A device high-risk operation risk assessment method based on a normal-like process

Abstract

The application discloses a kind of equipment high-risk operation risk assessment method based on normal process, it is extended to the combination form of normal and normal process by traditional normal process, to better adapt to the evaluation needs of equipment high-risk operation;Thus the multiple sub-equipment involved in equipment high-risk operation is respectively solved by state monitoring data sub-equipment state failure time estimate value, and further solves risk assessment value;Finally, a kind of radial basis neural network with normal function and two normal functions as base function is established, and the network error is obtained by the existing equipment risk assessment historical data, and the adaptive weight adjustment law based on error feedback is further designed, and the neural network weight is trained until the error converges;Finally, the state monitoring value of the equipment to be evaluated is substituted, and the high-risk operation process risk assessment total score of the equipment to be evaluated is solved, and the risk assessment is completed.

Description

Technical Field

[0001] This invention relates to a risk assessment method for high-risk equipment operations based on a quasi-normal process, belonging to the field of equipment reliability assessment and risk evaluation prediction. Background Technology

[0002] Safety assessment originated in the United States as Risk Assessment. When it was introduced to Japan in the 1970s, Professor Inoue Takeyasu, vice president of the Japan Society for Safety Engineering, changed the name to Safety Assessment to avoid using the same character. my country adopted this terminology, hence it is also called risk assessment or hazard evaluation. Some also translate assessment as evaluation. Safety assessment aims to achieve system safety. It uses the principles and methods of safety systems engineering to identify and analyze the risk factors present in a system, determine the likelihood and severity of accidents and occupational hazards, and thus provide a scientific basis for developing preventative measures and management decisions.

[0003] Traditional equipment performance degradation and health assessment, risk assessment, and lifespan prediction are all based on probability density integral calculations using the exponential function of a normal process. However, with the increasing complexity of equipment and the development of computer technology, a simple normal distribution cannot best describe the high-precision requirements of equipment performance degradation and risk assessment. Furthermore, the development of neural network technology has further reduced the accuracy requirements of the model, and the increased computational burden caused by model complexity has also been resolved with the advancement of neural network technology. Based on these background factors, this invention proposes a method combining a hybrid distribution based on a quasi-normal process with neural networks. This method can better address the accuracy issues in performance degradation and risk assessment during high-risk operations of complex equipment and can better simulate the physical nature of equipment performance degradation. Therefore, this invention not only has significant theoretical value for risk assessment but also high practical value. Summary of the Invention

[0004] The technical problem to be solved by the present invention is: a risk assessment method for high-risk equipment operations based on a quasi-normal process, so as to solve the problem of insufficient accuracy in risk assessment of high-risk equipment operations using a normal distribution process in the above-mentioned prior art.

[0005] The technical solution adopted in this invention is: a risk assessment method for high-risk equipment operations based on a quasi-normal process, the method comprising the following steps: Step S10: Divide all equipment involved in the entire high-risk operation process into multiple sub-equipment; then monitor the status of each sub-equipment to obtain sub-equipment status performance monitoring data, which is recorded as follows. ,in ;in This refers to the total number of devices. This represents the total number of sub-devices for each device.

[0006] Step S20: Based on historical statistical data and engineering experience of the sub-equipment, set the mean and variance of equipment state performance based on a quasi-normal process; set the ideal value of the sub-equipment state, and then solve the sub-equipment state failure deviation value based on the aforementioned mean and variance of equipment state performance; then perform a combination of absolute value transformation and fractional power transformation to obtain the sub-equipment state deviation combination value; then perform an exponential transformation to obtain the quasi-normal decay value of sub-equipment state failure; introduce the equipment performance normal degradation ratio parameter, and then further solve the sub-equipment state degradation probability density function; finally, perform integration based on the time signal to solve the estimated value of sub-equipment state failure time.

[0007] Step S30: Set the constant parameters of the reliability model for each sub-device, introduce an exponential function, and obtain the first... The reliability density function of each sub-device; the first... The first piece of equipment The estimated failure time of each sub-device is used as the upper limit of integration. Integrating the reliability density function yields the first sub-device failure time. The first piece of equipment Risk assessment value of individual sub-devices.

[0008] Step S40, according to the above Risk assessment values ​​of all sub-equipment of the equipment, and Historical data from the overall risk assessment by senior experts during high-risk operations of equipment were used. 100 network node center values ​​and node interval values ​​were selected. A hybrid radial basis function (RBF) neural network was established to simulate the complex nonlinear impact of risks occurring in each sub-equipment on the overall risk assessment data. Based on the risk assessment values ​​of the sub-equipment, the network node center values, and the node interval values, normal exponential transformation and two types of hybrid radial basis function (HPF) transformations were performed. These were then multiplied by the network weights to obtain the network's normal radial basis function output and the outputs of the two types of HPF. These were then superimposed to obtain the total network output. Finally, the total network output data was compared with the overall risk assessment data from senior experts during high-risk operations to obtain the network training error data. Based on the network training error data, an adaptive weight adjustment law based on error feedback was designed. The neural network weights were iteratively trained through integral accumulation until the network error converged to a small region near 0, at which point the network weight update was stopped.

[0009] Step S50, based on the number to be evaluated The status performance monitoring data of the equipment is used to solve the problem in the same way. The first piece of equipment Estimate the failure time of each sub-device; then solve for the first... The first piece of equipment The risk assessment value of each sub-device is then input into the trained neural network to obtain the network's assessment of the risk of the first sub-device. The total output of risk assessment data for each piece of equipment; used as the total risk assessment score for high-risk operation processes of the equipment to be evaluated.

[0010] In one exemplary embodiment of the present invention, the sub-equipment state failure deviation value is calculated based on the mean and variance of the equipment state performance; then, a combination of absolute value transformation and fractional power transformation is performed to obtain the sub-equipment state deviation combination value; then, an exponential transformation is performed to obtain the sub-equipment state failure class normal decay value; a normal degradation ratio parameter of equipment performance is introduced, and then the sub-equipment state degradation probability density function is further solved; finally, the sub-equipment state failure time estimate is obtained by integrating the time signal, including: ; ; ; ; ; in For the first The ideal state value of each sub-device is a constant parameter. Indicates the first The first piece of equipment Individual device condition failure deviation value; Indicates the first The first piece of equipment Combined values ​​of state deviations of individual sub-devices; , , , It is a constant positive integer, and , ; , , These are constant combination parameters; Indicates a time signal. This represents the integral over the time signal; For the first The first piece of equipment Status and performance monitoring data of individual sub-devices; and For devices based on normal-like processes, the device number is... The mean and variance of the condition performance of each sub-device; these are constant values. For the first The first piece of equipment Normal decay value of the state failure type of each sub-device; It is an exponential function; For the first The first piece of equipment The state degradation probability density function of each sub-device; For the first The first piece of equipment Estimated failure time of individual sub-devices; For the first The performance degradation ratio of each sub-device; it is a constant parameter used to adjust the proportion of the normal decay factor in performance degradation.

[0011] In one exemplary embodiment of the present invention, constant parameters are set for the reliability model of each sub-device, and an exponential function is introduced to obtain the first... The reliability density function of each sub-device; the first... The first piece of equipment The estimated failure time of each sub-device is used as the upper limit of integration. Integrating the reliability density function yields the first sub-device failure time. The first piece of equipment The risk assessment values ​​for each sub-device include: ; ; in For the first The reliability density function of each sub-device; where and For constant model parameters, according to the first Reliability empirical data for each sub-device was selected; This is a constant parameter used to adjust the size of the distribution range of risk assessment values ​​for sub-equipment. For the first The first piece of equipment Risk assessment value of individual sub-equipment; In one exemplary embodiment of the present invention, according to the... Risk assessment values ​​of all sub-equipment of the equipment, and Historical data of the overall risk assessment by superior experts during high-risk operations of equipment were used. 100 network node center values ​​and node interval values ​​were selected. A hybrid radial basis function (RBF) neural network was established to simulate the complex nonlinear impact of risks occurring in each sub-equipment on the overall risk assessment data of equipment operations. Based on the risk assessment values ​​of the sub-equipment, the network node center values, and the node interval values, normal exponential transformation and two types of hybrid radial basis function (HPF) transformations were performed. These were then multiplied by the network weights to obtain the network's normal radial basis function output and the outputs of the two types of HPF transformations. These were then superimposed to obtain the total network output. Finally, the total network output data was compared with the overall risk assessment data of superior experts during high-risk operations of equipment to obtain network training error data. Based on the network training error data, an adaptive weight adjustment law based on error feedback was designed, and the neural network weights were iteratively trained through integral accumulation operations, including: ; ; ; ; ; ; ; ; ; ; ; in The network node center values ​​are selected in total, with a total of 100 network node center values ​​chosen. These are values ​​within a range of network nodes, all of which are constant parameters. For the network to the first The first piece of equipment The normal radial basis output of each sub-device; For the network to the first The first piece of equipment The first type of normal radial basis output of each sub-device; For the network to the first The first piece of equipment The second-type normal radial basis output of each sub-device; For the network to the first Total output of risk assessment data for each piece of equipment; For the first Risk assessment data from higher-level experts for high-risk operations of the equipment; This is network training error data; , , These are the weights of the neural network; , , For the adaptive weighting principle, It refers to the 100th network node. indivual; , , This is a constant parameter used to adjust the convergence speed of the neural network weights.

[0012] In one exemplary embodiment of the present invention, according to the first to be evaluated The status performance monitoring data of the equipment is used to solve the problem in the same way. The first piece of equipment Estimate the failure time of each sub-device; then solve for the first... The first piece of equipment The risk assessment value of each sub-device is then substituted into the trained neural network to obtain the network's assessment of the risk of the first sub-device. The total output of the risk assessment data for each piece of equipment includes: ; ; ; ; ; ; in For the first Status and performance monitoring data of the equipment; For the first The first piece of equipment Estimated cumulative time of individual device status; For the first The first piece of equipment Risk assessment value of individual sub-equipment; For the network to the first The first piece of equipment The normal radial basis output of each sub-device; For the network to the first The first piece of equipment The first type of normal radial basis output of each sub-device; For the network to the first The first piece of equipment The second-type normal radial basis output of each sub-device; For the network to the first The total output of the risk assessment data for each piece of equipment is the total risk assessment score for the high-risk operation process of the equipment to be evaluated.

[0013] Beneficial effects of the present invention Compared with existing technologies, the method employed in this invention has three major innovations. First, it extends the traditional methods for equipment condition assessment and residual design based on normal processes, proposing a hybrid exponential model based on normal and two types of quasi-normal processes, which can better adapt to the complex characteristics of equipment condition performance degradation. Second, it proposes a dedicated radial basis function neural network using normal and two types of quasi-normal exponential functions as basis functions, and designs an error feedback adaptive adjustment law; this better matches and adapts to the hybrid exponential model used in the analysis and calculation, thereby improving the accuracy of risk assessment, reducing the risk of weight divergence, and accelerating the network training and weight convergence process. Third, by monitoring and assessing the condition of all sub-equipment involved in high-risk equipment operation processes, and combining this with closed-loop calculation using neural networks, it utilizes the fuzzy parallel computing capabilities of neural networks, thereby reducing the accuracy requirements for the parameters of the normal and quasi-normal models of equipment performance degradation, while also ensuring high accuracy in the comprehensive risk assessment solution. Attached Figure Description

[0014] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention. It is obvious that the drawings described below are merely some embodiments of the invention, and those skilled in the art can obtain other drawings based on these drawings without any inventive effort.

[0015] Figure 1 This is a flowchart of a risk assessment method for high-risk equipment operations based on a quasi-normal process. Figure 2 This is a graph showing the convergence of neural network training error according to the method provided in this embodiment of the invention; Figure 3 This is the network of the method provided in the embodiments of the present invention. Convergence diagram of weighted data; Figure 4 This is the network of the method provided in the embodiments of the present invention. Convergence diagram of weighted data; Figure 5 This is the network of the method provided in the embodiments of the present invention. Chart showing the convergence of weighted data. Detailed Implementation

[0016] To make the objectives, technical solutions, and advantages of the present invention clearer, specific embodiments are described below, with reference to the appendix. Figure 1 The present invention will be further described in detail below.

[0017] Example 1: Using the status and performance monitoring data of 300 pieces of equipment stored in a warehouse during high-risk operations and the risk assessment data from higher-level experts to train the network; finally, the equipment to be evaluated is designated as equipment number 301; and the status and performance monitoring data of its sub-equipment is used to complete the risk assessment task, this example illustrates a risk assessment method for high-risk equipment operations based on a quasi-normal process. This method includes the following steps: Step S10: Divide all equipment involved in the entire high-risk operation process into 14 sub-equipment units; then monitor the status of all 14 sub-equipment units to obtain sub-equipment status performance monitoring data, which is recorded as follows: ,in ;in This refers to the total number of devices in this case study; This represents the total number of sub-devices for each device.

[0018] Step S20 can be broken down into the following five sub-steps. First, based on historical statistical data and engineering experience of the sub-equipment, set the mean and variance of the equipment state performance based on a quasi-normal process; set the ideal value of the sub-equipment state, and then calculate the sub-equipment state failure deviation value based on the aforementioned mean and variance of the equipment state performance as follows: ; in For the first The first piece of equipment Status and performance monitoring data of individual sub-devices; For the first The ideal state value of each sub-device is a constant parameter. and For devices based on normal-like processes The mean and variance of the condition performance of each sub-device; these are constant values. Indicates the first The first piece of equipment Individual device status failure deviation value.

[0019] The second step involves combining the absolute value change and fractional power transformation of the sub-device's state failure deviation value to obtain the following combined sub-device state deviation value: ; in Indicates the first The first piece of equipment Combined values ​​of state deviations of individual sub-devices; , , , It is a constant positive integer, and , ; , , For constant combination parameters; in this example, we choose . , , , .

[0020] The third step involves performing an exponential transformation based on the combined state deviation values ​​of the sub-devices to obtain the normal decay values ​​for sub-device state failures as follows: ; in For the first The first piece of equipment Normal decay value of the state failure type of each sub-device; It is an exponential function.

[0021] The fourth step involves introducing the normal degradation ratio parameter of equipment performance, and then further solving the probability density function of sub-equipment state degradation as follows: ; in For the first The first piece of equipment The state degradation probability density function of each sub-device.

[0022] Fifth, finally, integrate the time signal to obtain the estimated failure time of the sub-device, as follows: ; in Indicates a time signal. This represents the integral over the time signal; For the first The first piece of equipment Estimated failure time of individual sub-devices; For the first The performance degradation ratio of each sub-device; it is a constant parameter used to adjust the proportion of the normal decay factor in performance degradation.

[0023] Step S30 can be broken down into the following two sub-steps. First, set the constant parameters for the reliability model of each sub-device, introduce an exponential function, and obtain the... The reliability density function of each sub-device is as follows: ; Where the first The first piece of equipment The estimated failure time of each sub-device is used as the upper limit of integration, and the reliability density function is integrated to obtain the first sub-device failure time. The first piece of equipment The risk assessment values ​​for each sub-device are as follows: ; in For the first The reliability density function of each sub-device; where and For constant model parameters, it is based on the first... Reliability empirical data for each sub-device was selected; This is a constant parameter used to adjust the distribution range of risk assessment values ​​for sub-equipment; in this case, it is selected as... , For the first The first piece of equipment Risk assessment value of individual sub-devices.

[0024] Step S40 can be broken down into the following five sub-steps. Step 1, according to the aforementioned... The risk assessment values ​​of all sub-devices of the device are used to select the center values ​​of 100 network nodes and the node interval values. Based on the risk assessment values ​​of the sub-devices, the center values ​​of the network nodes, and the node interval values, a normal exponential transformation and two types of normal fractional exponential transformations are performed. Finally, these are multiplied by the network weights to obtain the network normal radial basis function output and the two types of network normal radial basis function outputs, as follows: ; ; ; in The network node center values ​​are selected in total, with a total of 100 network node center values ​​chosen. These are values ​​within a range of network nodes, all of which are constant parameters. For the network to the first The first piece of equipment The normal radial basis output of each sub-device; For the network to the first The first piece of equipment The first type of normal radial basis output of the individual device; For the network to the first The first piece of equipment The second type of normal radial basis output of the sub-device.

[0025] The second step involves establishing a quasi-normal hybrid radial basis function neural network to simulate the complex nonlinear impact of risks occurring in each sub-device on the overall risk assessment data of equipment operation. The total network output is obtained by superimposing the network's normal radial basis function output with the quasi-normal radial basis function outputs of the two types of networks: ; in For the network to the first The total output of risk assessment data for each piece of equipment.

[0026] The third step is to base our work on past history. The network training error data is obtained by comparing the overall risk assessment data from higher-level experts during high-risk operations of the equipment with the total network output data, i.e., the equipment risk assessment data, as follows: ; in For the first Risk assessment data from higher-level experts for high-risk operations of the equipment; This is the network training error data; the convergence of the network training error data is shown in the attached figure. Figure 2 As shown.

[0027] The fourth step is to design an adaptive weight adjustment rule based on error feedback, according to the network training error data, as follows: ; ; ; in , , For the adaptive weighting principle, It refers to the 100th network node. indivual; , , This is a constant parameter used to adjust the convergence speed of the neural network weights; , , In this case, for different Simply select the same value for all of them.

[0028] The fifth step involves iteratively training the neural network weights through integral accumulation operations as follows: ; ; ; in , , Here are the weights of the neural network, where The convergence of the weights is shown in the attached figure. Figure 3 As shown, The convergence of the weights is shown in the attached figure. Figure 4 As shown, The convergence of the weights is shown in the attached figure. Figure 5 As shown.

[0029] The sixth step is to stop training and stop updating the network weights once the network error converges to a small region near 0.

[0030] Step S50 can be broken down into the following two sub-steps. The first step is to determine the evaluation criteria based on the number of... The status performance monitoring data of the equipment is used to solve the problem in the same way. The first piece of equipment Estimate the failure time of each sub-device; then solve for the first... The first piece of equipment The risk assessment values ​​for each sub-device are as follows: ; ; in For the first Status and performance monitoring data of the equipment; For the first The first piece of equipment Estimated cumulative time of individual device status; For the first The first piece of equipment Risk assessment values ​​for individual sub-devices; in this case, exist The values ​​are as follows: [8.53, 9.21, 7.88, 8.34, 9.32, 9.54, 9.12, 8.92, 7.93, 8.64, 8.82, 9.17, 9.05, 8.75].

[0031] The second step is to... The risk assessment values ​​of all sub-devices of the device are fed into the trained neural network to obtain the network's assessment of the first... The total output of the risk assessment data for this equipment is as follows: ; ; ; ; in , For the network to the 301st device The normal radial basis output of each sub-device; For the network to the 301st device The first type of normal radial basis output of each sub-device; For the network to the 301st device The second-type normal radial basis output of each sub-device; This is the network's total output of risk assessment data for the 301st device, which is the total risk assessment score for the high-risk operation process of the device being evaluated; in this case... Therefore, the risk assessment value of the high-risk operation process of the 301st piece of equipment is 89.57, and the conclusion is that the condition is good and the risk level is low.

[0032] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for risk assessment of high-risk equipment operations based on a quasi-normal process, characterized by the following steps: Step S10: Divide all equipment involved in the entire high-risk operation process into multiple sub-equipment; then monitor the status of each sub-equipment to obtain sub-equipment status performance monitoring data, which is recorded as follows. ,in ;in This refers to the total number of devices. This represents the total number of all sub-devices for each device. Step S20: Based on the historical statistical data and engineering experience of the sub-equipment, set the mean and variance of the equipment state performance based on a normal-like process; The ideal values ​​for the sub-device status are set, and then the sub-device status failure deviation value is calculated based on the mean and variance of the device status performance. Next, a combination of absolute value transformation and fractional power transformation is performed to obtain the combined sub-device status deviation value. Then, an exponential transformation is performed to obtain the normal decay value of the sub-device status failure class. The normal degradation ratio parameter of device performance is introduced, and then the probability density function of sub-device status degradation is further solved. Finally, the time signal is integrated to obtain the estimated value of the sub-device status failure time as follows: ； ； ； ； ； in For the first The first piece of equipment Status and performance monitoring data of individual sub-devices; For the first The ideal state value of each sub-device is a constant parameter. Indicates the first The first piece of equipment Individual device condition failure deviation value; Indicates the first The first piece of equipment Combined values ​​of state deviations of individual sub-devices; , , , It is a constant positive integer, and , ; , , These are constant combination parameters; Indicates a time signal. This represents the integral over the time signal; For the first The first piece of equipment Status and performance monitoring data of individual sub-devices; and For devices based on normal-like processes The mean and variance of the condition performance of each sub-device; these are constant values. For the first The first piece of equipment Normal decay value of the state failure type of each sub-device; It is an exponential function; For the first The first piece of equipment The state degradation probability density function of each sub-device; For the first The first piece of equipment Estimated failure time of individual sub-devices; For the first The performance degradation ratio of each sub-device; it is a constant parameter used to adjust the proportion of the normal decay factor in performance degradation; Step S30: Set the constant parameters of the reliability model for each sub-device, introduce an exponential function, and obtain the first... The reliability density function of each sub-device; the first... The first piece of equipment The estimated failure time of each sub-device is used as the upper limit of integration. Integrating the reliability density function yields the first sub-device failure time. The first piece of equipment The risk assessment values ​​for each sub-device are as follows: ； ； in For the first The reliability density function of each sub-device; where and For constant model parameters, according to the first Reliability empirical data for each sub-device was selected; This is a constant parameter used to adjust the size of the distribution range of risk assessment values ​​for sub-equipment. For the first The first piece of equipment Risk assessment value of individual sub-equipment; Step S40, according to the above Risk assessment values ​​of all sub-equipment of the equipment, and Historical data of the risk assessment by higher-level experts during high-risk operations of the equipment were used to select the central values ​​and interval values ​​of 100 network nodes. A hybrid normal radial basis function (HBF) neural network was established to simulate the complex nonlinear impact of risks in each sub-equipment on the overall risk assessment data of equipment operation. Based on the risk assessment values ​​of the sub-equipment, the center values ​​of network nodes, and the node interval values, normal exponential transformation and two types of normal fractional exponential transformations were performed. These results were then multiplied by the network weights to obtain the network's normal radial basis function output and the outputs of the two types of network HBF. These were then superimposed to obtain the total network output. Finally, the total network output data was compared with the overall risk assessment data from higher-level experts during high-risk equipment operation processes to obtain the network training error data. Based on the network training error data, an adaptive weight adjustment rule based on error feedback is designed. The neural network weights are iteratively trained through integral accumulation operations until the network error converges to a small region near 0, at which point the network weight updates are stopped as follows: ； ； ； ； ； ； ； ； ； ； ； in The network node center values ​​are selected in total, with a total of 100 network node center values ​​chosen. These are values ​​within a range of network nodes, all of which are constant parameters. For the network to the first The first piece of equipment The normal radial basis output of each sub-device; For the network to the first The first piece of equipment The first type of normal radial basis output of each sub-device; For the network to the first The first piece of equipment The second-type normal radial basis output of each sub-device; For the network to the first Total output of risk assessment data for each piece of equipment; For the first Risk assessment data from higher-level experts for high-risk operations of the equipment; This is network training error data; , , These are the weights of the neural network; , , This is a weight adaptive law. It refers to the 100th network node. indivual; , , This is a constant parameter used to adjust the convergence speed of the neural network weights; Step S50, based on the number to be evaluated The status performance monitoring data of the equipment is used to solve the problem in the same way. The first piece of equipment Estimate the failure time of each sub-device; then solve for the first... The first piece of equipment The risk assessment value of each sub-device is then substituted into the trained neural network to obtain the network's assessment of the risk of the first sub-device. The total output of risk assessment data for each piece of equipment; the total risk assessment score for the high-risk operation process of the equipment to be evaluated is as follows: ； ； ； ； ； ； in For the first Status and performance monitoring data of the equipment; For the first The first piece of equipment Estimated cumulative time of individual device status; For the first The first piece of equipment Risk assessment value of individual sub-equipment; For the network to the first The first piece of equipment The normal radial basis output of each sub-device; For the network to the first The first piece of equipment The first type of normal radial basis output of each sub-device; For the network to the first The first piece of equipment The second-type normal radial basis output of each sub-device; For the network to the first The total output of the risk assessment data for each piece of equipment is the total risk assessment score for the high-risk operation process of the equipment to be evaluated.

Images