged whether it is in the interval (mu t1 -6sigma t1 , mu t1 +6sigma t1 ): if it exceeds the six sigma interval, it is determined as an abnormal situation, otherwise, it is a normal situation for evaluation;

[0024] 5) diagnosing the alarm, fault and abnormal situation through the fault diagnosis model.

[0025] The monitoring device can form a sample set by collecting the operation data of the permanent magnet propulsion motor, design a fault diagnosis model, learn and train to form the normal distribution of various working conditions of the permanent magnet propulsion motor, identify possible hidden dangers online based on the six sigma principle, analyze and diagnose the possible causes of the fault in the fault diagnosis model, and evaluate to form a preventive maintenance report.

[0026] The monitoring method can diagnose abnormal conditions of the permanent magnet propulsion motor in operation online, provide preventive maintenance suggestions for the permanent magnet propulsion motor with potential hidden dangers, and provide repair maintenance suggestions for the permanent magnet propulsion motor with alarm or fault, thereby improving the task reliability of the permanent magnet propulsion motor and enabling centralized training and learning with other sample sets of the permanent magnet propulsion motor. BRIEF DESCRIPTION OF DRAWINGS

[0027] Figure 1 It is an online monitoring device structure diagram of the application.

[0028] Figure 2 It is an online monitoring process diagram of the application.

[0029] Figure 3 It is a single fault diagnosis model fault tree diagram related to the application.

[0030] Figure 4 It is a comprehensive fault diagnosis model fault tree diagram related to the application.

[0031] The reference numerals are as follows: 1-core computing circuit, 2-large capacity memory, 3-communication circuit, 4-controller. DETAILED DESCRIPTION

[0032] The technical solutions of the application are further specifically described as follows in combination with the drawings and embodiments.

[0033] Embodiment 1

[0034] The embodiment is as shown in the figure. Figure 1 The permanent magnet propulsion motor of the certain ship is composed of a permanent magnet motor and a frequency converter, and is operated and controlled by a controller.

[0035] The application discloses an online monitoring device for health states of permanent magnet propulsion motors, which mainly comprises a core computing circuit 1, a large capacity memory 2 connected with the core computing circuit 1, a communication circuit 3 and other circuits.

[0036] The core computing circuit 1 is selected from JM9271 GPU chips, has strong parallel computing capacity, deep learning and real-time processing capacity and the like.

[0037] The large capacity memory 2 is a 500 TB solid state hard disk array adopting main control devices, NAND flash memory chips and DRAM cache chips and the like core devices, has the characteristics of large storage capacity, high reliability, small size and the like.

[0038] The communication circuit 3 is an Ethernet communication, connected to the gateway of the permanent magnet propulsion motor controller, communicates with the permanent magnet propulsion motor, collects real-time running data, and stores in the mass storage 2 through the core computing circuit 1. Other circuits include power supply circuit, data reading interface circuit and the like.

[0039] The communication circuit 3 is connected with the controller of the permanent magnet propulsion motor to communicate and obtain real-time running data of the permanent magnet propulsion motor; the core computing circuit 1 stores the data obtained by the communication circuit 3 in the mass storage 2 in the form of a sample set, the core computing circuit 1 continuously learns the sample set data in the mass storage 2, combines the fault threshold value stored in the mass storage 2, forms a normal distribution function of various working conditions of the permanent magnet propulsion motor, and stores the normal distribution function parameters in the mass storage 2.

[0040] The mass storage 2 can be flexibly taken out, and the sample set data and other sample set data of the permanent magnet propulsion motor are stored in the data center for mutual learning and training to estimate more optimal evaluation parameters.

[0041] Embodiment 2

[0042] As shown in Figure 2 , the disclosed permanent magnet propulsion motor health state online monitoring method comprises the following steps.

[0043] S1, the core computing circuit 1 stores the real-time running data obtained by the communication circuit 3 in the mass storage 2 in the form of a sample set, the core computing circuit 1 continuously learns the sample set data, combines the fault threshold value stored in the mass storage 2, forms a normal distribution function of various working conditions of the permanent magnet propulsion motor , which is used to represent the probability that the value of a random variable is less than or equal to x, wherein the independent variable x is the point (threshold value) to be calculated, the integral variable t is used to traverse all possible values from -∞ to x in the integral operation, the mean μ is also called the expected value, which is the center position or average value of the normal distribution, and its value range is -∞<μ<+∞, the standard deviation σ measures the dispersion degree of data, the larger the σ, the more dispersed the data, and the more "stout" the curve, the smaller the σ, the more concentrated the data, and the more "slim" the curve, and σ>0, the variance σ 2 is the square of the standard deviation, which is also used to measure the dispersion degree of data, e is the natural constant, and the normalization constant ensures that the integral of the entire probability density function from -∞ to +∞ is equal to 1, that is, it satisfies the axiom of probability.

[0044] S2, the core computing circuit 1 stores the normal distribution function parameters in the mass storage 2, and the core computing circuit 1 compares the real-time running data with the normal distribution function to identify abnormal data online.

[0045] S3, design a fault diagnosis model.

[0046] Assume that the sample set conforms to a normal distribution X ~ N(μ,σ 2 ), where μ is the mathematical expectation (mean) of the normal distribution, and σ 2 is the variance of the normal distribution. Based on the collected sample set data, the core computing circuit 1 estimates the μ and σ 2 parameters according to the least square method. According to the 6σ (six sigma) principle, the core computing circuit 1 determines whether the real-time running data is between data points that are six standard deviations away from the mean, i.e., in the interval (μ-6σ, μ+6σ). At the same time, it is required that μ-6σ and μ+6σ do not reach the alarm threshold and fault threshold set for the permanent magnet propulsion motor. If it exceeds the six sigma interval, it is determined that the running data is abnormal. The fault diagnosis model of the present application takes into account both single fault phenomenon fault diagnosis and comprehensive diagnosis of multiple fault phenomena.

[0047] S4, analyze the possible causes of potential faults or hidden dangers to form a preventive maintenance report through the designed fault diagnosis model.

[0048] After the core computing circuit 1 identifies the possible hidden dangers online based on the six sigma principle, it analyzes and diagnoses the possible causes of the faults in the fault diagnosis model, gives preventive maintenance suggestions, and evaluates to form a preventive maintenance report.

[0049] This step takes a single fault phenomenon, i.e., a fault mode, as a top event, and through several fault trees that have been designed, infers multiple possible causes as bottom events. Thus, an abnormal running data forms a preventive maintenance report.

[0050] The fault diagnosis model analyzes and identifies the most initial source, i.e., the top event, of multiple fault phenomena, and through fault tree analysis, infers the most possible cause, i.e., the bottom event. For multiple fault phenomena that are not easy to identify a single top event, their respective fault trees are analyzed, and the intersection of the possible causes of each is taken as the bottom event.

[0051] Abnormal data forms a preventive maintenance report with this comprehensive fault diagnosis model; the preventive maintenance report contains suggestions for current safe operation, preventive maintenance timing, preventive maintenance components, tools, materials, and steps.

[0052] S5, the core computing circuit 1 identifies abnormal data and stores it together with possible hidden dangers and a preventive maintenance report in the mass storage 2. After the core computing circuit 1 identifies the hidden dangers, it transmits hidden danger information and a preventive maintenance report to the permanent magnet propulsion motor controller 4 through the communication circuit 3.

[0053] The permanent magnet propulsion motor controller 4 displays abnormal conditions, and can call up a preventive maintenance report by operation of an operator, and the operator can improve the reliability of the permanent magnet propulsion motor according to the preventive maintenance report. Through the designed fault diagnosis model, the possible causes of potential faults or hidden dangers are analyzed to form a preventive maintenance report.

[0054] Embodiment 3

[0055] The online monitoring method disclosed in this embodiment has the following steps.

[0056] 1) Start evaluation, read real-time running data of the permanent magnet propulsion motor to form sample set data.

[0057] 2) Estimate the mean and variance parameters of the normal distribution function, and judge whether the sample set data exceeds the alarm threshold and the fault threshold.

[0058] 3) If yes, select alarm or prompt fault, and then go to step 5).

[0059] 4) If no, judge whether the data exceeds the (μ-6σ, μ+6σ) interval, otherwise complete the evaluation; if yes, go to step 5) according to the abnormal condition.

[0060] 5) Judge whether it is a single fault.

[0061] 6) If yes, establish a fault tree, infer multiple possible causes, and go to step 8).

[0062] 7) If no, identify the fault mode, establish a fault tree, infer the most likely cause or the intersection of the most likely causes, and go to step 8).

[0063] 8) Judge whether it is necessary to alarm or prompt fault.

[0064] 9) If yes, give a corrective maintenance report, otherwise form a preventive maintenance report, and complete the evaluation.

[0065] Embodiment 4

[0066] As shown in Figure 3 , Figure 4 , the running data of the heat sink temperature of the A-phase power unit of the permanent magnet propulsion motor frequency converter is taken as an example to illustrate the process of the permanent magnet propulsion motor health state online monitoring as follows.

[0067] 1) Read the heat sink temperature t1, the water inlet temperature t2, the water inlet pressure p1 and the water outlet pressure p2.

[0068] 2) Store the above data to the mass storage 2.

[0069] 3), the core computing circuit 1 calculates the equal-interval scatter normal distribution function P(X=k)=C(n,k)p k (1-p) n-k , where p is the "success" probability of each test, 0≤p≤1, the combination number C(n,k)= is the number of all possible ways to select k successes from n tests, the "failure" probability (1-p) of each test is the probability that the "success" event does not occur in a single test, the random variable X represents the number of "successes" in n independent tests, the number of successes k is a specific number of successes of interest, k=0,1,2,…,n, and the total number of tests n is the total number of times a test is performed independently and repeatedly.

[0070] 4), the mean μ t1 and the variance σ t1 are estimated.

[0071] 5), it is determined whether the current heat sink temperature t1 exceeds the alarm threshold AT t1 and the failure threshold FT t1 ; if it exceeds the failure threshold, it is determined to be a failure, and if it exceeds the alarm threshold, it is determined to be an alarm.

[0072] 6), for data without alarms and failures, it is determined whether it is in the interval (μ t1 -6σ t1 , μ t1 +6σ t1 ) according to the six-sigma principle.

[0073] 7), if it exceeds the six-sigma interval, it is determined to be an abnormal situation, otherwise it is a normal situation and the evaluation is completed.

[0074] 8), alarms, failures, and abnormal situations are all diagnosed by a fault diagnosis model. The following further illustrates fault diagnosis for abnormal situations.

[0075] 9), further, only the heat sink temperature t1 is abnormal, which is a high temperature, and a single fault diagnosis model is established, as shown in Figure 3 .

[0076] 10), after single fault diagnosis, the possible reasons are: one is that the heat loss of the phase power unit device increases, two is that the cooling water flow of the phase power unit heat sink decreases, and three is that the water inlet temperature increases.

[0077] 11), after single fault diagnosis, by comparing the temperature changes of the heat sinks of other power units, only the temperature of this power unit is high, which excludes the third possible reason.

[0078] 12), through single fault diagnosis, a preventive maintenance report is formed, and the content includes inspection suggestions for the power unit power device and the heat sink.

[0079] 13), further, steps 3) to 7) are used to determine that the heat sink temperature t1 is abnormally high and the water inlet temperature t2 is abnormally high, a fault tree is established through a comprehensive fault diagnosis model as shown in Figure 4

[0080] 14), through the comprehensive fault diagnosis model, three possible reason sets {the heat loss of the phase power unit device increases, the cooling water flow of the phase power unit heat sink decreases, and the water inlet temperature t2 is abnormally high} of the abnormally high heat sink temperature t1 and the abnormally high water inlet temperature t2 are analyzed, and it is identified that the fault mode is the abnormally high water inlet temperature t2. Take this as the top event to establish a fault tree as shown in Figure 4

[0081] 15), through the comprehensive fault diagnosis model, a preventive maintenance report is formed, and it is suggested to check the plate exchange of the external cooling equipment.

[0082] The permanent magnet propulsion motor described in the patent comprises a permanent magnet motor and a propulsion frequency converter. The health state online monitoring described in the patent is not only a simple combination of the health state online monitoring of the permanent magnet motor and the propulsion frequency converter. The technical solution of the present application is different from patents CN114935483A, CN110554316A, CN114935483A, CN110133500A, and CN109765484A. Compared with patent CN117491869A, the technical solution and beneficial effects are different. Compared with patent CN117267066A, the technical solution is more intelligent, and the beneficial effects are more abundant. Compared with patent CN116298868A, the model is different, and the online diagnosis parameters of the present application are temperature and current, and the beneficial effects are different.

[0083] The scope of protection of the present application is not limited to the above-mentioned embodiments. For those skilled in the art, without departing from the inventive concept, several adjustments and improvements can be made, which are all within the scope of protection of the present application.​​

Claims

1. An online health status monitoring device for a permanent magnet propulsion motor, used to monitor a permanent magnet propulsion motor composed of a permanent magnet motor and a frequency converter, characterized in that: It includes a core computing circuit (1), a large-capacity memory (2) connected to the core computing circuit (1), and a communication circuit (3). The communication circuit (3) is connected to the controller (4) of the permanent magnet propulsion motor to communicate and obtain real-time operating data of the permanent magnet propulsion motor.

2. The online health status monitoring device for a permanent magnet propulsion motor according to claim 1, characterized in that, The core computing circuit (1) uses a JM9271 GPU chip, the large-capacity memory (2) uses a solid-state hard disk array based on the main controller, NAND flash memory chip and DRAM cache chip, and the communication circuit (3) uses Ethernet communication to connect to the gateway of the controller (4).

3. A monitoring method for the online health status monitoring device of the permanent magnet propulsion motor as described in claim 1, characterized in that, Includes the following steps: S1, the core computing circuit (1) stores the real-time operating data obtained by the communication circuit (3) in a large-capacity memory (2) according to the sample set format. The core computing circuit (1) continuously learns the sample set data and combines it with the fault threshold stored in the large-capacity memory (2) to form a normal distribution function of the normal operating parameters of the permanent magnet propulsion motor under various working conditions. In the formula, x is the point from which the probability is to be calculated, t is used to iterate through all possible values ​​from -∞ to x, μ is the mean, σ is the standard deviation, σ > 0, and is the normalization constant. Ensure that the integral of the entire probability density function from -∞ to +∞ is equal to 1; S2, the core computing circuit (1) stores the normal distribution function parameters in a large-capacity memory (2), and the core computing circuit (1) compares the real-time running data with the normal distribution function to identify abnormal data online; S3, Design a fault diagnosis model: Assume the sample set follows a normal distribution X ~ N(μ,σ). 2 ), where μ is the mathematical expectation of the normal distribution, σ 2 The variance is the normal distribution. The core calculation circuit (1) estimates μ and σ based on the sample set data using the least squares method. 2 The parameters are used to determine whether the real-time operating data is within the range of (μ-6σ, μ+6σ) according to the 6σ principle. At the same time, μ-6σ and μ+6σ are required not to reach the alarm threshold and fault threshold set by the permanent magnet propulsion motor. If they exceed this range, the operating data is judged to be abnormal. S4, the core computing circuit (1) identifies potential hidden dangers online, analyzes and diagnoses the causes of possible failures, and generates a preventive maintenance report; S5, the core computing circuit (1) identifies abnormal data and stores it together with potential hazards and preventive maintenance reports in a large-capacity memory (2). At the same time, it transmits hazard information and preventive maintenance reports to the controller (4) via the communication circuit (3).

4. The method for online monitoring of the health status of a permanent magnet propulsion motor according to claim 3, characterized in that, In S4, a single fault phenomenon is taken as the top event, and multiple possible causes are deduced through several fault trees and set as bottom events. A preventive maintenance report is formed based on abnormal operating data.

5. The method for online monitoring of the health status of a permanent magnet propulsion motor according to claim 4, characterized in that, When using the heat sink temperature of the A-phase power unit in the inverter obtained by the core calculation circuit (1) as operating data, the steps are as follows: 1) Read the heat sink temperature t1, inlet water temperature t2, inlet water pressure p1 and outlet water pressure p2, and store them in a large capacity memory (2). 2) The core computing circuit (1) analyzes t1 in the sample set data and calculates its equidistant normal distribution function P(X=k)=C(n,k)p k (1-p) n-k Estimate its mean μ t1 and variance σ t1 ; 3) Determine whether the current heat sink temperature t1 exceeds the alarm threshold AT. t1 and fault threshold FT t1 If the fault threshold is exceeded, it is determined to be a fault; if the alarm threshold is exceeded, it is determined to be an alarm. 4) For data without alarms or faults, determine whether it falls within (μ) according to the Six Sigma principle. t1 -6σ t1 ,μ t1 +6σ t1 If the range exceeds the six sigma range, it is considered an abnormal situation; otherwise, the evaluation is completed under normal circumstances. 5) The fault diagnosis model is used to diagnose alarms, faults and abnormal conditions.

Images