Fault prediction method based on cloud model and simple basic probability distribution
Through cloud model and a fault prediction method with simple basic probability distribution, the accuracy and adaptability of chemical equipment failure prediction are solved, and efficient fault diagnosis and prediction are achieved.
Patent Information
- Application Number
- CN202510468242.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-04-15
- Publication Date
- 2025-07-29
AI Technical Summary
Chemical equipment failure prediction has problems in the chemical industry, such as low prediction accuracy, poor model adaptability, easy sensor damage, unstable data quality and difficult to accurately diagnose under complex working conditions.
The cloud model is used for modeling, and the calculation process is simplified through simple basic probability distribution, and the failure probability fusion is combined with the D-S evidence fusion algorithm to output the fault category and level.
It improves the accuracy of intelligent diagnosis of chemical equipment faults, reduces the calculation amount, adapts to the complex and changing working conditions of chemical equipment, and improves the timeliness and accuracy of predictions.
Smart Images
Figure CN120387133A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of equipment fault prediction in the chemical industry, and particularly to a fault prediction method based on a cloud model and a simple basic probability distribution. Background Art
[0002] For the chemical industry, the stability of chemical equipment during operation can directly affect the benefits of chemical enterprises. The fault repair modes of chemical equipment mainly include preventive maintenance, condition-based maintenance, and breakdown maintenance. Regardless of the severity of the fault, the overall working efficiency of chemical enterprises will be affected to a certain extent during the fault repair period. If the fault is relatively serious, it will even affect the overall benefits of chemical enterprises. Therefore, for chemical enterprises, the repair mode is usually prevention-oriented and supplemented by maintenance. By regularly maintaining and servicing the equipment, the service life of chemical equipment can be effectively extended, and the safety performance of chemical equipment can be better ensured.
[0003] The chemical production process is usually continuous, and the cost of shutdown maintenance is high, so the requirements for the timeliness and accuracy of prediction are higher. The data of chemical equipment may be of high dimension, non-linear, and time-varying. Moreover, due to safety considerations, some key parameters may be difficult to monitor in real time. In addition, historical data may be insufficient, especially for new processes or new equipment, lacking sufficient fault samples, resulting in difficulties in training prediction models.
[0004] Chemical equipment is often in a harsh environment, such as high temperature, high pressure, corrosive gases or liquids, which will accelerate equipment aging and affect the life and data quality of sensors. The sensors themselves may be easily damaged or drifted in such an environment, resulting in inaccurate data.
[0005] Existing machine learning models may perform well under specific working conditions, but the chemical production conditions are variable, and the models may not be able to adapt to the changes in different working conditions, resulting in unstable prediction results. In addition, the multi-variable coupling relationship in the chemical process is complex, and traditional methods are difficult to capture these non-linear relationships. Traditional preventive maintenance may not be able to effectively cope with sudden faults, while predictive maintenance requires highly accurate model support. However, the complexity of chemical equipment makes it difficult for the model to cover all possible fault modes, resulting in missed alarms or false alarms. The chemical industry has extremely high safety requirements, and any prediction error may lead to serious consequences.
[0006] Therefore, proposing a fault prediction method based on a cloud model and a simple basic probability distribution to solve the difficulties existing in the prior art is an urgent problem to be solved by those skilled in the art. Summary of the Invention
[0007] In view of this, the present invention provides a fault prediction method based on cloud model and simple basic probability distribution, which solves the problem of low prediction accuracy of the intelligent diagnosis method for equipment faults in the chemical industry field. The cloud model is used for modeling, and the calculation process and fusion process of the basic probability distribution are greatly simplified through simple basic probability, thereby greatly reducing the amount of calculation and improving the accuracy of intelligent diagnosis of equipment faults.
[0008] To achieve the above object, the present invention adopts the following technical solutions:
[0009] A fault prediction method based on cloud model and simple basic probability distribution, comprising the following steps:
[0010] S1. Obtain data: Obtain equipment fault feature data;
[0011] S2. Data preprocessing: Preprocess the obtained equipment fault feature data to obtain a preprocessed fault feature data set;
[0012] S3. Model establishment: Establish a fault cloud model for each type of fault according to the fault feature data set;
[0013] S4. Probability result: Determine the basic probability assignment under different faults based on simple basic probability distribution, and use the D-S evidence fusion algorithm to fuse the basic probability assignment to obtain the final fused fault probability result;
[0014] S5. Result output: In the final fused fault probability result, select the fault state corresponding to the maximum basic probability as the diagnosis result, and output the fault category and fault level.
[0015] Optionally, the equipment fault feature data obtained in S1 includes temperature feature data, pressure feature data, and vibration feature data.
[0016] Optionally, in S2, the obtained equipment fault feature data is subjected to data cleaning and outlier removal, data normalization, denoising, and data enhancement preprocessing to obtain a preprocessed fault feature data set.
[0017] Optionally, the specific content of establishing a fault cloud model for each type of fault in S3 is as follows:
[0018] S31. Classify the preprocessed fault feature data according to fault types, and each type of data corresponds to a fault mode;
[0019] S32. Define the qualitative concept of the fault type, set semantic labels, and generate corresponding fault cloud models according to the semantic labels respectively;
[0020] S33. Based on the corresponding fault cloud model, output the fault type result.
[0021] Optionally, the semantic tags set in S32 include temperature characteristic data, pressure characteristic data, and vibration characteristic data. Cloud models between temperature characteristics and fault types, between pressure characteristics and fault types, and between vibration characteristics and fault types are generated respectively according to the semantic tags.
[0022] Optionally, the specific content of determining the basic probability assignment under different faults based on the simple basic probability distribution in S4 and using the D-S evidence fusion algorithm to fuse the basic probability assignment to obtain the final fused fault probability result is as follows:
[0023] Assume the current fault type set, θ = {fault 1, fault 2, fault 3};
[0024] For a certain feature, for example, the probability value a is calculated in the cloud model of fault 1 corresponding to the temperature feature. Then the basic probability distribution obtained for fault 1 under the corresponding temperature feature is:
[0025] m(fault 1) = a, m({fault 1, fault 2, fault 3}) = 1 - a;
[0026] Fuse the basic probability distributions of different faults under the temperature feature, and repeat the above steps for different faults to obtain the probability distributions of different fault types under the temperature feature;
[0027] Repeat the above steps to fuse the basic probability distributions of different faults under different features to obtain the final fused fault probability result.
[0028] Optionally, the fault levels output in S5 include four levels: normal state, potential fault, minor fault, and serious fault;
[0029] During the normal state, regular inspections are carried out; during the potential fault, predictive maintenance is carried out; during the minor fault, planned shutdown for maintenance is carried out; during the serious fault, immediate shutdown and notification for maintenance are carried out.
[0030] It can be seen from the above technical solutions that compared with the prior art, the present invention provides a fault prediction method based on cloud models and simple basic probability distributions, which has the following beneficial effects:
[0031] (1) For each feature under each fault of equipment faults in the chemical industry field, the present invention corresponds a cloud model. Then the cloud model can calculate a simple basic probability distribution through samples, and fuse the simple basic probability distributions under all features through the DS evidence fusion algorithm, so as to finally obtain the probability of the corresponding fault;
[0032] (2) When the present invention performs intelligent diagnosis on equipment failures in the chemical industry field, a cloud model is used for modeling, making the model closer to the complex and changeable actual on-site environment. At the same time, the calculation process and fusion process of the basic probability distribution are greatly simplified through simple basic probability, thereby greatly reducing the calculation amount and improving the accuracy of intelligent diagnosis of equipment failures. BRIEF DESCRIPTION OF THE DRAWINGS
[0033] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings described below are only the embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained according to the provided drawings.
[0034] Figure 1 It is a flowchart of a fault prediction method based on a cloud model and a simple basic probability distribution provided by the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS
[0035] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.
[0036] Refer to Figure 1 As shown, the present invention discloses a fault prediction method based on a cloud model and a simple basic probability distribution, including the following steps:
[0037] S1. Obtain data: Obtain equipment fault characteristic data;
[0038] S2. Data preprocessing: Preprocess the obtained equipment fault characteristic data to obtain a preprocessed fault characteristic data set;
[0039] S3. Model establishment: Establish a fault cloud model for each type of fault according to the fault characteristic data set;
[0040] S4. Probability result: Determine the basic probability assignment under different faults based on the simple basic probability distribution, and use the D-S evidence fusion algorithm to fuse the basic probability assignment to obtain the final fused fault probability result;
[0041] S5. Result output: In the final fused fault probability result, select the fault state corresponding to the maximum basic probability as the diagnosis result, and output the fault category and fault level.
[0042] Further, the device fault feature data obtained in S1 includes temperature feature data, pressure feature data, and vibration feature data.
[0043] Further, in S2, the obtained device fault feature data is subjected to data cleaning and outlier removal, data normalization, denoising, and data augmentation preprocessing to obtain a preprocessed fault feature data set.
[0044] Further, the specific content of establishing the fault cloud model for each type of fault in S3 based on the fault feature data set is as follows:
[0045] S31. Classify the preprocessed fault feature data by fault type, and each type of data corresponds to a fault mode.
[0046] S32. Define the qualitative concept of the fault type, set semantic labels, and generate corresponding fault cloud models according to the semantic labels respectively.
[0047] S33. Based on the corresponding fault cloud model, output the fault type result.
[0048] Further, the semantic labels set in S32 include temperature feature data, pressure feature data, and vibration feature data. Cloud models between temperature features and fault types, cloud models between pressure features and fault types, and cloud models between vibration features and fault types are generated respectively according to the semantic labels.
[0049] Specifically, the fault types include bearing faults, reactor faults, container faults, etc. The fault cloud models include cloud models between temperature features and fault types, including cloud models between temperature features and bearing faults, cloud models between temperature features and reactor faults, and cloud models between temperature features and container faults;
[0050] Cloud models between pressure features and fault types include cloud models between pressure features and bearing faults, cloud models between pressure features and reactor faults, and cloud models between pressure features and container faults;
[0051] Cloud models between vibration features and fault types include cloud models between vibration features and bearing faults, cloud models between vibration features and reactor faults, and cloud models between vibration features and container faults.
[0052] Further, the specific content of determining the basic probability assignment under different faults based on the simple basic probability distribution and using the D-S evidence fusion algorithm to fuse the basic probability assignment to obtain the final fused fault probability result in S4 is as follows:
[0053] Assume the current fault type set, θ = {fault 1, fault 2, fault 3};
[0054] For a certain feature, such as the probability value a calculated in the cloud model of fault 1 corresponding to the temperature feature, the basic probability distribution of fault 1 under the corresponding temperature feature is as follows:
[0055] m(Fault 1) = a, m({Fault 1, Fault 2, Fault 3}) = 1 - a;
[0056] Fuse the basic probability distributions of different faults under the temperature feature, and repeat the above steps for different faults to obtain the probability distributions of different fault types under the temperature feature;
[0057] Repeat the above steps, fuse the basic probability distributions of different faults under different features, and obtain the final fused fault probability result.
[0058] Furthermore, the fault levels output in S5 include four levels: normal state, potential fault, minor fault, and serious fault;
[0059] During the normal state, conduct regular inspections; during potential faults, conduct predictive maintenance; during minor faults, conduct planned shutdown for maintenance; during serious faults, immediately shut down and notify for maintenance.
[0060] In a specific embodiment, it includes the following content:
[0061] Obtain the temperature feature data, pressure feature data, and vibration feature data of the device;
[0062] Perform data cleaning, outlier removal, data normalization, denoising, and data augmentation preprocessing on the obtained device fault feature data to obtain the preprocessed fault feature dataset;
[0063] Classify the preprocessed fault feature data by fault type, and each type of data corresponds to a fault mode;
[0064] Define the qualitative concept of the fault type, set semantic labels, and the set semantic labels include temperature feature data, pressure feature data, and vibration feature data. Generate the cloud model between the temperature feature and the fault type, the cloud model between the pressure feature and the fault type, and the cloud model between the vibration feature and the fault type according to the semantic labels;
[0065] Based on the corresponding fault cloud model, output the fault type result;
[0066] The cloud model between the temperature feature and the fault type, the cloud model between the pressure feature and the fault type, and the cloud model between the vibration feature and the fault type obtain a simple basic probability distribution through sample calculation. Through the DS evidence fusion algorithm, fuse the simple basic probability distributions under all features to obtain the final corresponding fault probability;
[0067] In the final fused fault probability result, select the fault state corresponding to the maximum basic probability as the diagnosis result, and output the fault category and fault level.
[0068] The fault levels include four levels: normal state, potential fault, minor fault, and severe fault.
[0069] During the normal state, conduct regular inspections; during the potential fault state, conduct predictive maintenance; during the minor fault state, conduct planned shutdown for maintenance; during the severe fault state, immediately shut down and notify for maintenance.
[0070] The various embodiments in this specification are described in a progressive manner. Each embodiment focuses on the differences from other embodiments. For the same or similar parts among the various embodiments, reference can be made to each other.
[0071] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present invention. Various modifications to these embodiments will be obvious to those skilled in the art. The general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to these embodiments shown herein, but rather will be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims
1. A fault prediction method based on cloud model and simple basic probability distribution, characterized in that It includes the following steps: S1. Obtain data: Obtain device fault feature data; S2. Data preprocessing: Preprocess the obtained device fault feature data to obtain a preprocessed fault feature data set; S3. Model establishment: Establish a fault cloud model for each type of fault based on the fault feature data set; S4. Probability result: Determine the basic probability assignment under different faults based on a simple basic probability distribution, and use the D-S evidence fusion algorithm to fuse the basic probability assignment to obtain the final fused fault probability result; S5. Result output: In the final fused fault probability result, select the fault state corresponding to the maximum basic probability as the diagnosis result, and output the fault category and fault level.
2. A fault prediction method based on a cloud model and a simple basic probability distribution according to claim 1, wherein the device fault feature data obtained in S1 includes temperature feature data, pressure feature data, and vibration feature data.
3. A fault prediction method based on a cloud model and a simple basic probability distribution according to claim 1, wherein in S2, the obtained device fault feature data is subjected to data cleaning and outlier removal, data normalization, denoising, and data augmentation preprocessing to obtain a preprocessed fault feature data set.
4. A fault prediction method based on a cloud model and a simple basic probability distribution according to claim 1, wherein the specific content of establishing a fault cloud model for each type of fault based on the fault feature data set in S3 is as follows: S31. Classify the preprocessed fault feature data by fault type, and each type of data corresponds to a fault mode; S32. Define the qualitative concept of the fault type, set semantic labels, and generate corresponding fault cloud models according to the semantic labels respectively; S33. Based on the corresponding fault cloud model, output the fault type result.
5. A fault prediction method based on a cloud model and a simple basic probability distribution according to claim 4, wherein the semantic labels set in S32 include temperature feature data, pressure feature data, and vibration feature data, and cloud models between temperature features and fault types, between pressure features and fault types, and between vibration features and fault types are generated according to the semantic labels respectively.
6. A fault prediction method based on a cloud model and a simple basic probability distribution according to claim 1, wherein the specific content of determining the basic probability assignment under different faults based on a simple basic probability distribution and using the D-S evidence fusion algorithm to fuse the basic probability assignment to obtain the final fused fault probability result in S4 is as follows: Assume the current set of fault types, θ = {fault 1, fault 2, fault 3}; For a certain feature, for example, calculate the corresponding probability value a in the fault 1 cloud model corresponding to the temperature feature, then the basic probability distribution obtained for fault 1 under the corresponding temperature feature is: m(fault 1) = a, m({fault 1, fault 2, fault 3}) = 1 - a; Fuse the basic probability distributions of different faults for the temperature feature, and repeat the above steps for different faults to obtain the probability distributions of different fault types under the temperature feature; Repeat the above steps to fuse the basic probability distributions of different faults under different features, and obtain the final fused fault probability result.
7. A fault prediction method based on a cloud model and a simple basic probability distribution according to claim 1, characterized in that The fault levels output in S5 include four levels: normal state, potential fault, minor fault, and serious fault; During the normal state, regular inspections are carried out; during potential faults, predictive maintenance is carried out; during minor faults, planned shutdown and maintenance are carried out; during serious faults, immediate shutdown is carried out and maintenance is notified.