Sand making line status analysis method and system based on data analysis

Abstract

The embodiments of the present application relate to the field of data analysis technology, and specifically to a sand-making line state analysis method and system based on data analysis. The embodiments of the present application can accurately locate the scheduling demand events of the target production line by obtaining sample data with a one-to-one matching relationship and the prior annotation of the sample sand-making machine operation data. The state trend vector and the associated trend vector are obtained by the original state analysis algorithm, and then the features are obtained through time series element identification and fluctuation statistical analysis, so as to fully mine the data information. Based on these, the scheduling demand identification labels can be determined to efficiently and accurately judge the scheduling demand of the sand-making machine. The error comparison link helps to accurately find out the deficiencies of the original algorithm in demand identification, time series analysis and fluctuation statistics. According to these errors, the original algorithm is debugged to update it to the sand-making line scheduling processing algorithm, which improves the accuracy and reliability of the production line scheduling demand identification for the sand-making machine operation parameters to be analyzed, and optimizes the overall operation efficiency of the sand-making line.

Description

Technical Field

[0001] The embodiments of the present application relate to the field of data analysis technology, and specifically, to a sand making line status analysis method and system based on data analysis. Background Art

[0002] In the sand making industry, efficient and stable operation of the sand making line is of vital importance. The traditional sand making machine operation management method is difficult to accurately judge the production line scheduling needs, and there are many technical problems. On the one hand, there is a lack of comprehensive and targeted analysis of the sand making machine operation data and the sand making line process status data, and it is difficult to extract effective information from complex data to determine the scheduling needs. For example, it is impossible to accurately grasp the trend changes in the sand making machine operation data and the relationship between these changes and scheduling needs. On the other hand, there is no effective means to optimize the analysis algorithm. The existing algorithms may have errors in demand identification, time series analysis, and fluctuation statistics, but they cannot be adjusted and improved in time, resulting in inaccurate identification of the production line scheduling needs of the sand making machine operation parameters, which in turn affects the production efficiency of the sand making line and increases the risk of equipment failure. Summary of the invention

[0003] In order to at least overcome the above-mentioned deficiencies in the prior art, one of the purposes of the embodiments of the present application is to provide a sand making line status analysis method and system based on data analysis.

[0004] The embodiment of the present application provides a sand-making line state analysis method based on data analysis, and the method is applied to a sand-making line state analysis system, and the method includes: obtaining sample sand-making machine operation data and sand-making line process state data with a one-to-one matching relationship, and the sample sand-making machine operation data carries a priori annotations for characterizing target production line scheduling demand events in the sample sand-making machine operation data; using the original state analysis algorithm to perform trend analysis on the sample sand-making machine operation data to obtain a state trend vector and an associated trend vector with a dimension greater than the state trend vector; performing time series element identification on the state trend vector to obtain time series element features, and performing fluctuation statistical analysis on the associated trend vector to obtain fluctuation statistical features; based on the state trend vector, the associated trend vector potential vector, the timing element characteristics and the fluctuation statistical characteristics, determine the scheduling demand identification label of the sample sand making machine operation data; based on the scheduling demand identification label and the prior annotation, an error comparison is performed to obtain the demand identification error, based on the timing annotation of the prior annotation and the timing element characteristics, an error comparison is performed to obtain the timing analysis error, based on the sand making line process state data and the fluctuation statistical characteristics, an error comparison is performed to obtain the fluctuation statistical error; based on the timing analysis error, the fluctuation statistical error and the demand identification error, the original state analysis algorithm is debugged to update the original state analysis algorithm into a sand making line scheduling processing algorithm, and the sand making line scheduling processing algorithm is used to identify the production line scheduling demand for the sand making machine operation parameters to be analyzed.

[0005] An embodiment of the present application also provides a sand making line status analysis system, comprising a processor and a memory and a bus connected to the processor; wherein the processor and the memory communicate with each other through the bus; the processor is used to call program instructions in the memory to execute the above-mentioned sand making line status analysis method based on data analysis.

[0006] An embodiment of the present application also provides a computer-readable storage medium on which a program is stored, and when the program is executed by a processor, the above-mentioned sand making line state analysis method based on data analysis is implemented.

[0007] A sand-making line state analysis method and system based on data analysis provided in an embodiment of the present application obtains sample data with a one-to-one matching relationship, wherein the prior annotations of the sample sand-making machine operation data can accurately locate the scheduling demand events of the target production line, laying an accurate foundation for subsequent analysis. The state trend vector and the associated trend vector are obtained by the original state analysis algorithm, and then the features are obtained through time series element identification and fluctuation statistical analysis, so as to fully mine the data information. Based on these, the scheduling demand identification labels can be determined to efficiently and accurately judge the scheduling demand of the sand-making machine. The error comparison link helps to accurately find out the deficiencies of the original algorithm in demand identification, time series analysis and fluctuation statistics. Based on these errors, the original algorithm is debugged to update it to a sand-making line scheduling processing algorithm, which greatly improves the accuracy and reliability of identifying the production line scheduling demand for the sand-making machine operation parameters to be analyzed, and optimizes the overall operation efficiency of the sand-making line. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0009] Figure 1 A flow chart of a sand making line status analysis method based on data analysis provided in an embodiment of the present application.

[0010] Figure 2 A block diagram of a sand making line status analysis system provided in an embodiment of the present application.

[0011] icon:

[0012] 100-Sand making line status analysis system;

[0013] 101 - processor; 102 - memory; 103 - bus. DETAILED DESCRIPTION

[0014] The exemplary embodiments disclosed in the embodiments of the present application will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the embodiments of the present application are shown in the accompanying drawings, it should be understood that the embodiments of the present application can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided in order to enable a more thorough understanding of the embodiments of the present application, and to enable the scope of the embodiments of the present application to be fully communicated to those skilled in the art.

[0015] In order to better understand the above technical scheme, the technical scheme of the embodiment of the present application is described in detail below through the accompanying drawings and specific embodiments. It should be understood that the embodiment of the present application and the specific features in the embodiment are detailed descriptions of the technical scheme of the embodiment of the present application, rather than limitations on the technical scheme of the embodiment of the present application. In the absence of conflict, the embodiment of the present application and the technical features in the embodiment can be combined with each other.

[0016] Figure 1 This is a flow chart of a sand making line status analysis method based on data analysis provided according to an embodiment of the present application, which is applied to a sand making line status analysis system and includes steps 10 to 50.

[0017] Step 10: Obtain sample sand-making machine operation data and sand-making line process status data with a one-to-one matching relationship, wherein the sample sand-making machine operation data carries a priori annotations for characterizing target production line scheduling demand events in the sample sand-making machine operation data.

[0018] Step 20: Use the original state analysis algorithm to perform trend analysis on the sample sand making machine operation data to obtain a state trend vector and an associated trend vector whose dimension is greater than the state trend vector; perform time series element identification on the state trend vector to obtain time series element characteristics, and perform fluctuation statistical analysis on the associated trend vector to obtain fluctuation statistical characteristics.

[0019] Step 30: Based on the state trend vector, the associated trend vector, the time series element characteristics and the fluctuation statistical characteristics, determine the scheduling requirement identification label of the sample sand making machine operation data.

[0020] Step 40: Perform an error comparison based on the scheduling demand identification label and the prior annotation to obtain the demand identification error, perform an error comparison based on the timing annotation of the prior annotation and the timing element characteristics to obtain the timing analysis error, and perform an error comparison based on the sand making line process status data and the fluctuation statistical characteristics to obtain the fluctuation statistical error.

[0021] Step 50: Based on the timing analysis error, the fluctuation statistical error and the demand identification error, the original state analysis algorithm is debugged to update the original state analysis algorithm into a sand making line scheduling processing algorithm, and the sand making line scheduling processing algorithm is used to identify the production line scheduling demand for the sand making machine operating parameters to be analyzed.

[0022] First, in step 10, the sand making line state analysis system needs to obtain specific data. The sample sand making machine operation data and the sand making line process state data here have a one-to-one matching relationship. The sample sand making machine operation data also has a priori annotations, which are used to characterize the target production line scheduling demand events in the sample sand making machine operation data. For example, the sample sand making machine operation data may contain the changes of many parameters such as the current value, voltage value, speed of the sand making machine, vibration amplitude, etc. over time. For example, the motor current suddenly rises from a stable 100 amperes to 150 amperes in a certain period of time. This change may be related to a specific production line scheduling demand event, such as the material blockage inside the sand making machine causing the load to increase, and the priori annotations can mark this relationship. The sand making line process state data may include information such as the feed rate and the particle size distribution of the output material. For example, the feed rate suddenly decreases from 50 tons per hour to 30 tons at a certain moment, which is related to the change of the sand making machine operation data.

[0023] Next, in step 20, the sand making line state analysis system uses the original state analysis algorithm to perform trend analysis on the sample sand making machine operation data. Taking the motor current of the sand making machine as an example, the state trend vector is obtained by analyzing the current data within a period of time (such as the past 24 hours). For example, the current data shows a trend of first slowly rising and then stabilizing within these 24 hours. This trend can be represented by a state trend vector. At the same time, an associated trend vector with a dimension greater than the state trend vector can also be obtained. For example, in addition to the current trend, other factors related to the current, such as the speed of the sand making machine, the ambient temperature and other factors, are also considered to change over time, and the associated trend vector is formed by combining them. For the state trend vector, the time series element is identified to obtain the time series element feature. For example, in the trend of the motor current, the time period when the current rises is identified to be from 9 am to 11 am, and this time period is a time series element feature. The associated trend vector is subjected to fluctuation statistical analysis to obtain the fluctuation statistical feature. For example, the fluctuation range of the sand making machine speed in a day is between 100 and 120 rpm, and the average fluctuation amplitude is 5 rpm, which is the fluctuation statistical feature.

[0024] Then, in step 30, the scheduling demand identification label of the sample sand making machine operation data is determined based on the previously obtained state trend vector, associated trend vector, time series element characteristics and fluctuation statistical characteristics. For example, if the state trend vector shows that the power of the sand making machine is continuously decreasing, the associated trend vector shows that the speed and feed rate of the sand making machine have also changed accordingly, the time series element characteristics indicate that such changes occur in a specific production time period, and the fluctuation statistical characteristics show that the fluctuation range of these parameters exceeds the normal range, the combination of this information can be used to affix a scheduling demand identification label to the sample data, such as "the feed rate needs to be adjusted to stabilize the operation of the sand making machine."

[0025] Entering step 40, the sand making line state analysis system needs to perform an error comparison. Based on the scheduling demand identification label and the prior annotation, an error comparison is performed to obtain the demand identification error. For example, the scheduling demand identification label is "need to check the wear of the sand making machine tool", and the prior annotation is marked as "need to adjust the feeding speed", which means there is a demand identification error. Based on the timing annotation of the prior annotation and the timing element characteristics, an error comparison is performed to obtain the timing analysis error. For example, the timing annotation of the prior annotation indicates that a state change of the sand making machine should occur between 2 and 3 pm, but the timing element characteristics show that this change actually occurred between 3 and 4 pm, which results in a timing analysis error. Based on the sand making line process state data and the fluctuation statistical characteristics, an error comparison is performed to obtain the fluctuation statistical error. For example, the sand making line process state data shows that the discharge particle size fluctuates greatly within a certain period of time, while the normal fluctuation range given in the fluctuation statistical characteristics is small, which results in a fluctuation statistical error.

[0026] Finally, in step 50, the sand making line state analysis system debugs the original state analysis algorithm according to the timing analysis error, fluctuation statistical error and demand identification error. If the demand identification error is large, it means that the original state analysis algorithm has problems in identifying the scheduling demand, and it may be necessary to adjust the setting of various parameter weights in the algorithm or adjust the logical relationship of judging the scheduling demand. For example, it may be necessary to increase the weight of the power change of the sand making machine in judging the scheduling demand. If the timing analysis error is large, it may be necessary to optimize the processing part of the timing elements in the algorithm, such as improving the resolution of time series data or improving the extraction method of timing features. If the fluctuation statistical error is large, it is necessary to re-examine the calculation and analysis part of the fluctuation statistics in the algorithm, such as changing the method of calculating the fluctuation range or adjusting the fluctuation threshold. Through such debugging, the original state analysis algorithm is updated to the sand making line scheduling processing algorithm. This new algorithm can identify the production line scheduling demand for the sand making machine operating parameters to be analyzed. For example, for the new sand making machine operation data, the new algorithm can accurately judge whether it is necessary to adjust the feed rate, whether there is a hidden danger of equipment failure and other production line scheduling needs, thereby improving the operation efficiency and stability of the sand making line.

[0027] Through a series of steps, from data acquisition to algorithm debugging and updating, the sand making line status analysis system realizes the precise analysis of the operating status of the sand making machine and the accurate identification of the production line scheduling needs. This is of great significance for the sand making industry to improve production efficiency, reduce costs and ensure stable operation of equipment.

[0028] To elaborate further, in the process of acquiring data in step 10, the accuracy and completeness of the data are crucial to the implementation of the entire invention. For the sample sand making machine operation data, in addition to the common parameters mentioned above, it may also include temperature sensor data inside the equipment, lubricating oil pressure data, etc. The collection frequency of these data also needs to be set reasonably. For example, for motor current and voltage data, it may be necessary to collect them once every 10 seconds to accurately capture their dynamic changes; and for the temperature data inside the equipment, since its changes are relatively slow, it may only be necessary to collect them once every 30 seconds. The discharge particle size distribution data in the sand making line process status data may need to be measured using a high-precision particle size analyzer, and multiple measurements should be performed at different discharge port positions to take the average value to ensure that the data can truly reflect the discharge status of the sand making line.

[0029] In the trend analysis and feature extraction process of step 20, there are many specific implementation methods for the original state analysis algorithm. For trend analysis, in addition to simply observing the rising or falling trend of data over time, more complex methods can also be used. For example, the trend can be analyzed after smoothing the data using the moving average method. Taking the vibration amplitude data of the sand making machine as an example, by calculating the 5-minute moving average, the long-term trend of the vibration amplitude can be seen more clearly, avoiding the interference of short-term fluctuations. When identifying time series elements, the production time of a day can be divided into different time periods, such as the morning shift, the middle shift and the evening shift, and the operating status of the equipment in each time period may have different characteristics. For example, in the morning shift, the equipment has just started, and the parameters may be in the preheating stage, with large fluctuations; while in the middle shift, the equipment is in a stable operation period, and the parameters are relatively stable. For fluctuation statistical analysis, in addition to calculating the fluctuation range and the average fluctuation amplitude, the frequency of fluctuations can also be analyzed. For example, if the speed of the sand making machine fluctuates frequently in a short period of time, this may be related to the control system of the equipment or the uniformity of the feed.

[0030] When determining the scheduling demand identification tag in step 30, it is necessary to comprehensively consider the complex relationship of multiple factors. For example, when the power of the sand making machine decreases, the vibration amplitude increases, and the output particle size becomes coarser, this may not be a problem caused by just one single factor. It may be that the crushing components inside the equipment are worn, affecting the crushing efficiency, and it may also be related to the uneven feeding amount. In this case, the scheduling demand identification tag may need to comprehensively consider the inspection of the crushing components, the adjustment of the feeding equipment, and other aspects of the needs, such as "check the crushing components and adjust the feeding equipment to optimize the sand making effect."

[0031] In the error comparison of step 40, different types of errors reflect problems in different aspects of the algorithm. Demand identification errors may be caused by the algorithm's insufficient understanding of the coupling relationship between various parameters. For example, there is a nonlinear relationship between the power and feed rate of the sand making machine, and the original state analysis algorithm may deviate when processing this complex relationship. Timing analysis errors may be related to the algorithm's inaccurate grasp of the time characteristics of the production process. For example, the timing of the production process may be adjusted in different seasons or under different production orders of the sand making line. If the algorithm does not adapt to this change in time, a timing analysis error will occur. Fluctuation statistical errors may be caused by the algorithm's unreasonable definition of the normal fluctuation range under different working conditions. For example, in a high temperature environment, the fluctuation range of some parameters of the sand making machine may be larger than that in a normal temperature environment. If the algorithm does not take into account the influence of such environmental factors, a large fluctuation statistical error will occur.

[0032] When debugging the original state analysis algorithm in step 50, different debugging strategies need to be adopted for different error types. For demand identification errors, if they are caused by unreasonable parameter weight settings, retraining can be performed through a large amount of sample data to adjust the weights of each parameter. For example, through the analysis of 100 groups of sample data, it is found that after the weight of the vibration frequency of the sand making machine in the scheduling demand identification is increased from 0.3 to 0.5, the demand identification error is significantly reduced. For timing analysis errors, more time-related factors can be introduced to optimize the algorithm. For example, time factors such as the time schedule of production orders and equipment maintenance cycles are added to the algorithm so that the algorithm can analyze timing elements more accurately. For fluctuation statistical errors, the fluctuation threshold and the method of calculating the fluctuation range can be reset according to different working conditions. For example, the fluctuation threshold of the speed of the sand making machine is adjusted from ±5 rpm to ±8 rpm according to factors such as ambient temperature and humidity to adapt to the actual situation under different working conditions.

[0033] During the entire application implementation process, the sand making line status analysis system continuously optimizes and improves the algorithm to adapt to the complex and changeable operating environment and production needs of the sand making line, thereby achieving efficient and stable scheduling and management of the sand making line.

[0034] To elaborate further, in step 10, the equipment for collecting the sample sand making machine operation data and the sand making line process status data needs to have high precision and high reliability. For example, the accuracy of the sensor for collecting motor current may need to reach ±0.1 ampere to ensure that small current changes can be accurately captured. For the feed amount in the sand making line process status data, the feed amount measuring equipment such as the belt scale used needs to be calibrated regularly to avoid measurement errors. Moreover, when acquiring data with a one-to-one matching relationship, it is necessary to establish an effective data management system to ensure the accuracy and integrity of the data during storage and transmission. For example, a database management system can be used to classify and store data, and index it according to time series, equipment number, etc., to facilitate subsequent data query and analysis.

[0035] In terms of trend analysis in step 20, the original state analysis algorithm may use some advanced data analysis techniques. For example, when analyzing the correlation trend vector, the principal component analysis (PCA) method can be used. For example, there are multiple related parameters of the sand making machine, such as motor power, vibration amplitude, feed amount, and discharge particle size, which constitute the correlation trend vector. Through PCA, these parameters can be projected into a new low-dimensional space to extract the main components, thereby simplifying the analysis of the data. When identifying time series elements, the seasonal decomposition method in time series analysis can be used for reference. Taking the daily output data of the sand making machine as an example, it is decomposed into trend items, seasonal items, and random items. By analyzing the seasonal items, it can be found that the output of the sand making machine may have periodic changes in different seasons or months due to the influence of factors such as raw material humidity and temperature. These rules are important time series element characteristics.

[0036] When determining the scheduling demand identification label in step 30, in order to improve the accuracy of the label, the idea of ​​classification algorithms such as decision trees can be adopted. For example, the various parameter states of the sand making machine are used as input nodes of the decision tree, and by constructing different branch conditions, such as "if the power drops by more than 10% and the vibration amplitude increases by more than 5%, the scheduling demand is to check the internal components of the equipment", a complete decision tree model is gradually constructed, and the final scheduling demand identification label is determined based on this model.

[0037] In the error comparison of step 40, for the calculation of demand identification error, a method based on confusion matrix can be used. For example, the actual scheduling demand (determined by prior annotations) is compared with the scheduling demand identified by the algorithm, and the number of true positives, false positives, true negatives and false negatives is calculated. Then, the relevant indicators of demand identification error, such as accuracy, recall rate, etc., are calculated based on these numbers. For timing analysis error, it can be measured by calculating the sum of the absolute values ​​of time deviations or statistics such as mean square error. For example, the sum of the absolute differences between the time points in the time series annotations of the prior annotations and the time series element features is calculated. The larger the sum, the larger the timing analysis error. For fluctuation statistical errors, the relative error calculation method can be used, such as calculating the relative difference between the fluctuation range calculated by the algorithm and the fluctuation range in the actual sand making line process state data.

[0038] When debugging the original state analysis algorithm in step 50, in addition to the debugging strategies for different error types mentioned above, an integrated learning method can also be used to optimize the algorithm. For example, multiple different original state analysis algorithms (such as different trend analysis methods or feature extraction methods) are integrated to obtain the final result through weighted voting or averaging. During the debugging process, the weights of each sub-algorithm are adjusted according to error feedback to improve the overall performance of the algorithm.

[0039] During the implementation of the entire invention, the sand making line status analysis system continuously adopts advanced technologies and methods in each step, from data collection to algorithm debugging, to comprehensively improve the ability of sand making line status analysis and scheduling demand identification, ensure the efficient and stable operation of the sand making line, and improve the production efficiency and competitiveness of the sand making enterprise.

[0040] To further elaborate, in step 10, the diversity and representativeness of the data are crucial for subsequent analysis. In addition to the sand making machine operation data and sand making line process status data mentioned above, historical maintenance record data of the equipment can also be included. For example, which parts have been replaced in the past maintenance of the equipment, the time interval between maintenance, and other information. These data can provide additional references for the operation status analysis of the sand making machine. For example, if a sand making machine has frequently replaced its cutter in the past year, when analyzing the current operation data, it is necessary to pay more attention to parameters related to cutter wear, such as the discharge particle size, the power of the sand making machine, etc. At the same time, in order to ensure the representativeness of the data, the selection of sample data should cover different production conditions. For example, including sand making machine operation data and sand making line process status data under different raw material types, different production scales, etc.

[0041] In step 20, the data processing capability of the original state analysis algorithm needs to be continuously improved. In trend analysis, the analysis of state trend vectors and associated trend vectors can be combined with wavelet analysis methods. Taking the vibration signal of a sand making machine as an example, wavelet analysis can decompose the vibration signal into sub-signals of different frequencies, so as to more carefully analyze the trend changes of the vibration signal at different frequencies. This multi-scale analysis method helps to discover subtle changes hidden in the data. When identifying time series elements, a hidden Markov model (HMM) can be used. For example, the operating state of a sand making machine is regarded as a hidden Markov chain, and different operating states (such as normal operation, impending failure, etc.) correspond to different hidden states, while the observable sand making machine operating data (such as current, vibration, etc.) are the observed values ​​under these hidden states. HMM can better identify the characteristics of time series elements, such as the transition time points between different operating states.

[0042] When determining the scheduling requirement identification label in step 30, the concept of fuzzy logic can be considered. Since the relationship between the operating status of the sand making machine and the scheduling requirement is often not absolute, there is a certain ambiguity. For example, when the power of the sand making machine decreases slightly and the vibration amplitude increases slightly, it is difficult to determine whether it is a clear scheduling requirement. Through fuzzy logic, some fuzzy sets can be defined, such as "power slightly decreased" and "vibration amplitude slightly increased". Fuzzy concepts such as "slightly increased" can then be used to determine the scheduling requirement identification label based on the fuzzy relationship between these fuzzy concepts, such as "a simple equipment inspection may be required."

[0043] In the error comparison of step 40, for demand identification errors, in addition to the confusion matrix-based method, the ROC curve (Receiver Operating Characteristic Curve) can also be used to more comprehensively evaluate the performance of the algorithm. The ROC curve can show the relationship between the true positive rate and the false positive rate of the algorithm under different decision thresholds. By analyzing the area under the ROC curve (AUC), the pros and cons of the algorithm in demand identification can be more intuitively judged. For timing analysis errors, in addition to calculating the statistics of time deviation, it can also be analyzed from the perspective of the frequency domain. For example, the time series feature and the time series in the prior annotation are converted into frequency domain signals, and the timing analysis error is evaluated by comparing the difference between the two in the frequency domain. For fluctuation statistical errors, in addition to the calculation of relative errors, the cumulative distribution function (CDF) can also be used to compare the fluctuation statistical characteristics calculated by the algorithm and the fluctuation in the actual sand making line process state data.

[0044] When debugging the original state analysis algorithm in step 50, in addition to the above-mentioned methods, the idea of ​​reinforcement learning can also be used. The debugging process of the algorithm is regarded as a decision-making process of an intelligent agent in the environment. With the goal of reducing errors, the algorithm is optimized based on the error feedback (reward signal) obtained by continuously trying different debugging strategies (such as adjusting parameters, changing algorithm structure, etc.). For example, when the error is significantly reduced after adopting a certain debugging strategy, a positive reward signal is given to encourage the algorithm to continue to optimize in this direction.

[0045] The sand making line status analysis system continuously explores and applies new technologies and methods in each step, and continuously improves its analysis and scheduling capabilities to adapt to the increasingly complex production needs and technological development trends in the sand making industry, thereby providing sand making companies with more efficient and intelligent sand making line management solutions.

[0046] In an optional embodiment, the fluctuation statistical analysis of the associated trend vector to obtain the fluctuation statistical characteristics includes: determining the abnormal characteristics for the associated trend vector based on the environmental characteristic attention strategy; mapping the associated trend vector according to the abnormal characteristics to obtain the environmental characteristic attention vector; and performing fluctuation statistical analysis on the environmental characteristic attention vector to obtain the fluctuation statistical characteristics.

[0047] Furthermore, the scheduling requirement identification label of the sample sand making machine operation data is determined based on the state trend vector, the associated trend vector, the time series element feature and the fluctuation statistical feature, including: fusing the state trend vector and the associated trend vector to obtain a fused trend vector; splicing the fused trend vector with the abnormal feature, the time series element feature and the fluctuation statistical feature to obtain a splicing result; and identifying the scheduling requirement identification label of the sample sand making machine operation data based on the splicing result.

[0048] In a more detailed step, the state trend vector and the associated trend vector are fused to obtain a fused trend vector, including: determining a timing jump feature for the fluctuation statistical feature based on a timing feature attention strategy; mapping the environmental feature attention vector according to the timing jump feature to obtain a cross-mapping feature; and fusing the state trend vector and the cross-mapping feature to obtain a fused trend vector.

[0049] In the next step, the fused trend vector is spliced ​​with the abnormal characteristics, the time series element characteristics and the fluctuation statistical characteristics to obtain a splicing result, including: clustering based on the fusion result of the abnormal characteristics and the time series element characteristics to obtain the clustering characteristics of the sample sand making machine operation data; splicing the fluctuation statistical characteristics and the clustering characteristics for the fused trend vector to obtain a splicing result.

[0050] In this optional embodiment, the sand making line state analysis system performs a series of complex operations to achieve the goal. First, in the link of performing fluctuation statistical analysis on the associated trend vector to obtain fluctuation statistical features. Based on the environmental feature attention strategy, the abnormal characteristics are determined for the associated trend vector. The environmental feature attention strategy here is a mechanism for considering factors related to the operating environment of the sand making line, which involves the various environmental conditions in which the sand making line is located and the potential impact of these conditions on the operation of the sand making machine. For example, the associated trend vector contains multi-dimensional data such as the speed, feed amount, ambient temperature, and humidity of the sand making machine. In the actual sand making production process, the ambient temperature and humidity have an impact on the operation of the sand making machine that cannot be ignored. When the ambient temperature is in a higher numerical range, such as 30-40 degrees Celsius, some parts of the sand making machine may undergo slight size changes due to the principle of thermal expansion and contraction, which may affect the mechanical matching accuracy inside the sand making machine. At the same time, changes in humidity will also affect the characteristics of the material. For example, the material may become more viscous in a high humidity environment, which will affect the smoothness of the feeding, and then affect the relationship between the speed and feed amount of the sand making machine.

[0051] The environmental feature focus strategy will be based on an in-depth understanding of the relationship between environmental factors and the operation of the sand making machine during the sand making production process, and determine that some data features that are closely related to environmental changes are abnormal features. For example, under specific ambient temperature and humidity conditions, the relationship between the speed and feed amount of the sand making machine shows a specific change trend. For example, when the ambient temperature rises to 35 degrees Celsius and the humidity reaches 80%, the speed of the sand making machine will drop to a certain extent, and the feed amount will also decrease accordingly. This coordinated change feature between the speed and feed amount under specific environmental conditions is determined as an abnormal feature. This abnormal feature is not just a simple numerical change, but also includes the complex relationship between environmental factors and the operating parameters of the sand making machine.

[0052] The environmental feature attention vector is obtained by mapping the associated trend vector according to the abnormal characteristics. For example, if the abnormal characteristics indicate that the speed and feed rate of the sand making machine have a specific coordinated change under high temperature and high humidity environment, then in the mapping process, the data elements related to this coordinated change in the associated trend vector will be highlighted to form the environmental feature attention vector. This mapping process is a screening and reorganization process based on abnormal characteristics. For example, the original associated trend vector contains data in 10 dimensions, such as the speed, feed rate, motor current, motor voltage, discharge particle size, ambient temperature, ambient humidity, equipment vibration amplitude, equipment operation time, equipment maintenance cycle, etc. of the sand making machine. When mapping according to the abnormal characteristics, the dimensions related to the coordinated change of speed and feed rate under a specific environment will be focused on. For example, the data of the speed, feed rate, ambient temperature and humidity may be focused on. These data are reorganized and weighted to form the environmental feature attention vector. Compared with the original associated trend vector, this environmental feature attention vector focuses more on the operation-related characteristics of the sand making machine under the influence of environmental factors. For example, in this mapped environmental feature attention vector, the data of rotation speed and feed rate may be given higher weights to highlight their importance in a specific environment, while other dimensional data that are not closely related to the abnormal features may be weakened or appropriately adjusted.

[0053] Then, the fluctuation statistical analysis of the environmental feature attention vector is performed to obtain the fluctuation statistical features. For example, the speed data of the sand making machine in the environmental feature attention vector presents a specific fluctuation pattern within a certain period of time. For example, in a continuous production cycle, the changes in the speed data of the sand making machine are observed within 8 hours. The speed of the sand making machine may fluctuate between 100-120 rpm, with an average fluctuation of 5 rpm and a maximum fluctuation of 10 rpm. At the same time, the feed amount data will also change accordingly with the fluctuation of the speed. The feed amount fluctuates between 40-50 tons per hour, with an average fluctuation of 2 tons per hour and a maximum fluctuation of 3 tons per hour. These data on the fluctuation range, average fluctuation range and maximum fluctuation range of the speed and feed amount under the specific environmental feature attention vector are part of the fluctuation statistical features. In addition, the fluctuation statistical features may also include the frequency information of the fluctuation, such as the number of fluctuations per hour of the sand making machine speed within these 8 hours, and the periodic characteristics of the fluctuation. These fluctuation statistical features can more comprehensively describe the operating stability and change law of the sand making machine under the influence of specific environmental factors.

[0054] Next, in terms of determining the scheduling demand identification label of the sample sand making machine operation data based on the state trend vector, the associated trend vector, the time series element characteristics and the fluctuation statistical characteristics. First, the state trend vector and the associated trend vector are fused to obtain the fused trend vector. Specifically, the time series jump feature is determined for the fluctuation statistical feature based on the time series feature attention strategy. For example, the power fluctuation of the sand making machine in the fluctuation statistical feature suddenly increases in a certain time period (such as 9 am-10 am). The time point and amplitude change of this sudden increase are the time series jump features. In the sand making production process, the power fluctuation of the sand making machine may be affected by many factors. For example, in this specific time period, it may be because the feed of the previous process suddenly increases, resulting in an instantaneous increase in the load of the sand making machine, thereby causing a sudden increase in power. It may also be that a certain component inside the sand making machine has a slight hidden danger of failure during this time period, resulting in abnormal power fluctuations.

[0055] The environmental feature attention vector is mapped according to the timing jump feature to obtain the cross-mapping feature. For example, the speed data of the sand-making machine in the environmental feature attention vector may have special performance in the time period corresponding to the timing jump feature. For example, in the time period of 9:00-10:00 in the morning when the power suddenly increases, the speed of the sand-making machine will drop briefly and then quickly rise again. The mapping process will extract the data elements related to this special performance to form a cross-mapping feature. For example, if there are 5 data elements in the environmental feature attention vector that are related to the special performance of the sand-making machine speed in this time period, then these 5 elements constitute the cross-mapping feature. These 5 data elements may include the mechanical transmission parameters inside the sand-making machine that are directly related to the speed change, the control parameters of the motor, and the related parameters that interact with environmental factors.

[0056] The state trend vector and the cross-mapping feature are fused to obtain a fused trend vector. For example, the power trend of the sand making machine in the state trend vector is gradually decreasing, which may be caused by component wear or other potential fault factors caused by the long-term operation of the equipment. The special performance of the speed of the sand making machine in a specific time period in the cross-mapping feature is that it first increases and then decreases. This change in speed may be related to the adjustment of the power transmission system or control system inside the sand making machine during this time period. The fused fused trend vector will combine these two features to form a new vector containing information from both. This fused trend vector can more comprehensively reflect the operating state of the sand making machine, including both the long-term trend of power and the special changes in speed in a specific time period. For example, the fused trend vector may represent this comprehensive information in a specific data structure, such as a vector containing multiple data segments, one of which represents the quantitative information of the long-term downward trend of power, and the other represents the change curve information of the speed in a specific time period.

[0057] After that, the splicing result is obtained by splicing the abnormal characteristics, time series element characteristics and fluctuation statistical characteristics of the fused trend vector. Clustering is performed based on the fusion results of the abnormal characteristics and time series element characteristics to obtain the clustering characteristics of the sample sand making machine operation data. For example, the abnormal characteristics indicate the operation changes of the sand making machine under high temperature environment, and the time series element characteristics show that a certain operation state of the sand making machine changes during the time period of 2 pm to 3 pm. For example, under high temperature environment, the operation efficiency of the sand making machine will be reduced, which is manifested as the power of the sand making machine decreases, the speed is unstable, and the particle size of the discharge becomes coarser. In the time period of 2 pm to 3 pm, the sand making machine may have problems such as excessive wear or poor lubrication of some internal parts due to the heat accumulation of the equipment reaching a certain level, further aggravating the change of the operation state.

[0058] After the two are merged, the data may be divided into several different clusters. For example, it is divided into 3 clusters, each cluster represents a different combination of operating states. The first cluster may represent the operating state of the sand making machine in a high temperature environment and in the time period of 2-3 pm, with a large power drop, violent speed fluctuations, and very coarse particle size; the second cluster may represent the operating state in a high temperature environment, but not in the time period of 2-3 pm, with a certain degree of power drop, a small speed fluctuation, and a slightly coarse particle size; the third cluster may represent the operating state of the sand making machine in a non-high temperature environment, where the operating state is relatively stable, and the power, speed, and particle size of the discharge are within the normal range. These three cluster characteristics are the clustering characteristics.

[0059] The splicing result is obtained by splicing the fluctuation statistical features and cluster features to the fusion trend vector. For example, the fusion trend vector contains the comprehensive trend information of the power and speed of the sand making machine, the fluctuation statistical features contain information such as the fluctuation range of the power and speed, and the cluster features contain the classification information of different operating state combinations. Splicing these information together forms a splicing result that comprehensively reflects the operating state of the sand making machine. This splicing result can be a very complex data structure, containing information at multiple levels and dimensions. For example, it may be represented by a nested data structure, the outermost layer is the information of the fusion trend vector, the middle layer is the information of the fluctuation statistical features, and the innermost layer is the information of the cluster features. Each layer has a close logical connection and data interaction relationship.

[0060] Finally, the scheduling demand identification label of the sample sand making machine operation data is identified based on the splicing results. This splicing result integrates all aspects of the sand making machine operation information. Through a specific algorithm or rule, such as a decision rule obtained by training a large amount of historical data, this splicing result is analyzed to determine the scheduling demand identification label. For example, if the splicing result shows that the sand making machine has a power drop, abnormal speed fluctuations, and is in a specific operating state cluster, such as the first cluster represents an operating state in a high temperature environment and in the time period of 2-3 pm, with a large power drop, violent speed fluctuations, and very coarse discharge particle size, then the scheduling demand identification label can be "adjust the feed rate of the sand making machine and check the internal components". This is because in this operating state, the feed rate may be too much, resulting in excessive load on the sand making machine, which in turn causes power drop and speed fluctuations. At the same time, the coarsening of the discharge particle size may be caused by wear or failure of internal components, so it is necessary to adjust the feed rate and check the internal components.

[0061] In this way, through a series of complex and orderly operations, the operation data of the sand making machine is analyzed and integrated in multiple dimensions and levels. From the process of determining abnormal characteristics and timing jump characteristics, the operation characteristics of the sand making machine in different environments and timings are accurately captured. The construction of fusion trend vectors, clustering characteristics, etc. makes the description of the operation status of the sand making machine more comprehensive and accurate. The final scheduling demand identification label can more effectively guide the scheduling management of the sand making machine, improve the operation efficiency of the sand making machine, reduce the risk of failure, and optimize the overall production efficiency of the sand making line. By analyzing in detail the impact of environmental characteristics, timing characteristics and other factors on the operation of the sand making machine, potential problems can be discovered in advance to avoid excessive loss of equipment and reduced production efficiency. This data-driven scheduling demand identification method can better adapt to the complex and changeable situations in the sand making production process, and provide sand making enterprises with a more scientific and reasonable basis for production decision-making.

[0062] Based on the above optional embodiments, the method further includes: performing clustering based on the associated trend vector to obtain a trend mutation feature cluster; performing error comparison based on the trend mutation feature cluster and the prior annotation to obtain a trend mutation identification error. Based on this, the debugging of the original state analysis algorithm based on the timing analysis error, the fluctuation statistical error, and the demand identification error includes: performing global processing based on the timing analysis error, the fluctuation statistical error, the demand identification error, and the trend mutation identification error to obtain a global debugging error; and debugging the original state analysis algorithm according to the global debugging error.

[0063] Based on the previous embodiments, the sand making line status analysis system further performs a series of operations to improve the entire technical solution.

[0064] In the process of obtaining trend variation feature clusters based on the associated trend vector, the associated trend vector covers the associated data trend information of multiple aspects of the sand making machine operation. For example, the trend information of the sand making machine's motor speed, feed rate, discharge particle size, ambient temperature, equipment vibration amplitude, motor current and other parameters are all included. There are complex relationships between these parameters, and during the operation of the sand making machine, their changing trends will reflect the operating status of the sand making machine and possible problems.

[0065] Take the motor speed of the sand making machine as an example. For example, in a long production cycle, the trend data of the motor speed shows that it starts from the initial 1000 rpm, stabilizes near this value for a period of time, then gradually drops to 900 rpm, and then suddenly drops to 800 rpm at a certain moment, and then maintains a certain fluctuation at a lower level. In terms of feed rate, the initial setting is 50 tons per hour. Before the motor speed starts to drop, the feed rate has a gradual decrease trend from 50 tons per hour to 45 tons per hour. When the motor speed suddenly drops, the feed rate also drops sharply to 30 tons per hour, and then fluctuates between 30-35 tons per hour.

[0066] The data of the particle size of the discharged material showed that the average value of the particle size of the discharged material was 5 mm at the beginning. With the change of the motor speed and the feeding amount, the particle size of the discharged material gradually increased. When the motor speed dropped to 800 rpm and the feeding amount was stable at 30-35 tons per hour, the particle size of the discharged material increased to 8 mm. The ambient temperature gradually increased from 20 degrees Celsius at the beginning to 25 degrees Celsius. The vibration amplitude of the equipment gradually increased from the initial small amplitude during the process of the motor speed decreasing, and the motor current also decreased with the decrease of the motor speed.

[0067] For such a complex correlation trend vector, clustering is a detailed division method based on the internal feature relationship of the data. In the clustering process, factors such as the similarity, causal relationship, and degree of mutual influence between the trends of various parameters need to be considered. For example, there is an obvious causal relationship between the downward trend of motor speed and the decreasing trend of motor current, because when the motor speed decreases, the motor load decreases, which leads to a decrease in motor current. These two feature data may be divided into a trend mutation feature cluster due to this close causal relationship.

[0068] Looking at the relationship between feed rate and discharge particle size, the reduction of feed rate will cause the degree of material crushing in the sand making machine to change, thus causing the discharge particle size to increase. When the feed rate decreases from 50 tons per hour to 45 tons per hour, the discharge particle size gradually increases from 5 mm to 6 mm. When the feed rate drops sharply to 30 tons per hour, the discharge particle size further increases to 8 mm. This strong correlation between feed rate and discharge particle size makes it possible for them to be classified into the same trend variation feature cluster.

[0069] There is also a correlation between the vibration amplitude of the equipment and the motor speed. When the motor speed decreases, the balance state of the equipment may be affected, resulting in an increase in the vibration amplitude. For example, when the motor speed decreases from 1000 rpm to 900 rpm, the vibration amplitude of the equipment starts from a smaller amplitude and has an obvious upward trend. This correlation may cause the relevant trend data of the equipment vibration amplitude and motor speed to be classified into a trend abnormality feature cluster.

[0070] Then, based on the trend anomaly feature cluster and prior annotations, the error comparison is performed to obtain the trend anomaly identification error. Prior annotations are the marking information of the target production line scheduling demand events in the sample sand making machine operation data. It contains the pre-judgment and identification of various situations during the operation of the sand making machine. For example, the prior annotation may mark that when the motor speed drops to a certain level and the feed amount is reduced to a specific value, it is due to blockage or failure of the feeding equipment, resulting in uneven feeding, which affects the normal operation of the sand making machine. In this case, the discharge particle size will increase and the vibration amplitude of the equipment will increase accordingly.

[0071] In the actual trend change feature cluster analysis, if this relationship is not accurately reflected or the degree of reflection is different from the prior annotation, trend change identification errors will occur. For example, the prior annotation clearly states that when the motor speed drops to 800 rpm and the feed rate is reduced to 30 tons per hour, the probability of this situation due to a feeder failure is 0.8 (this is an example of a correlation value based on experience or historical data), while the probability of this situation due to a feeder failure in the trend change feature cluster analysis is 0.6, so the difference of 0.2 constitutes part of the trend change identification error.

[0072] For another example, the prior annotation may indicate that when the ambient temperature rises to 25 degrees Celsius and the motor speed drops to 900 rpm, the vibration amplitude of the equipment should increase within a specific range, and this increase is related to the thermal expansion of the equipment and the slight deformation of the internal mechanical structure. If the value of the increase in the vibration amplitude of the equipment obtained in the trend anomaly feature cluster deviates from the expected value range in the prior annotation, for example, the prior annotation states that the vibration amplitude of the equipment should increase to 2-3 mm, but the actual analysis in the trend anomaly feature cluster shows that it increases to 3-4 mm, this will also cause trend anomaly identification errors.

[0073] Next, in terms of debugging the original state analysis algorithm based on the time series analysis error, fluctuation statistical error, demand identification error and trend change identification error, a global debugging error is obtained by global processing based on these errors. This global processing process is a complex process that comprehensively considers the impact of each error on the overall performance of the algorithm.

[0074] The timing analysis error reflects the deviation of the algorithm when processing time series data. For example, when analyzing the time series data of the motor speed of the sand making machine, the algorithm may not accurately capture the rapid decline trend of the motor speed in a certain time period. For example, the motor speed dropped rapidly from 1000 rpm to 900 rpm between 10:00 and 10:10 am, while the algorithm analysis concluded that the decline time was from 10:00 to 10:20 am. This time deviation constitutes part of the timing analysis error.

[0075] The fluctuation statistical error reflects the inaccurate grasp of the data fluctuation characteristics by the algorithm. For example, the output particle size data of the sand making machine has a certain fluctuation over a period of time. The actual output particle size fluctuates between 5-8 mm, with an average fluctuation of 1.5 mm, while the fluctuation range calculated by the algorithm is 4-9 mm, with an average fluctuation of 2 mm. This difference in the fluctuation range and average fluctuation leads to the fluctuation statistical error.

[0076] Demand identification errors indicate the inadequacy of the algorithm in identifying scheduling requirements. For example, according to the operating status of the sand making machine, when the motor speed decreases, the feed amount decreases, and the output particle size increases, the actual scheduling requirement is to check the feeding equipment and the crushing components inside the sand making machine. However, the scheduling requirement identified by the algorithm may be to check only the feeding equipment, which results in demand identification errors.

[0077] Trend anomaly identification errors show problems with the algorithm in analyzing trend anomaly features, such as the errors caused by comparing with prior annotations as described above.

[0078] Global processing may assign different weights to these errors according to their importance or impact, and then perform weighted summation and other operations to obtain the global debugging error. For example, if the timing analysis error is 0.1 (this is a standardized or quantified numerical example), the fluctuation statistics error is 0.15, the demand identification error is 0.2, and the trend change identification error is 0.1, and according to the characteristics and experience of the algorithm, the timing analysis error weight is 0.2, the fluctuation statistics error weight is 0.25, the demand identification error weight is 0.3, and the trend change identification error weight is 0.25, then the global debugging error = 0.1×0.2+0.15×0.25+0.2×0.3+0.1×0.25=0.1675.

[0079] Finally, the original state analysis algorithm is debugged according to the global debugging error. This debugging process may involve operations such as adjusting the internal parameters of the algorithm and optimizing the algorithm structure.

[0080] If the global debugging error is large, it indicates that there are major problems with the overall performance of the algorithm, and it may be necessary to adjust the weight distribution of different feature data in the algorithm. For example, if it is found that the trend change identification error accounts for a large proportion of the global debugging error, it may be necessary to re-examine the parameters of the clustering algorithm or improve the basis for clustering. For example, the judgment criteria for the causal relationship and degree of correlation between the trends of various parameters may need to be adjusted during the clustering process. If the time series analysis error accounts for a large proportion, it may be necessary to optimize the module that processes time series data in the algorithm, such as using a more accurate time series analysis method or increasing the resolution of time series data. If the fluctuation statistical error accounts for a large proportion, it may be necessary to improve the statistical method of data fluctuation characteristics in the algorithm, such as using a more complex fluctuation analysis model or adjusting the method of calculating the fluctuation range and average fluctuation amplitude.

[0081] In this way, by introducing trend anomaly feature clusters and trend anomaly identification errors and incorporating them into the debugging process of the original state analysis algorithm, the debugging of the algorithm is made more comprehensive and accurate. By clustering the associated trend vectors to obtain trend anomaly feature clusters, the characteristic relationship within the sand making machine operation data can be more deeply explored, which helps to discover the complex causal relationships and mutual influence relationships hidden in the data. Based on the error comparison of trend anomaly feature clusters and prior annotations, the dimension of algorithm performance evaluation is further improved, so that the algorithm evaluation is not only limited to traditional timing, fluctuation and demand identification, but also can be evaluated from the perspective of trend anomaly features. Finally, the original state analysis algorithm is debugged according to the global debugging error, which helps to improve the accuracy and reliability of the algorithm in sand making machine operation state analysis and scheduling demand identification, thereby optimizing the operation efficiency of the sand making line, reducing production problems caused by inaccurate algorithms, enhancing the adaptability of the sand making line state analysis system to complex production conditions, and improving the overall benefits of sand making production.

[0082] In a preferred technical solution, the method of using the original state analysis algorithm to perform trend analysis on the sample sand-making machine operation data to obtain a state trend vector and an associated trend vector whose dimension is greater than the state trend vector includes: using the original state analysis algorithm to perform trend analysis on the sample sand-making machine operation data to obtain multiple initial state trend one-hot encodings of different dimensions; determining a first trend one-hot encoding and a second trend one-hot encoding from the multiple initial state trend one-hot encodings according to feature screening rules, wherein the dimension of the second trend one-hot encoding is greater than the state trend vector; normalizing the first trend one-hot encoding to obtain a state trend vector, and normalizing the second trend one-hot encoding to obtain an associated trend vector.

[0083] Furthermore, the normalizing of the second trend one-hot encoding to obtain the associated trend vector includes: if there are multiple second trend one-hot encodings, for the target initial state trend one-hot encoding among the multiple second trend one-hot encodings, determining the derived state trend one-hot encoding of the target initial state trend one-hot encoding according to the remaining initial state trend one-hot encodings whose dimensions are greater than the target initial state trend one-hot encoding among the multiple second trend one-hot encodings; performing feature interaction on the target initial state trend one-hot encoding and the derived state trend one-hot encoding to obtain the state trend interaction encoding corresponding to the target initial state trend one-hot encoding; and normalizing the state trend interaction encodings corresponding to the multiple second trend one-hot encodings to obtain the associated trend vector.

[0084] Furthermore, the number of the multiple second trend one-hot encodings is P, and the target initial state trend one-hot encoding among the multiple second trend one-hot encodings is determined based on the remaining initial state trend one-hot encodings among the multiple second trend one-hot encodings whose dimensions are greater than the target initial state trend one-hot encoding, including: determining that the derived state trend one-hot encoding of the first second trend one-hot encoding is empty for the multiple second trend one-hot encodings in descending order of dimension; determining the derived state trend one-hot encoding of the u-th second trend one-hot encoding based on the derived state trend one-hot encoding of the previous u-1 second trend one-hot encoding and the u-th second trend one-hot encoding, where u is not greater than P.

[0085] In a further refined embodiment, the feature interaction of the target initial state trend one-hot encoding and the derived state trend one-hot encoding is performed to obtain a state trend interaction encoding corresponding to the target initial state trend one-hot encoding, including: performing feature interaction on the target initial state trend one-hot encoding and the derived state trend one-hot encoding to obtain an intermediate interaction vector corresponding to the target initial state trend one-hot encoding; performing full connection processing and residual processing on the intermediate interaction vector to obtain a state trend interaction encoding corresponding to the target initial state trend one-hot encoding.

[0086] In an alternative embodiment of the above-mentioned preferred technical solution, before normalizing the first trend one-hot encoding to obtain a state trend vector, the method also includes: if there are multiple first trend one-hot encodings, respectively strengthening the attention of the multiple first trend one-hot encodings based on the set attention index; before determining the derived state trend one-hot encoding of the target initial state trend one-hot encoding among the multiple second trend one-hot encodings according to the remaining initial state trend one-hot encodings whose dimensions are greater than the target initial state trend one-hot encoding among the multiple second trend one-hot encodings, the method also includes: respectively strengthening the attention of the multiple second trend one-hot encodings based on the set attention index.

[0087] In this preferred technical solution, the sand making line status analysis system processes the sample sand making machine operation data according to a series of complex operating procedures.

[0088] First, the original state analysis algorithm is used to perform trend analysis on the sample sand making machine operation data to obtain multiple initial state trend one-hot encodings of different dimensions. These initial state trend one-hot encodings are the preliminary results of the trend analysis of the sand making machine operation data. Each one-hot encoding contains trend information of the sand making machine operation in a certain aspect, and because it reflects information of different dimensions, it differs in dimension. For example, one initial state trend one-hot encoding may reflect the upward trend of the sand making machine motor speed over a period of time, and another may reflect the downward trend of the feed rate in the same time period. Their dimensions may differ due to the specific data features contained.

[0089] Next, the first trend one-hot encoding and the second trend one-hot encoding are determined from multiple initial state trend one-hot encodings according to the feature screening rule, wherein the dimension of the second trend one-hot encoding is greater than the state trend vector. The feature screening rule is a predefined rule based on the importance and relevance of the characteristics of the sand making machine operation data. For example, according to the different degrees of influence on the production efficiency during the operation of the sand making machine, the initial state trend one-hot encodings that have a greater impact on the production efficiency and have a strong data correlation are screened as the first trend one-hot encoding and the second trend one-hot encoding. For example, the initial state trend one-hot encodings of data such as the motor speed and the crushing chamber pressure related to the core crushing function of the sand making machine may be preferentially screened out, some of which are determined as the first trend one-hot encoding, and the other part is determined as the second trend one-hot encoding.

[0090] Then, the first trend one-hot encoding is normalized to obtain the state trend vector, and the second trend one-hot encoding is normalized to obtain the associated trend vector. The normalization operation is to process the data into a unified standard range for subsequent calculation and analysis. Taking the initial state trend one-hot encoding of the motor speed of the sand making machine as an example, if its original value range is between 500-1500 rpm, it may be mapped to a value range between 0-1 after normalization, and this normalized value becomes part of the state trend vector or the associated trend vector.

[0091] In the process of normalizing the second trend one-hot encoding to obtain the associated trend vector, if there are multiple second trend one-hot encodings, for the target initial state trend one-hot encoding among the multiple second trend one-hot encodings, the derived state trend one-hot encoding of the target initial state trend one-hot encoding is determined according to the remaining initial state trend one-hot encodings whose dimensions are greater than the target initial state trend one-hot encoding among the multiple second trend one-hot encodings. The multiple second trend one-hot encodings each contain sand making machine operation data trend information of different dimensions. For example, one of the second trend one-hot encodings may contain multi-dimensional data trends such as vibration amplitude, temperature and humidity of the sand making machine, and another may contain multi-dimensional data trends such as feed amount, discharge particle size and equipment operation time. Arrange the multiple second trend one-hot encodings in descending order of dimension, and determine that the derived state trend one-hot encoding of the first second trend one-hot encoding is empty. Then, according to the derived state trend one-hot encoding of the first u-1 second trend one-hot encoding and the u-th second trend one-hot encoding, determine the derived state trend one-hot encoding of the u-th second trend one-hot encoding (u is not greater than P). For example, when u=2, the derived state trend one-hot code of the second second trend one-hot code is determined by specific rules (for example, based on the physical relationship or statistical correlation between the data) according to the first second trend one-hot code (for example, it contains dimensions such as motor speed, current and power) and the second second trend one-hot code (for example, it contains dimensions such as feed amount, discharge particle size and equipment vibration amplitude). This derived state trend one-hot code can be a new trend code generated on the basis of the second second trend one-hot code and combined with some relevant information in the first second trend one-hot code, which is used to supplement and improve the trend information contained in the second second trend one-hot code.

[0092] After that, the target initial state trend one-hot encoding and the derived state trend one-hot encoding are subjected to feature interaction to obtain the state trend interaction encoding corresponding to the target initial state trend one-hot encoding. First, the target initial state trend one-hot encoding and the derived state trend one-hot encoding are subjected to feature interaction to obtain the intermediate interaction vector. For example, the target initial state trend one-hot encoding contains the trend information of the motor speed and current of the sand making machine, and the derived state trend one-hot encoding contains the supplementary trend information related to the feed amount and the vibration amplitude of the equipment and the motor speed and current. Through feature interaction (which can be based on some mathematical operation or logical relationship, such as multiplication, addition, or logical AND, logical OR, etc.), these information are fused together to obtain the intermediate interaction vector. Then, the intermediate interaction vector is subjected to full connection processing and residual processing to obtain the state trend interaction encoding corresponding to the target initial state trend one-hot encoding. The full connection processing can comprehensively connect and map each element in the intermediate interaction vector, so that the relationship between different elements can be more deeply explored. The residual processing helps to retain the original information in the intermediate interaction vector and avoid excessive loss of information during the processing process, thereby obtaining a more accurate state trend interaction encoding.

[0093] In an alternative embodiment of the above-mentioned preferred technical solution, before normalizing the first trend unique hot coding to obtain the state trend vector, if the number of first trend unique hot coding is multiple, the attention of multiple first trend unique hot coding is strengthened based on the set attention index. This set attention index is a measurement standard pre-set according to the characteristics of the sand making machine operation data and the analysis focus. For example, for those first trend unique hot codings that have a greater impact on the operating stability of the sand making machine, such as coding related to motor speed and equipment vibration amplitude, a higher attention index is set according to its importance in the overall operation of the sand making machine. Through attention strengthening, the role of these important first trend unique hot codings in subsequent calculations and analysis can be highlighted. Similarly, for the target initial state trend unique hot coding in multiple second trend unique hot codings, according to the remaining initial state trend unique hot codings with a dimension greater than the target initial state trend unique hot coding in multiple second trend unique hot codings, before determining the derived state trend unique hot coding of the target initial state trend unique hot coding, the attention of multiple second trend unique hot codings is strengthened based on the set attention index. For example, a higher attention index is set for the second trend one-hot encoding containing data related to the core components of the sand making machine (such as the crushing chamber, etc.) so that more attention can be paid to these important encoding information in subsequent operations such as determining the derivative state trend one-hot encoding and performing feature interaction.

[0094] It can be seen that the sample sand making machine operation data is processed through a series of meticulous operations. Starting from obtaining the initial state trend one-hot encoding of different dimensions, after operations such as feature screening, normalization, determination of derived state trend one-hot encoding, feature interaction, and attention enhancement, the trend information in the sand making machine operation data can be mined more comprehensively and accurately. By determining the first trend one-hot encoding and the second trend one-hot encoding, and processing them separately, trend information of different importance and dimensions can be distinguished. The determination of the derived state trend one-hot encoding and the feature interaction operation help to deeply explore the relationship between data. Attention enhancement highlights the role of important data, improves the accuracy and effectiveness of the entire analysis process for the analysis of the sand making machine operation status, and helps to optimize the operation and management of the sand making line.

[0095] In summary, the embodiments of the present application obtain sample data with a one-to-one matching relationship, in which the prior annotations of the sample sand-making machine operation data can accurately locate the scheduling demand events of the target production line, laying an accurate foundation for subsequent analysis. The state trend vector and the associated trend vector are obtained by the original state analysis algorithm, and then the features are obtained through time series element identification and fluctuation statistical analysis, so as to fully mine the data information. Based on these, the scheduling demand identification labels can be determined to efficiently and accurately judge the scheduling demand of the sand-making machine. The error comparison link helps to accurately find out the shortcomings of the original algorithm in demand identification, time series analysis and fluctuation statistics. According to these errors, the original algorithm is debugged to update it to the sand-making line scheduling processing algorithm, which greatly improves the accuracy and reliability of the production line scheduling demand identification for the sand-making machine operation parameters to be analyzed, and optimizes the overall operation efficiency of the sand-making line.

[0096] An embodiment of the present application provides a computer-readable storage medium on which a program is stored. When the program is executed by a processor, the sand making line state analysis method based on data analysis is implemented.

[0097] An embodiment of the present application provides a processor, which is used to run a program, wherein the sand making line state analysis method based on data analysis is executed when the program is running.

[0098] In the present application embodiment, Figure 2 As shown, the sand making line state analysis system 100 includes at least one processor 101, and at least one memory 102 and a bus 103 connected to the processor 101; wherein the processor 101 and the memory 102 communicate with each other through the bus 103; the processor 101 is used to call the program instructions in the memory 102 to execute the above-mentioned sand making line state analysis method based on data analysis.

[0099] The embodiments of the present application are described with reference to the flowcharts and / or block diagrams of the methods, sand making line state analysis systems (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0100] In a typical configuration, the sand making line status analysis system includes one or more processors (CPU), memory and bus. The sand making line status analysis system may also include input / output interface, network interface and the like.

[0101] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip. The memory is an example of a computer-readable medium.

[0102] Computer readable media include permanent and non-permanent, removable and non-removable media that can be used to store information by any method or technology. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage computer readable storage media or any other non-transmission media that can be used to store information that can be accessed by the sand making line state analysis system. As defined in this article, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carriers.

[0103] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or computer-readable storage medium that includes a series of elements includes not only those elements, but also other elements that are not explicitly listed, or also includes elements inherent to such process, method, commodity or computer-readable storage medium. In the absence of further restrictions, the elements defined by the sentence "comprises a ..." do not exclude the presence of other identical elements in the process, method, commodity or computer-readable storage medium that includes the elements.

[0104] Those skilled in the art will appreciate that the embodiments of the present application can be provided as methods, systems or computer program products. Therefore, the embodiments of the present application can be in the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware. Moreover, the embodiments of the present application can be in the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0105] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.

Claims

1. A sand making line state analysis method based on data analysis, characterized in that: The method is applied to a sand making line state analysis system, and the method comprises: Acquire sample sand-making machine operation data and sand-making line process status data with a one-to-one matching relationship, wherein the sample sand-making machine operation data carries a priori annotations for characterizing a target production line scheduling demand event in the sample sand-making machine operation data; Performing trend analysis on the sample sand making machine operation data using an original state analysis algorithm to obtain a state trend vector and an associated trend vector having a dimension greater than the state trend vector; Performing time series element identification on the state trend vector to obtain time series element features, and performing fluctuation statistical analysis on the associated trend vector to obtain fluctuation statistical features: determining abnormal characteristics for the associated trend vector based on the environmental feature attention strategy; mapping the associated trend vector according to the abnormal characteristics to obtain an environmental feature attention vector; performing fluctuation statistical analysis on the environmental feature attention vector to obtain fluctuation statistical features; wherein the environmental feature attention strategy is used to pay attention to the relevant factors of the sand making line operating environment, including the various environmental conditions of the sand making line and the potential impact of the various environmental conditions on the operation of the sand making machine; the abnormal characteristics characterize the relationship between environmental factors and the operating parameters of the sand making machine; Based on the state trend vector, the associated trend vector, the time series element feature and the fluctuation statistical feature, determine the scheduling demand identification label of the sample sand making machine operation data: fuse the state trend vector and the associated trend vector to obtain a fused trend vector; splice the abnormal feature, the time series element feature and the fluctuation statistical feature for the fused trend vector to obtain a splicing result; identify the scheduling demand identification label of the sample sand making machine operation data based on the splicing result; wherein the scheduling demand identification label is: adjust the feed rate to stabilize the operation of the sand making machine or check the wear of the sand making machine tool; Based on the scheduling demand identification label and the prior annotation, an error comparison is performed to obtain a demand identification error; based on the timing annotation of the prior annotation and the timing element characteristics, an error comparison is performed to obtain a timing analysis error; based on the sand making line process state data and the fluctuation statistical characteristics, an error comparison is performed to obtain a fluctuation statistical error; The original state analysis algorithm is debugged based on the timing analysis error, the fluctuation statistical error and the demand identification error, so that the original state analysis algorithm is updated into a sand making line scheduling processing algorithm, and the sand making line scheduling processing algorithm is used to identify the production line scheduling demand for the sand making machine operating parameters to be analyzed.

2. The method according to claim 1, characterized in that The step of fusing the state trend vector and the associated trend vector to obtain a fused trend vector includes: Determine a timing jump feature for the fluctuation statistical feature based on a timing feature focus strategy; Mapping the environmental feature attention vector according to the timing jump feature to obtain a cross-mapping feature; The state trend vector and the cross-mapping feature are fused to obtain a fused trend vector.

3. The method according to claim 1, characterized in that The step of splicing the abnormal movement feature, the time series element feature and the fluctuation statistical feature for the fused trend vector to obtain a splicing result includes: Based on the fusion result of the abnormal movement feature and the time series element feature, clustering is performed to obtain the clustering features of the sample sand making machine operation data; The fluctuation statistical feature and the cluster feature are spliced ​​for the fused trend vector to obtain a splicing result.

4. The method according to claim 1, characterized in that The method further comprises: performing clustering based on the associated trend vector to obtain a trend variation feature cluster; performing error comparison based on the trend variation feature cluster and the prior annotation to obtain a trend variation identification error; The debugging of the original state analysis algorithm based on the timing analysis error, the fluctuation statistical error and the demand identification error includes: performing global processing based on the timing analysis error, the fluctuation statistical error, the demand identification error and the trend anomaly identification error to obtain a global debugging error; and debugging the original state analysis algorithm according to the global debugging error.

5. The method according to any one of claims 1 to 4, characterized in that: The method of performing trend analysis on the sample sand making machine operation data using the original state analysis algorithm to obtain a state trend vector and an associated trend vector having a dimension greater than the state trend vector includes: Performing trend analysis on the sample sand making machine operation data using an original state analysis algorithm to obtain a plurality of initial state trend unique hot codes of different dimensions; Determine a first trend one-hot encoding and a second trend one-hot encoding from the multiple initial state trend one-hot encodings according to a feature screening rule, wherein the dimension of the second trend one-hot encoding is greater than that of the state trend vector; The first trend one-hot encoding is normalized to obtain a state trend vector, and the second trend one-hot encoding is normalized to obtain an association trend vector.

6. The method according to claim 5, characterized in that The step of normalizing the second trend one-hot encoding to obtain the associated trend vector includes: If there are multiple second trend one-hot codes, for the target initial state trend one-hot code among the multiple second trend one-hot codes, determine the derived state trend one-hot code of the target initial state trend one-hot code according to the remaining initial state trend one-hot codes whose dimensions are greater than the target initial state trend one-hot code among the multiple second trend one-hot codes; Performing feature interaction on the target initial state trend one-hot encoding and the derived state trend one-hot encoding to obtain a state trend interaction encoding corresponding to the target initial state trend one-hot encoding; Normalizing the state trend interaction codes corresponding to the plurality of second trend one-hot codes to obtain an associated trend vector; The number of the plurality of second trend one-hot codes is P, and the target initial state trend one-hot code among the plurality of second trend one-hot codes is determined based on the remaining initial state trend one-hot codes among the plurality of second trend one-hot codes whose dimensions are greater than the target initial state trend one-hot code, and the derived state trend one-hot code of the target initial state trend one-hot code is determined, including: determining that the derived state trend one-hot code of the first second trend one-hot code is empty according to the descending order of the dimensions of the plurality of second trend one-hot codes; determining the derived state trend one-hot code of the u-th second trend one-hot code according to the derived state trend one-hot code of the first u-1 second trend one-hot code and the u-th second trend one-hot code, wherein u is not greater than P; The step of performing feature interaction on the target initial state trend one-hot encoding and the derived state trend one-hot encoding to obtain a state trend interaction encoding corresponding to the target initial state trend one-hot encoding comprises: performing feature interaction on the target initial state trend one-hot encoding and the derived state trend one-hot encoding to obtain an intermediate interaction vector corresponding to the target initial state trend one-hot encoding; performing full connection processing and residual processing on the intermediate interaction vector to obtain a state trend interaction encoding corresponding to the target initial state trend one-hot encoding; Before normalizing the first trend one-hot encoding to obtain the state trend vector, the method also includes: if there are multiple first trend one-hot encodings, respectively strengthening the attention of the multiple first trend one-hot encodings based on the set attention index; before determining the derived state trend one-hot encoding of the target initial state trend one-hot encoding among the multiple second trend one-hot encodings according to the remaining initial state trend one-hot encodings whose dimensions are greater than the target initial state trend one-hot encoding among the multiple second trend one-hot encodings, the method also includes: respectively strengthening the attention of the multiple second trend one-hot encodings based on the set attention index.

7. A sand making line state analysis system, characterized in that: It comprises a processor and a memory and a bus connected to the processor; wherein the processor and the memory communicate with each other via the bus; the processor is used to call program instructions in the memory to execute the sand making line status analysis method based on data analysis as described in any one of claims 1-6.

8. A computer-readable storage medium, characterized in that: A program is stored thereon, and when the program is executed by a processor, the sand making line state analysis method based on data analysis as described in any one of claims 1 to 6 is implemented.

Images