A refrigeration host health status assessment method based on assessment condition parameter estimation and refrigeration host

Abstract

The present invention provides a refrigeration host health status assessment method based on assessment condition parameter estimation and a refrigeration host; the method is: 1. Classify the correlation between fault characteristic parameters and working environment parameters; 2. Set the working environment parameter combination under the assessment condition. 3. Establish a membership function between the health degrees corresponding to the fault characteristic parameters. 4. Establish and dynamically update the correlation assessment model. Predict each fault characteristic parameter under the assessment condition and calculate the corresponding health degree. The present invention uses the newly added data to update the data set and re-establish the correlation assessment model in each detection cycle; thus, under non-assessment conditions, the updated correlation assessment model is used to predict the fault characteristic parameters under the assessment condition, unitize and unify the complex data at different times and under different working conditions, eliminate the changes in fault characteristic parameters caused by differences in working conditions, and accurately obtain multi-dimensional health degrees under different working conditions.

Description

Technical Field

[0001] The present invention relates to the technical field of refrigeration unit health status assessment, and in particular to a refrigeration host health status assessment method based on assessment working condition parameter estimation and a refrigeration host. Background Art

[0002] The refrigeration host is one of the most important equipment in the refrigeration system. The health status assessment of the refrigeration host can guide the maintenance of the equipment and avoid under-maintenance and over-maintenance of the equipment. It has important guiding significance for effectively reducing the probability of equipment failure and implementing pre-emptive maintenance.

[0003] Most of the existing equipment health status assessment technologies use advanced artificial intelligence algorithms to identify the equipment status corresponding to different operating parameters. However, existing patented technologies often focus on the combinatorial optimization of the algorithm itself, complicating the uncertainty problem and using machine learning to study the patterns and laws. On the one hand, this makes the algorithm too complex, has high requirements for computer hardware, and is difficult to deploy on site. On the other hand, these algorithms rely on massive amounts of historical data, including operating data of health status and fault status. In actual engineering projects, it is difficult to obtain equipment fault status parameters on site, and equipment manufacturers are rarely willing to risk host damage to conduct fault status test runs before the equipment leaves the factory.

[0004] Therefore, there is an urgent need for a refrigeration host health status assessment method that is streamlined, convenient, more in line with engineering practice, and can be completed using equipment health status data. Summary of the invention

[0005] The present invention provides a refrigeration host health status assessment method based on assessment working condition parameter estimation and a refrigeration host.

[0006] In a first aspect, the present invention provides a method for evaluating the health status of a refrigeration host based on estimation of assessment working condition parameters, which comprises the following steps:

[0007] Step 1: Determine multiple fault characteristic parameters P of the refrigeration host i ; For each fault characteristic parameter P i A plurality of associated working environment parameters are screened separately, and the degree of association is graded.

[0008] Step 2: Set the working environment parameter combination under the assessment conditions.

[0009] Step 3: Establish the characteristic parameters P of each fault under the test conditions i The corresponding health level H i The membership function between .

[0010] Step 4: For each fault characteristic parameter P i, one or more working environment parameters are selected as the prediction input parameter X according to the correlation level from high to low ij , establish the prediction input parameter X ij Predicted fault characteristic parameter P i The associated evaluation model.

[0011] Step 5: After each inspection cycle, each fault characteristic parameter P i Reselect the forecast input parameter X ij In each detection cycle, the associated evaluation model is used to predict the characteristic parameters P of each fault under the test condition. i ; Through the fault characteristic parameter P under the test condition i Get the corresponding health H i According to health H i Determine the health of different aspects of the refrigeration unit.

[0012] Preferably, the process of determining the working environment parameter combination under the assessment condition is: selecting the working environment parameter combination with the highest density value by a kernel density estimation method as the working environment parameter under the assessment condition.

[0013] As a preferred method, the process of establishing the association evaluation model is as follows:

[0014] (1) Establish a data set containing fault characteristic parameters and working environment parameters.

[0015] (2) For any fault characteristic parameter P i , take the working environment parameter with the highest correlation level as the prediction input parameter X ij .

[0016] (3) For the prediction input parameter X ij and fault characteristic parameter P i Fitting is performed to obtain the associated evaluation model.

[0017] (4) Perform a significance test on the association evaluation model; if the association evaluation model does not meet the significance requirement, add the next level working environment parameter as the new prediction input parameter X ij , and re-execute step (3) until the association evaluation model meets the significance requirement.

[0018] Preferably, the inspection period T is 7 to 60 days.

[0019] As a preferred method, obtain the health level H i After that, for each health degree H i Perform time domain analysis to predict the health level H i The trend of changes in health iAn alarm message is issued before the predicted time when the warning threshold is reached.

[0020] As a preferred method, obtain the health level H i After that, calculate the comprehensive health H s as follows:

[0021]

[0022] in, H i The corresponding weights;

[0023] According to the comprehensive health s Determine the health status of the refrigeration host, the comprehensive health degree H s The larger the value, the healthier the cooling unit is.

[0024] Preferably, the fault characteristic parameters include evaporation pressure, exhaust pressure, lubricating oil pressure and lubricating oil temperature in the lubricating oil system, operating current in the circuit system and motor stator winding temperature, which correspond to fault characteristic parameters P1 to P6 respectively.

[0025] Preferably, the membership function corresponding to the fault characteristic parameters P1 to P5 adopts an intermediate membership function, and the expression is as follows:

[0026]

[0027] Where i=1,2,...,5; w i,1 、w i,2 、w i,3 、w i,4 They are fault characteristic parameters P i The lower limit of the operating permission, the lower limit of the normal range, the upper limit of the normal range, and the upper limit of the operating permission.

[0028] The membership function corresponding to the fault characteristic parameter P6 adopts the smaller the better type of membership function, and the expression is as follows:

[0029]

[0030] Among them, i=6.

[0031] Preferably, the working environment parameters include chilled water flow, chilled water inlet temperature, compressor speed, guide vane opening, throttle valve opening, ambient temperature, main engine power, operating voltage, cooling water flow and cooling water inlet temperature.

[0032] The first-level associated working environment parameters corresponding to the evaporation pressure are the chilled water flow rate and the chilled water inlet temperature, the second-level associated working environment parameters are the compressor speed, the guide vane opening, and the throttle valve opening, and the third-level associated working environment parameters are the ambient temperature and the host power. The first-level associated working environment parameters corresponding to the exhaust pressure are the cooling water flow rate and the cooling water inlet temperature, the second-level associated working environment parameters are the compressor speed, the guide vane opening, and the throttle valve opening, and the third-level associated working environment parameters are the ambient temperature and the host power. The first-level associated working environment parameters corresponding to the lubricating oil pressure are the compressor speed, and the second-level associated working environment parameters are the ambient temperature and the host power. The first-level associated working environment parameters corresponding to the lubricating oil temperature are the compressor speed and the ambient temperature, and the second-level associated working environment parameters are the host power. The first-level associated working environment parameters corresponding to the working current are the compressor speed and the working voltage, and the second-level associated working environment parameters are the ambient temperature and the host power. The first-level associated working environment parameters corresponding to the motor stator winding temperature are the compressor speed and the ambient temperature, and the second-level associated working environment parameters are the host power.

[0033] In a second aspect, the present invention provides a refrigeration host based on fault characteristic parameter monitoring, which includes a refrigeration main body, a parameter acquisition module and a health assessment module, wherein the parameter acquisition module is used to acquire the fault characteristic parameters and working environment parameters of the refrigeration main body. The health assessment module is used to execute the above-mentioned refrigeration host health status assessment method.

[0034] The main innovative features of the present invention are as follows:

[0035] 1. The present invention uses newly added data to update the data set and re-establish the associated evaluation model in each detection cycle; thus, under non-assessment conditions, the updated associated evaluation model is used to predict the fault characteristic parameters under the assessment conditions, and the complex data of different times and different conditions are unitized and unified, eliminating the changes in fault characteristic parameters caused by differences in working conditions, so that the multi-dimensional health can be accurately obtained under different working conditions; based on this, the present invention only needs to perform time domain analysis on the health of the assessment conditions in different detection cycles, so as to quickly analyze the health status and health decay trend of the equipment.

[0036] 2. The present invention performs time domain analysis based on multi-dimensional health, which helps to predict the occurrence of failures in the refrigeration host. In addition, the present invention focuses on the attenuation trend of the health status of the equipment during normal operation, and can arbitrarily select the evaluation and inspection cycle. It requires less historical data and can achieve accurate health assessment when no substantial failures occur in the refrigeration host. The data acquisition threshold is low and the applicability is strong.

[0037] 3. The present invention divides the associated working environment parameters into correlation levels from the perspective of equipment mechanism for different fault characteristic parameters, and only selects some working environment parameters with high correlation to participate in modeling, thereby reducing the model complexity of the associated evaluation model and improving the modeling efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 is a flow chart of an embodiment of the present invention. DETAILED DESCRIPTION

[0039] The present invention will be further described below in conjunction with the accompanying drawings.

[0040] Example

[0041] A refrigeration host health status assessment method based on the estimation of assessment condition parameters is taken as the object of a MaxE YK chiller of Johnson Controls (rated cooling capacity: 3516KW, rated power: 652KW).

[0042] The refrigeration host health status assessment method comprises the following steps:

[0043] Step 1: According to the manufacturer's information and equipment principles, determine the fault characteristic parameters, health calculation method, and health weights corresponding to each parameter, and select the working environment parameters according to the equipment's gradual failure mechanism.

[0044] In this embodiment, the fault characteristic parameters include fault characteristic parameters P1 to P6, which are respectively the evaporation pressure (P1) and the exhaust pressure (P2) in the refrigeration cycle system, the lubricating oil pressure (P3) and the lubricating oil temperature (P4) in the lubricating oil system, and the working current (P5) and the motor stator winding temperature (P6) in the circuit system.

[0045] The working environment parameters include working environment parameters X1~X8, which are chilled water flow (X1), chilled water inlet temperature (X2), compressor speed (X3), guide vane opening (X4), throttle valve opening (X5), ambient temperature (X6), main engine power (X7) and working voltage (X8), cooling water flow (X9), cooling water inlet temperature (X1), compressor speed (X2), guide vane opening (X3), throttle valve opening (X4), ambient temperature (X5), main engine power (X6), working voltage (X7), cooling water flow (X8), cooling water inlet temperature (X1), compressor speed (X2), compressor guide vane opening (X4), throttle valve opening (X5), ambient temperature (X6), main engine power (X7), working voltage (X8), cooling water flow (X9), cooling water inlet temperature (X1), compressor speed (X2), compressor guide vane opening (X4), throttle valve opening (X5), ambient temperature (X6), main engine power (X7), working voltage (X 10 ).

[0046] The correlation between the fault characteristic parameters and the working environment parameters is graded into three levels, namely level 1, level 2 and level 3 with decreasing correlation degrees.

[0047] The level 1 associated working environment parameters corresponding to the evaporating pressure (P1) are the chilled water flow rate (X1) and the chilled water inlet temperature (X2); the level 2 associated working environment parameters are the compressor speed (X3), the guide vane opening (X4), and the throttle valve opening (X5); the level 3 associated working environment parameters are the ambient temperature (X6) and the main engine power (X7).

[0048] The first-level associated working environment parameters corresponding to the exhaust pressure (P2) are cooling water flow (X9), cooling water inlet temperature (X 10 ), the level 2 related working environment parameters are compressor speed (X3), guide vane opening (X4), throttle valve opening (X5), and the level 3 related working environment parameters are ambient temperature (X6) and main engine power (X7).

[0049] The level 1 associated working environment parameter corresponding to the lubricating oil pressure (P3) is the compressor speed (X3), the level 2 associated working environment parameters are the ambient temperature (X6) and the main engine power (X7), and there is no level 3 associated working environment parameter.

[0050] The level 1 associated working environment parameters corresponding to the lubricating oil temperature (P4) are the compressor speed (X3) and the ambient temperature (X6), the level 2 associated working environment parameter is the host power (X7), and there is no level 3 associated working environment parameter.

[0051] The level 1 associated working environment parameters corresponding to the working current (P5) are the compressor speed (X3) and the working voltage (X8); the level 2 associated working environment parameters are the ambient temperature (X6) and the host power (X7); there are no level 3 associated working environment parameters.

[0052] The level 1 associated working environment parameters corresponding to the motor stator winding temperature (P6) are the compressor speed (X3) and the ambient temperature (X6), the level 2 associated working environment parameter is the host power (X7), and there is no level 3 associated working environment parameter.

[0053] Step 2: Use the kernel density estimation method to select the working environment parameter combination with the highest density value as the working environment parameter under the test condition. In this embodiment, under the test condition, the value of the chilled water flow rate (X1) is 500m 3 / h, chilled water inlet temperature (X2) is 11℃, compressor speed (X3) is 3600rpm, guide vane opening (X4) is 80%, throttle valve opening (X5) is 80%, ambient temperature (X6) is 30℃, main engine power (X7) is 652kW, working voltage (X8) is 380V, cooling water flow rate (X9) is 720m 3 / h, cooling water inlet temperature (X 10 ) is taken as 32℃.

[0054] Step 3: The fault characteristic parameters P1 to P6 each correspond to a health level, namely health levels H1 to H6. i The corresponding health level H i The membership function between .

[0055] The membership function corresponding to the fault characteristic parameters P1 to P5 adopts an intermediate membership function, and the expression is as follows:

[0056]

[0057] Where i=1,2,...,5; w i,1 、w i,2 、w i,3 、w i,4 They are fault characteristic parameters P i The lower limit of the operation permission, the lower limit of the normal range, the upper limit of the normal range, and the upper limit of the operation permission. i,2 ~w i,3 is the fault characteristic parameter P i The suitable interval of P i ≤w i,1 or P i ≥w i,4 The refrigeration host should be shut down.

[0058] The membership function corresponding to the fault characteristic parameter P6 adopts the smaller the better type of membership function, and the expression is as follows:

[0059]

[0060] Where i=6; w i,2 、w i,3 They are fault characteristic parameters P i The lower limit of the normal interval and the upper limit of the normal interval. i,2 ~w i,3 is the fault characteristic parameter P i normal operating range.

[0061] In this embodiment, the membership function corresponding to the fault-free characteristic parameter adopts the larger the better type of membership function; in some other embodiments, there are fault characteristic parameters whose membership function adopts the larger the better type of membership function; for these fault characteristic parameters P i , and its corresponding membership function expression is as follows:

[0062]

[0063] Among them, w i,2 、w i,3 They are fault characteristic parameters P iThe lower limit of the normal interval and the upper limit of the normal interval. i,2 ~w i,3 is the fault characteristic parameter P i normal operating range.

[0064] Step 4: Establish an initial correlation evaluation model for each fault characteristic parameter P1-P6.

[0065] 4-1. Take the fault characteristic parameters and working environment parameters of the refrigeration host tested at the factory as the standard health status data to establish a data set. The data set is divided into a training set and a test set in a ratio of 7:3 by random sampling.

[0066] 4-2. For any fault characteristic parameter P i , take the corresponding level 1 associated working environment parameter as the prediction input parameter X ij Without considering the timing, the artificial intelligence algorithm is used to calculate the fault characteristic parameters P in the training set. i The corresponding prediction input parameter X ij Fitting is performed to obtain the fault characteristic parameter P i For example, the fault characteristic parameter P1 is the evaporation pressure, and its corresponding level 1 associated working environment parameters are the chilled water inlet temperature and the chilled water flow rate; therefore, the chilled water inlet temperature and the chilled water flow rate are used as the prediction input parameters X ij , by establishing a prediction model for evaporation pressure.

[0067] Prediction input parameter X ij In which, i is the corresponding fault characteristic parameter P i The serial number; j is the prediction input parameter X ij In this embodiment, the artificial intelligence algorithm adopts the support vector regression (SVR) algorithm. The typical steps of the support vector regression algorithm include data preprocessing, kernel function selection, model training, model evaluation, and model application, which belong to the existing algorithm and will not be repeated here. In other embodiments, the support vector regression algorithm can use other methods that can perform fitting solutions.

[0068] 4-3. Use the test set to perform a significance test on the association evaluation model obtained in step 4-2; for the association evaluation model that does not meet the significance requirement, the next level of association working environment parameters are also introduced as the prediction input parameter X ij , re-establish the association evaluation model and re-test the significance until the association evaluation model meets the significance requirements or all associated working environment parameters have been used as prediction input parameters X ij .

[0069] For example, if the associated evaluation model of the evaporation pressure as the fault characteristic parameter P1 does not meet the significance requirement, the compressor speed, guide vane opening, and throttle valve opening belonging to the second-level associated working environment parameters are added to the prediction input parameter X ij After that, the association evaluation model is re-established.

[0070] In this embodiment, the significance test adopts the F test method; in the F test method, the F parameter threshold is set to 1, the P parameter threshold is set to 0.05, and the association evaluation model with an F parameter greater than 1 and a P parameter less than 0.05 is considered to meet the significance requirement.

[0071] Step 5: During the operation of the refrigeration host, the associated evaluation model of the fault characteristic parameters P1 to P6 is updated periodically.

[0072] 5-1. During the operation of the refrigeration host, fault characteristic parameters and working environment parameters are continuously collected; at the end of each detection cycle T, a new data set is established using the fault characteristic parameters and working environment parameters collected during the detection cycle T; the data set is divided into a training set and a test set in a ratio of 7:3 by random sampling.

[0073] 5-2. Use the new training set to reconstruct the correlation evaluation model of each fault characteristic parameter P1 to P6. The process of constructing the correlation evaluation model is consistent with step 4-2 to step 4-3.

[0074] In this step, the value of the detection period T is determined by the following method: the Monte Carlo simulation method is used to simulate the performance of the associated evaluation model under different detection periods, and within the MSE error threshold range (MSE < 0.1), the shortest possible detection period is selected as the optimal detection period T; in this embodiment, the optimal detection period T is 14 days.

[0075] Step 6: In each detection cycle, the associated evaluation model is used to predict the fault characteristic parameters P1-P6 under the test condition; and then the health degree H1-H6 is calculated by the fault characteristic parameters P1-P6 under the test condition and their corresponding membership functions.

[0076] Step 7: For the health levels H1~H6 corresponding to the six fault characteristic parameters P1~P6, time domain analysis is performed respectively, as follows: i The relationship function H with time t i =f i (t). Through the relationship function H i =f(t) predicted health H i of the changing trend.

[0077] Calculate and establish the life prediction equation f i (t s)=0, calculate the health level H i The zeroing time t s ; If the zeroing time t s If the time difference with the current time is less than 5T, a fault characteristic parameter P is issued. i alarm; for example, the risk of abnormal evaporation pressure is high.

[0078] In this embodiment, for each detection cycle, the health H of the current detection cycle and the previous 9 detection cycles is taken. i ; Linear fitting is performed on the 10 data points obtained with respect to time, and the slope of the fitted straight line is used as the health decay rate of the current cycle; the health H of the current cycle is used to calculate the health decay rate of the current cycle. i and health decay rate prediction; this local linear analysis method greatly simplifies the calculation.

[0079] In some other embodiments, the number of data points used for linear fitting is obtained by optimization, specifically by using Monte Carlo simulation to simulate the fitting accuracy of different numbers of data points and take the optimal number of data points.

[0080] Step 8: Calculate the comprehensive health level H based on health levels H1 to H6 s as follows:

[0081]

[0082] in, H i Corresponding weights; in this embodiment to The value of was obtained by expert scoring method, and the values ​​were 0.25, 0.25, 0.15, 0.15, 0.1, and 0.1 respectively.

[0083] According to the comprehensive health s Determine the health status of the refrigeration host; in this embodiment, if H s ≥0.8, the health status of the refrigeration host is judged to be excellent; if 0.8>H s >0.6, the health status is judged to be good; if H s <0.6, the health status is judged to be poor.

[0084] Overall health s It can provide users with an intuitive sense of the overall health of a device.

[0085] Take the comprehensive health H s The derivative of time t (unit is 1 detection cycle, 14 days in the above example) is used as the comprehensive decay rate of health ; If the comprehensive health decay rate ≥0.5, then the health decay rate is judged to be fast); if the comprehensive health decay rate If the health decay rate is greater than 0.1 and less than 0.5, then the health decay rate is normal. ≤0.1, it is judged that the health decay rate is slow.

[0086] The comprehensive health decay rate helps users understand how fast the health of the device decays.

Claims

1. A method for evaluating the health status of a refrigeration host based on estimation of working condition parameters, characterized in that: The following steps are involved: Step 1: Determine multiple fault characteristic parameters of the refrigeration host P i ; For each fault characteristic parameter P i Screening multiple related working environment parameters respectively and grading the degree of correlation; Step 2: Set the working environment parameter combination under the assessment conditions; Step 3: Establish the characteristic parameters of each fault under the assessment conditions P i Corresponding health H i The membership function between Step 4: For each fault characteristic parameter P i , one or more working environment parameters are selected as prediction input parameters according to the correlation level from high to low X ij , establish a correlation evaluation model; The process of building a correlation evaluation model is as follows: (1) Establish a data set containing fault characteristic parameters and working environment parameters; (2) For any fault characteristic parameter P i , take the working environment parameter with the highest correlation level as the prediction input parameter X ij ; (3) Prediction input parameters X ij Fault characteristic parameters P i Perform fitting to obtain the associated evaluation model; (4) Conduct significance test on the association evaluation model; If the association evaluation model does not meet the significance requirement, the next level of working environment parameters will be added as new prediction input parameters. X ij , and re-execute step (3) until the association evaluation model meets the significance requirement; Step 5: After each inspection cycle, each fault characteristic parameter P i Reselect forecast input parameters X ij and establish a correlation evaluation model; in each detection cycle, the correlation evaluation model is used to predict the characteristic parameters of each fault under the test condition P i ; By examining the fault characteristic parameters under the working conditions P i Get the corresponding health H i .

2. A method for evaluating the health status of a refrigeration host based on estimation of working condition parameters according to claim 1, characterized in that: The process of determining the working environment parameter combination under the assessment working condition is as follows: selecting the working environment parameter combination with the highest density value by the kernel density estimation method as the working environment parameter under the assessment working condition.

3. The method for evaluating the health status of a refrigeration host based on estimation of working condition parameters according to claim 1, characterized in that: The inspection cycle T 7 to 60 days.

4. The method for evaluating the health status of a refrigeration host based on estimation of working condition parameters according to claim 1, characterized in that: Get health H i Afterwards, for each health H i Perform time domain analysis to predict health status H i trends and in terms of health H i An alarm message is issued before the predicted time when the warning threshold is reached.

5. The method for evaluating the health status of a refrigeration host based on estimation of working condition parameters according to claim 1, characterized in that: Get health H i After that, calculate the comprehensive health H s as follows: ； in, For health H i The corresponding weights; According to comprehensive health H s Determine the health status of the refrigeration host and the overall health H s The larger the value, the healthier the cooling unit is.

6. The method for evaluating the health status of a refrigeration host based on estimation of working condition parameters according to claim 1, characterized in that: The fault characteristic parameters include evaporation pressure, exhaust pressure, lubricating oil pressure and lubricating oil temperature in the lubricating oil system, working current in the circuit system and motor stator winding temperature, which correspond to fault characteristic parameters respectively. P 1~ P 6.

7. The method for evaluating the health status of a refrigeration host based on estimation of working condition parameters according to claim 1, characterized in that: Fault characteristic parameters P 1~ P 5 The corresponding membership function adopts the intermediate membership function; the fault characteristic parameter P The membership function corresponding to 6 adopts the smaller the better type of membership function.

8. The method for evaluating the health status of a refrigeration host based on estimation of working condition parameters according to claim 1, characterized in that: The working environment parameters include chilled water flow, chilled water inlet temperature, compressor speed, guide vane opening, throttle valve opening, ambient temperature, host power, working voltage, cooling water flow and cooling water inlet temperature.

9. A refrigeration host, comprising a refrigeration main body, a parameter acquisition module and a health assessment module, characterized in that: The parameter acquisition module is used to acquire fault characteristic parameters and working environment parameters of the refrigeration main body; the health assessment module is used to execute the refrigeration host health status assessment method as described in any one of claims 1-8.

Images