Online abnormity monitoring method and system for linear moving knife edge ore pulp sampler

By constructing statistical features and unsupervised algorithms for abnormal monitoring, combined with feature selection and stability indicators, the problem of missing measurement methods and calibration of linear mobile cutting edge slurry samplers is solved, and high-precision online abnormal monitoring is achieved.

CN119000173AInactive Publication Date: 2024-11-22BEIJING INST OF METROLOGY & TESTING SCI

Patent Information

Application Number
CN202411061560.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-05
Publication Date
2024-11-22
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

The existing linear moving cutting edge slurry samplers lack standard measurement methods and cannot be traced. The accuracy is poor due to the complex components of the ore slurry, and the calibration of continuous and uninterrupted operation is difficult to meet the application scenarios and working conditions.

Method used

By collecting time series data for preprocessing, building statistical features, using unsupervised algorithms for abnormal monitoring, and determining abnormal tags through most voting strategies, improving monitoring performance based on feature selection, correcting tags, and using stability indicators PSI to quantify offline and online evaluation The distribution difference of samples is optimized for stability.

Benefits of technology

It realizes effective identification of abnormal points during the slurry sampling process, improves monitoring accuracy, adapts to complex components of the slurry, ensures high-precision monitoring, and solves the problem of insufficient monitoring accuracy.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119000173A_ABST
    Figure CN119000173A_ABST
Patent Text Reader

Abstract

The invention discloses an online abnormity monitoring method and system for a linear moving knife edge ore pulp sampler, and relates to the technical field of industrial automation monitoring, and the method comprises the steps: collecting time sequence data of different monitoring items, and carrying out the data preprocessing; the method comprises the following steps: constructing statistical features in an offline learning stage, carrying out anomaly monitoring through an unsupervised algorithm, determining an abnormal label through a majority voting strategy, improving the monitoring performance of the unsupervised algorithm based on feature selection, correcting the label and training a model; and the on-line detection learning segment quantitatively evaluates the distribution difference between the off-line sample and the on-line sample through a stability index PSI so as to realize stability optimization. According to the online anomaly monitoring method for the linear moving knife edge ore pulp sampler, the anomaly label is determined through five kinds of unsupervised learning and the majority voting strategy in the offline learning stage, abnormal points in the ore pulp sampling process are effectively recognized, the method can adapt to complex components of ore pulp, and the monitoring precision is improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the technical field of industrial automation monitoring, and in particular to an online abnormality monitoring method and system for a linear moving knife-edge slurry sampler. Background Art

[0002] Ore dressing is a continuous process industry that enriches low-grade raw ores through flotation, gravity separation, magnetic separation and other methods. In the process of ore dressing, taking out representative pulp samples regularly and quantitatively is an important means and basis for guiding production and determining relevant data of the process flow, and it is also the bottleneck of online particle size, concentration and grade analysis and monitoring. The linear moving knife-edge pulp sampler refers to an automated sampling device used to take out representative pulp samples from the process pulp flow. The principle is that the pulp flow enters the sampling box from the horizontal inlet to form a parabolic waterfall surface or enters the sampling box from the vertical inlet to form a free fall. The internal design of the linearly movable sampling knife edge is driven by a cylinder or a motor to reciprocate and uniformly cut all sections of the pulp flow. The sampler is designed with box flushing water and knife edge flushing water to clean the residual samples before and after sampling.

[0003] Since the introduction of linear moving knife-edge slurry sampler technology, it has been promoted and applied to many mining enterprises such as copper mines, lead-zinc mines, gold mines, and uranium mines for process or online analytical instrument sampling, which has injected new concepts into the advancement of slurry sampling technology and provided new products and technologies. However, at present, there are still no relevant standards for linear moving knife-edge slurry samplers in China. Although this technology has been introduced for many years and has been put into use on a large scale in industrial production, the evaluation of linear moving knife-edge slurry samplers still has problems such as unclear technical indicators and unclear technical requirements. In order to standardize the use of linear moving knife-edge slurry samplers and promote the unified standards of the mining and metallurgical industry as soon as possible, the monitoring of linear moving knife-edge slurry samplers currently has the following difficulties:

[0004] 1. The linear moving knife-edge slurry sampler lacks a measurement method, and the actual measured value lacks a reference standard value and cannot be traced;

[0005] 2. The composition of the ore pulp is complex, and there are many factors affecting the measurement of ion concentration, so the actual value is quite different from the measured value;

[0006] 3. The linear moving knife-edge slurry sampler instrument usually runs continuously and uninterruptedly, and the calibration should be consistent with the application scenario and working conditions.

[0007] In order to solve the above problems, the present invention proposes an online abnormality monitoring method and system for a linear moving knife-edge slurry sampler. Summary of the invention

[0008] In view of the above-mentioned problems, the present invention is proposed.

[0009] Therefore, the technical problem solved by the present invention is: the existing evaluation method of linear moving knife-edge slurry sampler lacks a standard measurement method and cannot be traced, is affected by the complex composition of the slurry and has poor accuracy, and how to achieve continuous and uninterrupted operation, and the calibration should conform to the application scenarios and working conditions.

[0010] In order to solve the above technical problems, the present invention provides the following technical solutions: an online anomaly monitoring method for a linear moving knife-edge slurry sampler, comprising collecting time series data of different monitoring items and performing data preprocessing. Statistical features are constructed in the offline learning stage, anomaly monitoring is performed through an unsupervised algorithm, and anomaly labels are determined through a majority voting strategy. The monitoring performance of the unsupervised algorithm is improved based on feature selection, and the labels are corrected and the model is trained. In the online learning stage, the stability index PSI is used to quantitatively evaluate the distribution difference between offline samples and online samples to achieve stability optimization.

[0011] As a preferred solution of the online abnormal monitoring method for the linear moving knife-edge slurry sampler described in the present invention, wherein: the time series data of different monitoring items are collected and data preprocessing includes extracting the mobile knife-edge slurry sampler online monitoring system data set D, collecting time series data, and the linear moving knife-edge slurry samplers installed in different positions have different monitoring items. The model features are constructed with the maximum set to adapt to all linear moving knife-edge slurry sampler monitoring items.

[0012] Perform data quality checks, including data type, control records, missing rate, number of records with null values, minimum value, 1st percentile, 5th percentile, 25th percentile, 50th percentile, 75th percentile, 95th percentile, 99th percentile, maximum value, mean, and variance statistics.

[0013] The data collection frequency for all monitoring projects is 30 minutes. The data set contains timestamp indicators, linear moving knife-edge slurry sampler models and monitoring projects, and a data set library suitable for all linear moving knife-edge slurry samplers is constructed.

[0014] Clean the collected time series data, fill it with the previous non-empty value, and delete the records that are still empty after filling.

[0015] Calculate the Pearson correlation coefficient for each pair of features and remove features with a correlation coefficient greater than 0.7.

[0016] As a preferred solution of the online abnormality monitoring method of the linear moving knife-edge slurry sampler described in the present invention, wherein: the construction of statistical features in the offline learning stage includes performing feature construction on the retained features, including a statistical feature set, a time feature set and a time series feature set.

[0017] The construction of the statistical feature set includes calculating the skewness and kurtosis of the data, and combining it with the mean and variance that have already been calculated.

[0018] The construction of the time feature set includes the construction based on the time interval of data collection and the periodicity of data, and the use of Fourier transform to extract the periodic components in the data, which is expressed as:

[0019]

[0020] Among them, P j (f) is the periodic component of the jth feature at frequency f, p ij is the jth eigenvalue of the ith sample, t i Is the timestamp.

[0021] The construction of the time series feature set includes representing the correlation of the time series at different lag periods based on the autocorrelation function, which is expressed as:

[0022]

[0023] Among them, α is a smoothing parameter, which is used to control the smoothness of the autocorrelation function. is the smoothed autocorrelation coefficient of the jth feature at lag period k, where k is the lag period, μ j is the mean, p i j is the j-th eigenvalue of the ith sample.

[0024] As a preferred solution of the online anomaly monitoring method of the linear moving knife-edge slurry sampler described in the present invention, wherein: the anomaly monitoring is performed by an unsupervised algorithm, and the anomaly label is determined by a majority voting strategy, including calculating the anomaly label of each algorithm by 5 unsupervised learning algorithms.

[0025] The five unsupervised learning algorithms include ARIMA statistical model, Isolation Forest machine learning model, support vector machine, density clustering algorithm and principal component analysis.

[0026] After completing the analysis, a majority voting strategy is used to obtain anomaly labels.

[0027] As a preferred solution of the online abnormality monitoring method of the linear moving knife-edge slurry sampler described in the present invention, the feature selection-based improvement of the monitoring performance of the unsupervised algorithm includes eliminating the dimension through standardization processing, adopting the feature selection method of the wrapping method, the embedding method and the filtering method, and the wrapping method selects features through recursive feature elimination, which is expressed as:

[0028]

[0029] Among them, S is the feature subset, f is the unsupervised algorithm, and X S is the data corresponding to the feature subset, θ is the algorithm parameter, L is the loss function, y is the true label, and N is the number of samples.

[0030] When the unsupervised learning algorithm is the ARIMA statistical model, the loss function is expressed as:

[0031]

[0032] Among them, X t is the actual time series value, To predict the time series value, N′ is the length of the time series.

[0033] When the unsupervised learning algorithm is the Isolation Forest machine learning model, the loss function is expressed as:

[0034]

[0035] Among them, E(h(x i )) is the expected value of the path length of the i-th sample in the tree, x i is the i-th sample, and c(n) is the normalization constant.

[0036] When the unsupervised learning algorithm is a support vector machine, the loss function is expressed as:

[0037]

[0038] Among them, w is the weight vector, ξ is the slack variable, ρ is the bias, and ν is the penalty parameter.

[0039] When the unsupervised learning algorithm is a density clustering algorithm, the loss function is expressed as:

[0040]

[0041] Where 1 is the indicator function, the value is 1 when the condition is true, otherwise it is 0, x′ is the data point, x i ′ is the original data point, ∈ is the clustering radius, density(x i ) is the density of the ith data point, and minPts is the minimum density threshold.

[0042] When the unsupervised learning algorithm is principal component analysis, the loss function is expressed as:

[0043]

[0044] in, is the reconstructed data point.

[0045] The embedding method uses regularized Lasso regression for feature selection, expressed as:

[0046]

[0047] Among them, β is the regression coefficient, p is the characteristic number, and x ij is the jth feature of the i-th sample after standardization, y i is the label of the i-th sample after standardization, and λ is the regularization parameter.

[0048] The filtering method uses the chi-square test to select features, expressed as:

[0049]

[0050] Among them, χ 2 is the chi-square statistic, O i is the observed value, E i is the expected value.

[0051] The anomaly scores obtained by the five algorithms are weighted and fused according to the coefficients of 0.2:0.3:0.2:0.1:0.2, and the anomaly scores are divided into the interval of [0,100] using the minimum and maximum normalization method.

[0052] As a preferred solution of the online abnormality monitoring method of the linear moving knife-edge slurry sampler described in the present invention, wherein: the correction label includes correction based on the typical monitoring range of the industry, setting the monitoring technical requirements, the sampling time interval is allowed to differ by 10s, the sampling cycle is allowed to differ by 2s, the sampling frequency is the same as the set sampling frequency, the sampling knife edge movement speed is allowed to differ by 0.05m / s, the sampling knife edge cavity volume is allowed to differ by 10%, the sampling knife edge cap opening is dynamically adjusted, when the maximum mineral particle diameter in the slurry is ≥2mm, the opening is ≥3 times the maximum particle diameter, when the maximum mineral particle diameter in the slurry is <2mm, the opening is ≥6mm, the sampling volume error is allowed to differ by 5%, and the sampling volume repeatability is allowed to have 5%.

[0053] As a preferred solution of the online abnormality monitoring method of the linear moving knife-edge slurry sampler described in the present invention, wherein: the online monitoring learning segment quantitatively evaluates the distribution difference between offline samples and online samples through the stability index PSI to achieve stability optimization, including using pickle to save the results of offline data processing, feature engineering, and model training, and providing online abnormality monitoring in the form of a Restful API interface after going online, using the stability index PSI to quantitatively evaluate the distribution difference between offline samples and online samples, and statistically analyzing the degree of change of the distribution of each risk category across time windows. If the degree of change is less than 10%, there is no need to update the model. If the degree of change is between 10% and 20%, the cause of the change is identified, and an alarm is sent to the operation and maintenance personnel. If the degree of change is greater than 20%, the model is iterated until the degree of change is maintained at less than 10%.

[0054] Another object of the present invention is to provide an online abnormality monitoring system for a linear moving knife-edge slurry sampler, which can eliminate dimensions through standardized processing, and use feature selection methods such as wrapping, embedding and filtering to perform different processing on five unsupervised algorithms. While optimizing, the characteristics of the model are also taken into account, so that the monitoring performance is significantly improved while ensuring high-precision monitoring, solving the problem of insufficient monitoring accuracy in the current evaluation method of the linear moving knife-edge slurry sampler.

[0055] As a preferred solution of the online anomaly monitoring system of the linear moving knife-edge slurry sampler described in the present invention, it includes: a data acquisition module, an offline learning module, and an online learning module. The data acquisition module is used to collect time series data of different monitoring items and perform data preprocessing. The offline learning module is used to construct statistical features in the offline learning stage, perform anomaly monitoring through an unsupervised algorithm, and determine the anomaly label through a majority voting strategy, improve the monitoring performance of the unsupervised algorithm based on feature selection, correct the label and train the model. The online learning module is used to quantitatively evaluate the distribution difference between offline samples and online samples through the stability index PSI in the online learning stage to achieve stability optimization.

[0056] A computer device comprises a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement a step of an online abnormality monitoring method for a linear moving knife-edge slurry sampler.

[0057] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of an online abnormality monitoring method for a linear moving knife-edge slurry sampler.

[0058] Beneficial effects of the present invention: The online abnormality monitoring method of the linear moving knife-edge slurry sampler provided by the present invention constructs statistical features in the offline learning stage, performs abnormality monitoring through an unsupervised algorithm, and determines abnormal labels through a majority voting strategy. By constructing statistical features in the offline learning stage, it is possible to effectively identify abnormal points in the slurry sampling process, adapt to the complex composition of the slurry, improve monitoring accuracy, eliminate dimensions through standardized processing, and use the feature selection methods of wrapping, embedding and filtering to perform different processing on five unsupervised algorithms. While optimizing, the characteristics of the model are also considered, so that the monitoring performance is significantly improved while ensuring high-precision monitoring. The construction of multiple feature sets ensures the effectiveness and reliability of characterizing slurry data. The present invention achieves better results in terms of recognition accuracy and adaptability to slurry. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0060] Figure 1 An overall flow chart of an online abnormality monitoring method for a linear moving knife-edge slurry sampler provided in the first embodiment of the present invention.

[0061] Figure 2 This is an overall flow chart of an online abnormality monitoring system for a linear moving knife-edge slurry sampler provided in accordance with a third embodiment of the present invention. DETAILED DESCRIPTION

[0062] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0063] Example 1, reference Figure 1 , as one embodiment of the present invention, provides an online abnormality monitoring method for a linear moving knife-edge slurry sampler, comprising:

[0064] S1: Collect time series data of different monitoring projects and perform data preprocessing.

[0065] Furthermore, time series data of different monitoring projects are collected, and data preprocessing is performed, including extracting the mobile knife-edge slurry sampler online monitoring system data set D, collecting time series data, and the data may include temperature, pressure, flow, pH value, etc. The linear mobile knife-edge slurry samplers installed in different locations have different monitoring projects. The model features are constructed with the maximum set to adapt to all linear mobile knife-edge slurry sampler monitoring projects.

[0066] Perform data quality checks, including data type, control records, missing rate, number of records with null values, minimum value, 1st percentile, 5th percentile, 25th percentile, 50th percentile, 75th percentile, 95th percentile, 99th percentile, maximum value, mean, and variance statistics.

[0067] The data collection frequency for all monitoring items is 30 minutes. The data set contains timestamp indicators, linear moving knife-edge slurry sampler models and monitoring items. The monitoring items include sampling cycle, sampling frequency, sampling knife-edge movement speed, etc., and a data set library suitable for all linear moving knife-edge slurry samplers is constructed.

[0068] Clean the collected time series data, fill it with the previous non-empty value, and delete the records that are still empty after filling.

[0069] Calculate the Pearson correlation coefficient for each pair of features and remove features with a correlation coefficient greater than 0.7.

[0070] S2: Construct statistical features in the offline learning stage, perform anomaly monitoring through unsupervised algorithms, determine anomaly labels through majority voting strategies, improve the monitoring performance of unsupervised algorithms based on feature selection, correct labels and train models.

[0071] Furthermore, constructing statistical features in the offline learning stage includes performing feature construction on the retained features, including a statistical feature set, a time feature set, and a time series feature set.

[0072] The construction of the statistical feature set includes calculating the skewness and kurtosis of the data, and combining it with the mean and variance that have already been calculated.

[0073] The construction of the time feature set includes the construction based on the time interval of data collection and the periodicity of data, and the use of Fourier transform to extract the periodic components in the data, which is expressed as:

[0074]

[0075] Among them, P j (f) is the periodic component of the jth feature at frequency f, p ij is the jth eigenvalue of the ith sample, t i Is the timestamp.

[0076] The construction of the time series feature set includes representing the correlation of the time series at different lag periods based on the autocorrelation function, which is expressed as:

[0077]

[0078] Among them, α is a smoothing parameter, which is used to control the smoothness of the autocorrelation function. is the smoothed autocorrelation coefficient of the jth feature at lag period k, where k is the lag period, μ j is the mean, p i j is the j-th eigenvalue of the ith sample.

[0079] It should be noted that anomaly monitoring is performed by an unsupervised algorithm and anomaly labels are determined by a majority voting strategy, including calculating anomaly labels of each algorithm by five unsupervised learning algorithms.

[0080] The five unsupervised learning algorithms include ARIMA statistical model, Isolation Forest machine learning model, support vector machine, density clustering algorithm and principal component analysis.

[0081] After completing the analysis, a majority voting strategy is used to obtain anomaly labels.

[0082] It should also be noted that the correction label includes corrections based on the typical monitoring range of the industry and setting monitoring technical requirements. As shown in Table 1, the sampling time interval is allowed to differ by 10s, the sampling cycle is allowed to differ by 2s, the sampling frequency is the same as the set sampling frequency, the sample blade movement speed is allowed to differ by 0.05m / s, the sampling blade cavity volume is allowed to differ by 10%, and the sampling blade cap opening is dynamically adjusted. When the maximum mineral particle diameter in the slurry is ≥2mm, the opening is ≥3 times the maximum particle diameter. When the maximum mineral particle diameter in the slurry is <2mm, the opening is ≥6mm. The sampling volume error is allowed to differ by 5%, and the sampling volume repeatability is allowed to be 5%.

[0083] Table 1 Parameter range table

[0084]

[0085]

[0086] S3: In the online learning phase, the stability index PSI is used to quantitatively evaluate the distribution difference between offline samples and online samples to achieve stability optimization.

[0087] Furthermore, the online inspection phase uses the stability index PSI to quantitatively evaluate the distribution difference between offline samples and online samples to achieve stability optimization, including using pickle to save the results of offline data processing, feature engineering, and model training, and providing online anomaly monitoring in the form of a Restful API interface after going online. The stability index PSI is used to quantitatively evaluate the distribution difference between offline samples and online samples, and the degree of change of the distribution of each risk category across time windows is statistically analyzed. If the degree of change is less than 10%, there is no need to update the model. If the degree of change is between 10% and 20%, the cause of the change is identified and an alarm is sent to the operation and maintenance personnel. If the degree of change is greater than 20%, the model is iterated until the degree of change is maintained at less than 10%.

[0088] Example 2 is an embodiment of the present invention, which provides an online abnormality monitoring method for a linear moving knife-edge slurry sampler. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculation and simulation experiments.

[0089] First, the test was conducted in a copper mining enterprise, and three linear moving knife-edge slurry samplers at different positions were selected, namely sampler A and sampler B. The purpose of the test was to compare the differences between the prior art and the method of the present invention through data analysis in the online monitoring and offline learning stages, thereby reflecting the innovation and advantages of the present invention.

[0090] An online monitoring system is installed on samplers A, B and C to collect time series data every 30 minutes. The monitoring items include slurry concentration (%), slurry particle size (μm), slurry flow rate (m / s) and slurry grade (%).

[0091] Table 2 Experimental data comparison table

[0092]

[0093]

[0094] There are many abnormal labels in the prior art. For example, abnormal labels appear at time points such as 2023-07-01 02:00, 2023-07-01 05:00, and 2023-07-01 05:30. However, these points are not actually truly abnormal, and from the perspective of the movement speed of the sampling blade, the entire process remains within a reasonable range in the industry, which can indirectly prove the correctness of the data and is also affected by the mineral type. The composition of copper ore slurry is relatively simple. It can be seen that our invention can completely improve the evaluation of the current linear moving knife-edge slurry sampler.

[0095] Example 3, reference Figure 2, as an embodiment of the present invention, provides an online abnormality monitoring system for a linear moving knife-edge slurry sampler, including a data acquisition module, an offline learning module, and an online learning module.

[0096] Among them, the data acquisition module is used to collect time series data of different monitoring projects and perform data preprocessing. The offline learning module is used to construct statistical features in the offline learning stage, perform anomaly monitoring through unsupervised algorithms, determine anomaly labels through majority voting strategies, improve the monitoring performance of unsupervised algorithms based on feature selection, correct labels and train models. The online learning module is used to quantitatively evaluate the distribution differences between offline samples and online samples through the stability index PSI in the online learning stage to achieve stability optimization.

[0097] If the function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods of each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0098] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0099] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering, or processing in another suitable manner as necessary, and then stored in a computer memory.

[0100] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc. It should be noted that the above embodiments are only used to illustrate the technical solution of the present invention and are not limited. Although the present invention is described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical solution of the present invention can be modified or replaced by equivalents without departing from the spirit and scope of the technical solution of the present invention, which should be included in the scope of the claims of the present invention.

[0101] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for online abnormality monitoring of a linear moving knife-edge slurry sampler, characterized in that: include: Collect time series data of different monitoring projects and perform data preprocessing; In the offline learning phase, statistical features are constructed, anomaly monitoring is performed through unsupervised algorithms, and anomaly labels are determined through majority voting strategies. The monitoring performance of unsupervised algorithms is improved based on feature selection, labels are corrected, and models are trained. The online monitoring learning phase achieves stability optimization by quantitatively evaluating the distribution difference between offline samples and online samples through the stability index PSI.

2. The method for online abnormality monitoring of a linear moving knife-edge slurry sampler according to claim 1, characterized in that: The collecting of time series data of different monitoring items and data preprocessing include extracting a mobile knife-edge slurry sampler online monitoring system data set D, collecting time series data, and constructing model features with the maximum set to adapt to all linear mobile knife-edge slurry sampler monitoring items for different monitoring items of linear mobile knife-edge slurry samplers installed at different locations; Conduct data quality checks, including data type, control records, missing rate, number of null value records, minimum value, 1st percentile, 5th percentile, 25th percentile, 50th percentile, 75th percentile, 95th percentile, 99th percentile, maximum value, mean, and variance statistics; The data collection frequency for all monitoring items is 30 minutes. The data set contains timestamp indicators, linear moving knife-edge slurry sampler models and monitoring items, and a data set library suitable for all linear moving knife-edge slurry samplers is constructed; Clean the collected time series data, fill it with the previous non-empty value, and delete the records that are still empty after filling; Calculate the Pearson correlation coefficient for each pair of features and remove features with a correlation coefficient greater than 0.

7.

3. The method for online abnormality monitoring of a linear moving knife-edge slurry sampler according to claim 2, characterized in that: The constructing of statistical features in the offline learning stage includes constructing features for the retained features, including a statistical feature set, a time feature set, and a time series feature set; The construction of the statistical feature set includes calculating the skewness and kurtosis of the data, and combining it with the mean and variance that have already been calculated; The construction of the time feature set includes the construction based on the time interval of data collection and the periodicity of data, and the use of Fourier transform to extract the periodic components in the data, which is expressed as: Among them, P j (f) is the periodic component of the jth feature at frequency f, p ij is the jth eigenvalue of the ith sample, t i is the timestamp; The construction of the time series feature set includes representing the correlation of the time series at different lag periods based on the autocorrelation function, which is expressed as: Among them, α is a smoothing parameter, which is used to control the smoothness of the autocorrelation function. is the smoothed autocorrelation coefficient of the jth feature at lag period k, where k is the lag period, μ j is the mean, p i j is the j-th eigenvalue of the ith sample.

4. The method for online abnormality monitoring of a linear moving knife-edge slurry sampler according to claim 3, characterized in that: The anomaly monitoring by unsupervised algorithms and determining anomaly labels by majority voting strategy include calculating anomaly labels of each algorithm by five unsupervised learning algorithms; 5 unsupervised learning algorithms include ARIMA statistical model, Isolation Forest machine learning model, support vector machine, density clustering algorithm and principal component analysis; After completing the analysis, a majority voting strategy is used to obtain anomaly labels.

5. The method for online abnormality monitoring of a linear moving knife-edge slurry sampler according to claim 4, characterized in that: The feature selection-based method for improving the monitoring performance of the unsupervised algorithm includes eliminating the dimension through standardization, and adopting the feature selection methods of wrapping method, embedding method and filtering method. The wrapping method selects features through recursive feature elimination, which is expressed as: Among them, S is the feature subset, f is the unsupervised algorithm, and X S is the data corresponding to the feature subset, θ is the algorithm parameter, L is the loss function, y is the true label, and N is the number of samples; When the unsupervised learning algorithm is the ARIMA statistical model, the loss function is expressed as: Among them, X t is the actual time series value, is the predicted time series value, N′ is the length of the time series; When the unsupervised learning algorithm is the Isolation Forest machine learning model, the loss function is expressed as: Among them, E(h(x i )) is the expected value of the path length of the i-th sample in the tree, x i is the i-th sample, c(n) is the normalization constant; When the unsupervised learning algorithm is a support vector machine, the loss function is expressed as: Among them, w is the weight vector, ξ is the slack variable, ρ is the bias, and ν is the penalty parameter; When the unsupervised learning algorithm is a density clustering algorithm, the loss function is expressed as: Where 1 is the indicator function, the value is 1 when the condition is true, otherwise it is 0, x′ is the data point, x i ′ is the original data point, ∈ is the clustering radius, density(x i ) is the density of the i-th data point, minPts is the minimum density threshold; When the unsupervised learning algorithm is principal component analysis, the loss function is expressed as: in, is the reconstructed data point; The embedding method uses regularized Lasso regression for feature selection, expressed as: Among them, β is the regression coefficient, p is the characteristic number, and x ij is the jth feature of the i-th sample after standardization, y i is the label of the i-th sample after standardization, and λ is the regularization parameter; The filtering method uses the chi-square test to select features, expressed as: Among them, χ 2 is the chi-square statistic, O i is the observed value, E i is the expected value; The anomaly scores obtained by the five algorithms are weighted and fused according to the coefficients of 0.2:0.3:0.2:0.1:0.2, and the anomaly scores are divided into the interval of [0,100] using the minimum and maximum normalization method.

6. The method for online abnormality monitoring of a linear moving knife-edge slurry sampler according to claim 5, characterized in that: The correction label includes correction based on the typical monitoring range of the industry, setting monitoring technical requirements, the sampling time interval is allowed to differ by 10s, the sampling cycle is allowed to differ by 2s, the sampling frequency is the same as the set sampling frequency, the sample blade movement speed is allowed to differ by 0.05m / s, the sampling blade cavity volume is allowed to differ by 10%, the sampling blade cap opening is dynamically adjusted, when the maximum mineral particle diameter in the slurry is ≥2mm, the opening is ≥3 times the maximum particle diameter, when the maximum mineral particle diameter in the slurry is <2mm, the opening is ≥6mm, the sampling volume error is allowed to differ by 5%, and the sampling volume repeatability is allowed to have 5%.

7. The method for online abnormality monitoring of a linear moving knife-edge slurry sampler according to claim 6, characterized in that: The online monitoring learning stage achieves stability optimization by quantitatively evaluating the distribution difference between offline samples and online samples through the stability index PSI, including using pickle to save the results of offline data processing, feature engineering, and model training, and providing online anomaly monitoring in the form of a Restful API interface after going online. The stability index PSI is used to quantitatively evaluate the distribution difference between offline samples and online samples, and the degree of change of the distribution of each risk category across time windows is statistically analyzed. If the degree of change is less than 10%, there is no need to update the model. If the degree of change is between 10% and 20%, the cause of the change is identified and an alarm is sent to the operation and maintenance personnel. If the degree of change is greater than 20%, the model is iterated until the degree of change is maintained at less than 10%.

8. A system using the method for online abnormality monitoring of a linear moving knife-edge slurry sampler according to any one of claims 1 to 7, characterized in that: Including data acquisition module, offline learning module, online learning module; The data acquisition module is used to collect time series data of different monitoring items and perform data preprocessing; The offline learning module is used to construct statistical features in the offline learning stage, perform anomaly monitoring through an unsupervised algorithm, determine anomaly labels through a majority voting strategy, improve the monitoring performance of the unsupervised algorithm based on feature selection, correct labels and train the model; The online learning module is used to quantitatively evaluate the distribution difference between offline samples and online samples through the stability index PSI in the online learning phase to achieve stability optimization.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method for online abnormality monitoring of a linear moving knife-edge slurry sampler according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method for online abnormality monitoring of a linear moving knife-edge slurry sampler according to any one of claims 1 to 7 are implemented.

Citation Information

Patent Citations

  • Time sequence anomaly detection method for on-line monitoring system of dressing and smelting slag yard

    CN115545115A

  • Urban building abnormal energy consumption detection and optimization method, system, medium and equipment

    CN118013362A

  • Database exception monitoring method and device, computer device, and storage medium

    WO2021184554A1

  • Anomaly detection method for large-scale multivariate time series data in cloud environment

    WO2022160902A1

Cited By

  • Gold ore flotation full-process intelligent monitoring and optimal control method and system

    CN120276349A

  • Automatic viscous gold concentrate sampling system based on intelligent sensing

    CN120668946A

  • Online abnormity monitoring method and system for linear movement cutting ore pulp sampler

    CN120832618A

  • A method and system for online anomaly monitoring of a linear moving cutting slurry sampler

    CN120832618B