Batch circulating grain dryer fault diagnosis method and batch circulating grain dryer

Abstract

This application relates to the field of dryer technology, and particularly to a fault diagnosis method for batch circulating grain dryers and the batch circulating grain dryer itself. The method includes: acquiring multi-dimensional operating data containing temperature, moisture content, wind speed, and circulation parameters; constructing a regularized fault feature dataset by performing standardized preprocessing on the multi-dimensional operating data; constructing a fault time-series feature sequence by generating corresponding parameter events according to data type and arranging them in an orderly manner according to timestamps, accurately depicting the gradual change pattern of gradual faults; and identifying weak abnormal signals and gradual trends from the fault time-series feature sequence through deep feature mining and matching diagnosis, thereby determining the specific fault type and providing a fault optimization scheme. This method can solve the problem of significant deficiencies in the identification of gradual faults and early weak faults.

Description

Technical Field

[0001] This application belongs to the field of dryer technology, and particularly relates to a fault diagnosis method for batch circulating grain dryers and a batch circulating grain dryer. Background Technology

[0002] Batch circulating grain dryers are key equipment for post-harvest grain processing. They mainly consist of a drying silo, a hot air supply system, a ventilation fan system, a grain circulation lifting mechanism, a grain discharge mechanism, a temperature control system, and a sensor detection unit. They enable grain to circulate and dry in stages within the silo, and feature uniform moisture content, high drying efficiency, and wide applicability to various grain types. They are widely used in grain planting bases, grain depots, and grain processing enterprises.

[0003] Traditional fault diagnosis methods rely heavily on human experience and single-point threshold alarms, which are difficult to meet the needs of modern intelligent and continuous drying operations, especially in the identification of gradual faults and early weak faults. Summary of the Invention

[0004] This application provides a fault diagnosis method for a batch circulating grain dryer and a batch circulating grain dryer, which can solve the problem of obvious defects in the identification of gradual faults and early weak faults.

[0005] In a first aspect, embodiments of this application provide a fault diagnosis method for a batch circulating grain dryer, including: Acquire multi-dimensional operational data of a batch circulating grain dryer when it completes the drying of a batch of grain; wherein, the multi-dimensional operational data includes equipment temperature parameters, moisture content parameters, wind speed parameters, and circulation parameters; The temperature parameters, moisture content parameters, wind speed parameters, and circulation parameters are preprocessed to obtain a regularized fault feature dataset; wherein, the regularized fault feature dataset reflects the characteristic data set of the batch circulating grain dryer's operating status and potential faults; Based on the normalized fault feature dataset, corresponding parameter events are generated according to data types, and then arranged in order according to the timestamps of each parameter event to construct a fault time-series feature sequence; wherein, the parameter events include temperature events generated according to the temperature parameters, moisture content events generated according to the moisture content parameters, wind speed events generated according to the wind speed parameters, and cycle events generated according to the cycle parameters. Deep feature mining and matching diagnosis are performed on the fault time sequence to determine the fault diagnosis conclusion of the batch circulating grain dryer; wherein, the fault diagnosis conclusion includes the specific fault type and fault optimization scheme.

[0006] The technical solutions described in this application embodiment have at least the following technical effects: The fault diagnosis method for a batch circulating grain dryer provided in this application acquires multi-dimensional operational data of the batch circulating grain dryer when it completes the drying of a batch of grain. This multi-dimensional operational data includes equipment temperature parameters, moisture content parameters, wind speed parameters, and circulation parameters. The temperature parameters, moisture content parameters, wind speed parameters, and circulation parameters are preprocessed to obtain a regularized fault feature dataset. This regularized fault feature dataset reflects the characteristic data set of the batch circulating grain dryer's operating status and potential faults. Based on the regularized fault feature dataset, corresponding parameter events are generated according to data types, and then arranged in an ordered manner according to the timestamps of each parameter event to construct a fault time-series feature sequence. The parameter events include temperature events generated based on temperature parameters, moisture content events generated based on moisture content parameters, wind speed events generated based on wind speed parameters, and circulation events generated based on circulation parameters. Deep feature mining and matching diagnosis are performed on the fault time-series feature sequence to determine the fault diagnosis conclusion for the batch circulating grain dryer. The fault diagnosis conclusion includes the specific fault type and fault optimization scheme. This application acquires multidimensional operational data including temperature, moisture content, wind speed, and circulation parameters. By performing standardized preprocessing on this data, a regularized fault feature dataset is constructed. Fault time-series feature sequences are built by generating corresponding parameter events according to data type and arranging them in order based on timestamps, accurately depicting the gradual change pattern of progressive faults. Through deep feature mining and matching diagnosis of the fault time-series feature sequences, weak abnormal signals and gradual trends are identified, thereby determining the specific fault type and providing a fault optimization solution. This method addresses the significant deficiencies in the identification of progressive faults and early-stage weak faults.

[0007] In a second aspect, embodiments of this application provide a fault diagnosis device for a batch circulating grain dryer, applied to a batch circulating grain dryer, for implementing the fault diagnosis method for a batch circulating grain dryer described in any one of the first aspects above. The fault diagnosis device for the batch circulating grain dryer includes: The acquisition unit is used to acquire multi-dimensional operating data of the batch circulating grain dryer when it completes the drying of a batch of grain; wherein, the multi-dimensional operating data includes equipment temperature parameters, moisture content parameters, wind speed parameters, and circulation parameters; The normalization unit is used to preprocess the temperature parameters, moisture content parameters, wind speed parameters, and circulation parameters to obtain a normalized fault feature dataset; wherein, the normalized fault feature dataset reflects the characteristic data set of the batch circulating grain dryer's operating status and potential faults; The sequence unit is used to generate corresponding parameter events according to data type based on the normalized fault feature dataset, and then arrange them in order according to the timestamp of each parameter event to construct a fault time sequence feature sequence; wherein, the parameter events include generating temperature events according to the temperature parameters, generating moisture content events according to the moisture content parameters, generating wind speed events according to the wind speed parameters, and generating cycle events according to the cycle parameters. The diagnostic unit is used to perform deep feature mining and matching diagnosis on the fault time sequence to determine the fault diagnosis conclusion of the batch circulating grain dryer; wherein, the fault diagnosis conclusion includes the specific fault type and fault optimization scheme.

[0008] Thirdly, embodiments of this application provide a batch circulating grain dryer, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the method as described in any of the foregoing aspects.

[0009] Fourthly, embodiments of this application provide a computer program product that, when run on a batch circulating grain dryer, causes the batch circulating grain dryer to perform the method described in any one of the first aspects above.

[0010] It is understood that the beneficial effects of the second to fourth aspects mentioned above can be found in the relevant descriptions in the first aspect mentioned above, and will not be repeated here. Attached Figure Description

[0011] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0012] Figure 1 This is a schematic flowchart of a fault diagnosis method for a batch circulating grain dryer provided in an embodiment of this application; Figure 2 This is a schematic diagram of the frame of the batch circulating grain dryer provided in the embodiments of this application; Figure 3 This is a schematic diagram of the operation of a fault diagnosis method for a batch circulating grain dryer provided in an embodiment of this application; Figure 4 This is a schematic diagram of the fault diagnosis device for a batch circulating grain dryer provided in the embodiments of this application; Figure 5 This is a schematic diagram of the structure of the batch circulating grain dryer provided in the embodiments of this application. Detailed Implementation

[0013] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0014] It should be understood that, when used in this application specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or a collection thereof.

[0015] It should also be understood that the term “and / or” as used in this application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0016] As used in this application specification and the appended claims, the term "if" may be interpreted, depending on the context, as "when," "once," "in response to determination," or "in response to detection." Similarly, the phrase "if determined" or "if detected [the described condition or event]" may be interpreted, depending on the context, as meaning "once determined," "in response to determination," "once detected [the described condition or event]," or "in response to detection [the described condition or event]."

[0017] Furthermore, in the description of this application and the appended claims, the terms "first," "second," "third," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0018] References to "one embodiment" or "some embodiments" as described in this specification mean that one or more embodiments of this application include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.

[0019] Traditional fault diagnosis methods in related technologies often rely on human experience and single-point threshold alarms, which are difficult to meet the needs of modern intelligent and continuous drying operations, especially in the identification of gradual faults and early weak faults. Specifically, the fault diagnosis methods for batch circulating grain dryers in related technologies are relatively crude, mainly falling into two categories: one relies entirely on manual inspections and experience-based judgment by operators. Operators need to periodically visually inspect and perform simple instrument checks on core components such as the dryer's main support structure, grain circulation system, and hot air drying system, relying on their own operational experience to determine if the equipment is faulty. For example, they might rely on touching the drying cylinder shell to judge whether the temperature is abnormal, listen to the fan's operating sound to judge whether the wind speed is stable, and observe the grain discharge state to judge whether the moisture content meets the standard. This method is highly dependent on the operator's experience and sense of responsibility, resulting in low diagnostic efficiency and failing to detect gradual faults (such as a slow decrease in fan speed or a gradual decrease in wind speed caused by slight dust accumulation in the hot air duct) and early, subtle faults (such as parameter deviations caused by slight sensor drift or abnormal heat loss caused by slight damage to the drying cylinder insulation layer). Often, by the time the fault worsens and the equipment becomes unusable... One type of fault diagnosis involves simple single-point threshold alarms. This method involves setting a single sensor at a critical point in the dryer and pre-setting a fixed parameter threshold. When the parameter collected by the sensor exceeds the threshold, an alarm is triggered. For example, the preset hot air outlet temperature threshold is 55-65℃. When the temperature exceeds this range, an alarm is triggered to indicate an abnormal temperature. This method can only identify obvious faults with sudden parameter changes, but cannot identify gradual faults with slow parameter changes. It also does not consider the correlation between parameters (such as the synergistic effect of abnormal wind speed and abnormal temperature, and the linkage between circulation parameters and moisture content parameters). This can easily lead to false alarms and missed alarms. At the same time, it cannot clearly identify the fault type and specific fault location. Operators need to spend a lot of time troubleshooting the cause of the fault. It is difficult to adapt to the continuous and intelligent drying operation requirements of modern batch circulating grain dryers and cannot achieve early warning and accurate diagnosis of faults.

[0020] To address the aforementioned issues, this application provides a fault diagnosis method for a batch circulating grain dryer. The method includes: acquiring multi-dimensional operational data of the batch circulating grain dryer when it completes drying a batch of grain; wherein the multi-dimensional operational data includes equipment temperature parameters, moisture content parameters, wind speed parameters, and circulation parameters; preprocessing the temperature parameters, moisture content parameters, wind speed parameters, and circulation parameters to obtain a regularized fault feature dataset; wherein the regularized fault feature dataset reflects the characteristic data set of the batch circulating grain dryer's operating status and potential faults; based on the regularized fault feature dataset, generating corresponding parameter events according to data types, and then arranging them in an ordered manner according to the timestamps of each parameter event to construct a fault time-series feature sequence; wherein the parameter events include temperature events generated based on temperature parameters, moisture content events generated based on moisture content parameters, wind speed events generated based on wind speed parameters, and circulation events generated based on circulation parameters; performing deep feature mining and matching diagnosis on the fault time-series feature sequence to determine the fault diagnosis conclusion of the batch circulating grain dryer; wherein the fault diagnosis conclusion includes the specific fault type and fault optimization scheme. This application acquires multidimensional operational data including temperature, moisture content, wind speed, and circulation parameters. By performing standardized preprocessing on this data, a regularized fault feature dataset is constructed. Fault time-series feature sequences are built by generating corresponding parameter events according to data type and arranging them in order based on timestamps, accurately depicting the gradual change pattern of progressive faults. Through deep feature mining and matching diagnosis of the fault time-series feature sequences, weak abnormal signals and gradual trends are identified, thereby determining the specific fault type and providing a fault optimization solution. This method addresses the significant deficiencies in the identification of progressive faults and early-stage weak faults.

[0021] The fault diagnosis method for batch circulating grain dryers provided in this application embodiment can be applied to batch circulating grain dryers. In this case, the batch circulating grain dryer is the subject of execution of the fault diagnosis method for batch circulating grain dryers provided in this application embodiment. This application embodiment does not impose any restrictions on the specific type of batch circulating grain dryer.

[0022] Please see Figure 2 , Figure 2This is a schematic diagram of the frame of a batch circulating grain dryer provided in an embodiment of this application. For example, the batch circulating grain dryer mainly includes a main body and temperature sensors, wind speed sensors, moisture content sensors, speed sensors, and position sensors. The temperature sensors are installed at the hot air furnace outlet, inside the drying cylinder, in the hot air duct, and at the grain discharge port to collect temperature parameters. The wind speed sensors are installed inside the hot air duct and at the fan outlet to collect wind speed parameters. The moisture content sensors are installed at the feed hopper and discharge hopper to collect moisture content parameters. The speed sensors are installed on the motor shafts of the circulating elevator, screw conveyor, and fan to collect circulation parameters. The batch circulating grain dryer also includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the method described in any of the above aspects.

[0023] To better understand the fault diagnosis method for batch circulating grain dryers provided in this application embodiment, the specific implementation process of the fault diagnosis method for batch circulating grain dryers provided in this application embodiment will be described by way of example below.

[0024] Figure 1 A flowchart illustrating the fault diagnosis method for a batch circulating grain dryer provided in an embodiment of this application is shown. Figure 3 This is a schematic diagram illustrating the operation of a fault diagnosis method for a batch circulating grain dryer according to an embodiment of this application. The fault diagnosis method for the batch circulating grain dryer includes: S100 acquires multi-dimensional operational data of the batch circulating grain dryer when it completes the drying of a batch of grain. This multi-dimensional operational data includes equipment temperature parameters, moisture content parameters, wind speed parameters, and circulation parameters.

[0025] It is understandable that multidimensional operational data is a collection of various types of data reflecting the operational status of the equipment during its operating cycle. These include: equipment temperature parameters, which refer to various temperature data related to the operation of the batch circulating grain dryer, such as the furnace temperature, hot air outlet temperature, grain discharge temperature, and equipment shell temperature; moisture content parameters, which refer to data on changes in the moisture content of the grain during the drying process, such as the initial moisture content when the grain is fed in, the real-time moisture content during the drying process, and the final moisture content after drying; wind speed parameters, which refer to the operating wind speed data of the fans used to transport hot air within the batch circulating grain dryer, such as the inlet wind speed of the bottom parallel fans and the flow wind speed of hot air within the drying cylinder; and circulation parameters, which refer to data related to the circulation of grain within the batch circulating grain dryer, such as the grain circulation conveying speed, the time of a single circulation, and the number of circulations. For example, when a batch of wheat is put into a batch circulating grain dryer, the data collected from the start of the equipment to the time when the wheat moisture content drops from 25% to 13% (meeting the safe storage standard) includes furnace temperature of 85℃, hot air outlet temperature of 62℃, initial wheat moisture content of 25%, real-time moisture content of 18%, fan speed of 3.2m / s, and grain circulation speed of 0.8m / min. Data is collected once every minute, and a total of 120 sets of data are collected during the entire drying process. Among them, the furnace temperature of 85℃ (collected at 10:01) and the hot air outlet temperature of 62℃ (collected at 10:01) together constitute multi-dimensional operating data.

[0026] S200 preprocesses temperature, moisture content, wind speed, and circulation parameters to obtain a regularized fault feature dataset. This regularized fault feature dataset reflects the operating status and potential faults of the batch circulating grain dryer.

[0027] Preprocessing refers to a series of sorting and optimization processes performed on the collected raw multidimensional operational data to eliminate interference factors and make the data more suitable for subsequent fault diagnosis. A standardized fault feature dataset is a standardized data set that, after preprocessing, clearly and accurately reflects the equipment's operating status and potential fault hazards. Each set of data corresponds to a characteristic of a specific operational stage of the equipment and can be directly used for subsequent fault feature mining. For example, the collected raw data may contain issues such as instantaneous fluctuations in fan speed and missing grain moisture content data. After preprocessing to eliminate these interferences, the resulting standardized data set, showing furnace temperature of 82-86℃, hot air outlet temperature of 60-63℃, wheat moisture content steadily decreasing from 25% to 13%, fan speed stable at 3.1-3.3 m / s, and grain circulation speed stable at 0.78-0.82 m / min, is the standardized fault feature dataset. This dataset clearly shows whether the equipment has potential fault hazards such as abnormal temperature, unstable wind speed, or abnormal decrease in moisture content.

[0028] As an optional embodiment of this application, in step S200, temperature parameters, moisture content parameters, wind speed parameters, and circulation parameters are preprocessed to obtain a normalized fault feature dataset, including: S210, based on temperature parameters, moisture content parameters, wind speed parameters, and circulation parameters, obtains a fault diagnosis standard dataset.

[0029] It can be understood that the fault diagnosis standard dataset refers to the basic dataset formed by filtering out relevant fault diagnosis data and removing irrelevant redundant data from the collected original multidimensional operational data. It serves as the basis for subsequent data normalization and noise removal. For example, from the original multidimensional operational data, all temperature, moisture content, wind speed, and circulation parameter data during the wheat drying operation (within 2 hours from grain input to discharge) are filtered out. The standby temperature (25℃, irrelevant to the drying operation) in the 10 minutes before equipment startup and the redundant fan wind speed data (3.2m / s, no grain drying at this time, not reflecting potential faults) in the 5 minutes after drying are removed. The resulting continuous operational data within 2 hours is the fault diagnosis standard dataset. Based on the multidimensional operational data of a single batch drying cycle, the fault diagnosis standard dataset is obtained by filtering according to the above screening criteria and removal range, ensuring that the fault diagnosis standard dataset only contains valid operational data during the drying operation.

[0030] S220, numerical normalization is performed on the fault diagnosis standard dataset to obtain the fault diagnosis normalized dataset.

[0031] Numerical normalization refers to the process of converting various parameters with different dimensions and numerical ranges in a fault diagnosis standard dataset into a unified numerical range using a pre-defined mathematical algorithm. The core purpose of numerical normalization is to eliminate differences in dimensions and numerical values ​​between different parameters, avoiding deviations in subsequent feature mining due to excessively large parameter value differences. For example, in a fault diagnosis standard dataset, the numerical range of temperature parameters is 60-86℃, the range of moisture content parameters is 13-25%, the range of wind speed parameters is 3.1-3.3 m / s, and the range of circulation speed is 0.78-0.82 m / min. After normalization, all parameter values ​​are uniformly converted to the [0,1] interval. For example, 86℃ corresponds to 1, 60℃ corresponds to 0, 25% corresponds to 1, 13% corresponds to 0, and 3.3 m / s corresponds to 1 and 3.1 m / s corresponds to 0. The converted dataset is the fault diagnosis normalized dataset, at which point various parameters can be subjected to unified feature analysis and calculation.

[0032] S230, an outlier detection algorithm is used to identify and remove noisy data from the normalized fault diagnosis dataset to determine the normalized fault feature dataset.

[0033] As we can understand it, outlier detection algorithms refer to algorithms used to identify abnormal data (i.e., noisy data) in a dataset that deviates from the normal data range or does not conform to the normal operating rules of the equipment. Common examples include mean-standard deviation-based detection algorithms and box plot detection algorithms. Noisy data refers to abnormal data that deviates from the normal operating state of the equipment due to factors such as sensor failure and external interference. This type of data interferes with the accuracy of fault diagnosis and needs to be removed. For example, in a normalized dataset for fault diagnosis, the normalized values ​​of wind speed parameters are mostly stable between 0.5 and 0.6 (corresponding to the original wind speed of 3.15-3.25 m / s). If an outlier with a value of 0.9 (corresponding to the original wind speed of 3.45 m / s) appears, this outlier data does not conform to the wind speed pattern during normal equipment operation and is therefore considered noisy data. After the outlier data is identified and removed by the outlier detection algorithm, the remaining data that conforms to the normal pattern can be further processed to obtain the normalized fault feature dataset.

[0034] By adopting the above steps S210 to S230, it is helpful to solve the problems of messy, interference and incomparable parameters in the original multidimensional operating data, and provide high-quality data support for the subsequent construction of fault time series feature sequences and deep feature mining, so as to avoid fault diagnosis deviation, missed diagnosis or misdiagnosis due to data quality problems, and improve the accuracy and reliability of fault diagnosis.

[0035] In one possible implementation, S230, an outlier detection algorithm is used to identify and remove noisy data from the fault diagnosis normalized dataset to determine the normalized fault feature dataset, including: S231, calculate the mean and standard deviation of each multidimensional running data in the fault diagnosis normalized dataset, and determine the noise identification threshold range.

[0036] The mean refers to the average value of all data for a certain type of parameter in the fault diagnosis normalized dataset, reflecting the normal operating center level of that type of parameter. The standard deviation refers to the degree to which all data for a certain type of parameter in the fault diagnosis normalized dataset deviates from the mean, reflecting the dispersion of the data. The smaller the standard deviation, the more concentrated the data; conversely, the larger the standard deviation, the more dispersed the data. The noise identification threshold range refers to the numerical range determined based on the mean and standard deviation to judge whether data is noise. It is usually set as [mean - 3 × standard deviation, mean + 3 × standard deviation] (i.e., the 3σ principle, which can cover 99.73% of normal data). Data exceeding the noise identification threshold range can be preliminarily judged as suspected noise. For example, for the wind speed parameters in the fault diagnosis normalized dataset, the mean of all wind speed data is calculated to be 0.55 and the standard deviation is 0.03. According to the 3σ principle, the noise identification threshold range is [0.55-3×0.03, 0.55+3×0.03], that is, [0.46, 0.64]. Wind speed data that exceed the identification threshold range (such as 0.9, 0.4) will be initially judged as suspected noise.

[0037] S232 marks the multi-dimensional running data in the fault diagnosis normalization dataset that exceed the noise identification threshold range as suspected noise.

[0038] It is understandable that suspected noise refers to data that has been initially identified as abnormal, but has not yet been confirmed as actual noise. The purpose of marking suspected noise is to provide a clear target for subsequent verification and screening, and to avoid directly rejecting data that may reflect real equipment abnormalities. For example, based on the wind speed parameter noise identification threshold range [0.46, 0.64] determined by S231, if wind speed values ​​of 0.9 and 0.4 appear in the fault diagnosis normalized dataset, both of which exceed the threshold range, 0.9 and 0.4 will be marked as suspected noise.

[0039] S233 uses a sliding window filtering algorithm to verify suspected noise points, removes suspected noise points caused by sensor anomalies, and obtains the missing dataset.

[0040] The sliding window filtering algorithm can be understood as follows: it sets a fixed-length window (i.e., an interval containing a certain number of continuous data points), slides the window across the dataset, analyzes the distribution patterns of the data within the window, and then verifies whether suspected noise points are actually noise points. Suspected noise points caused by sensor anomalies refer to isolated abnormal data caused by momentary sensor malfunctions, signal interference, etc. This type of data does not reflect the actual operational anomalies of the equipment and needs to be removed. The blank dataset refers to the dataset formed by the blank positions (i.e., the positions of the removed data) that appear in the dataset after removing suspected noise points. For example, if the sliding window length is set to 5 (i.e., each window contains 5 consecutive wind speed data points), when the window slides to the position containing the suspected noise point 0.9, the other four data points within the window are 0.54, 0.55, 0.56, and 0.55, all within the normal threshold range. Only 0.9 is a suspected noise point, which can be determined to be noise caused by momentary sensor interference and removed. After removal, a blank dataset is formed.

[0041] For example, in S233, a sliding window filtering algorithm is used to verify suspected noise points, remove suspected noise points caused by sensor anomalies, and obtain a missing dataset, including: S2331, Set a preset length sliding window.

[0042] It is understandable that a preset-length sliding window refers to a fixed-length window set in advance based on the operating patterns and data acquisition frequency of the batch circulating grain dryer, used to analyze the continuity of data. The window length setting needs to balance data continuity and the accuracy of anomaly detection, and can be set to 3-10 data points (i.e., containing 3-10 consecutive multi-dimensional operational data). For example, considering the grain drying cycle (2 hours) and data acquisition frequency (data acquired every 1 minute) of the batch circulating grain dryer, setting the sliding window length to 5 means that each window contains 5 consecutive 1-minute data acquisitions, which can ensure the continuity of data within the window while quickly identifying isolated suspected noise points.

[0043] S2332, if only a single multidimensional running data point in the sliding window is suspected noise, the suspected noise point is determined to be interference noise data generated by sensor interference and is removed to obtain the missing dataset.

[0044] It's understandable that if only a single data point within the window is suspected noise, while the rest are normal, it indicates that the suspected noise is isolated and doesn't conform to the continuous evolution pattern of equipment malfunctions. Therefore, it's identified as interference noise caused by sensor interference and removed. After removal, the location of the suspected noise becomes blank, forming a missing dataset. For example, if the five normalized wind speed data points within the sliding window are 0.54, 0.55, 0.9 (suspected noise), 0.56, and 0.55, only 0.9 is suspected noise, while the other four are normal data. This indicates that 0.9 is an isolated noise point caused by transient sensor interference and is removed. After removal, a missing dataset containing blank positions is formed.

[0045] By employing the steps S2331 to S2332 described above, the problem of confusing isolated interference noise with real abnormal data of the equipment can be solved, further improving the purity of the data. This provides an accurate basis for filling in missing data and determining the normalized fault feature dataset, ensuring the accuracy of subsequent fault diagnosis.

[0046] S234, the mean of adjacent multidimensional running data within the sliding window is used to fill the missing positions in the missing dataset to obtain a normalized fault feature dataset.

[0047] It can be understood that the mean of adjacent multidimensional data within a sliding window refers to the average of the other continuous normal data within the sliding window containing the missing data. Filling the missing position means replacing the missing data with the mean, restoring the continuity of the dataset and preventing breaks in subsequent feature mining and time series construction due to data gaps. For example, in a missing dataset, the normal wind speed data within the sliding window containing a certain missing position are 0.54, 0.55, 0.56, and 0.55. The calculated mean is (0.54 + 0.55 + 0.56 + 0.55) ÷ 4 = 0.55. Filling the missing position with 0.55 makes the filled data continuous with other normal data, forming a complete, continuous, and clean normalized fault feature dataset.

[0048] By adopting the above steps S231 to S234, the continuity, purity and standardization of the regularized fault feature dataset can be improved, providing high-quality data support for the subsequent construction of fault time-series feature sequences, deep feature mining and matching diagnosis, avoiding fault diagnosis deviations and missed diagnoses caused by data noise and gaps, and further improving the accuracy and reliability of fault diagnosis. In particular, it can accurately capture the feature data of early weak faults, laying the foundation for early fault identification.

[0049] S300, based on a normalized fault feature dataset, generates corresponding parameter events according to data type, and then arranges them in order according to the timestamps of each parameter event to construct a fault time-series feature sequence. Among them, parameter events include temperature events generated based on temperature parameters, moisture content events generated based on moisture content parameters, wind speed events generated based on wind speed parameters, and cycle events generated based on cycle parameters.

[0050] It can be understood that data types refer to the four main categories of parameters (temperature, moisture content, wind speed, and circulation) in the normalized fault feature dataset, with each category corresponding to a data type. Parameter events refer to transforming the operational data of each parameter category into event representations with clear physical meaning that can be used for fault analysis. Essentially, it's a mapping between parameter data and operational status. Specifically: Temperature events are events generated based on temperature parameters that reflect the equipment's temperature operating status (e.g., normal temperature event, high temperature event, low temperature event). Moisture content events are events generated based on moisture content parameters that reflect changes in grain moisture content (e.g., normal moisture content decrease event, slow moisture content decrease event, abnormal moisture content fluctuation event). Wind speed events are events generated based on wind speed parameters that reflect the fan's wind speed operating status (e.g., stable wind speed event, high wind speed event, low wind speed event). Circulation events are events generated based on circulation parameters that reflect the grain's circulation operating status (e.g., normal circulation speed event, excessively fast circulation speed event, abnormal number of circulation events). A timestamp refers to the specific time each parameter data was collected (accurate to the moment of collection; for example, during the drying process of a batch of wheat, temperature data collected at 10:01 corresponds to timestamp 10:01). The fault timing feature sequence refers to a continuous and ordered sequence of events formed by arranging all parameter events in chronological order according to their timestamps. This sequence can intuitively reflect the temporal distribution and evolution of various parameter events during equipment operation. For example, the normalized fault feature dataset contains continuous data from 10:00 to 12:00 during the wheat drying process. Parameter events are generated according to data type: at 10:00, a normal temperature event (furnace temperature 84℃), a normal decrease in moisture content event (moisture content 24%), a stable wind speed event (wind speed 3.2m / s), and a normal circulation speed event (circulation speed 0.8m / min) are generated. At 10:05, a normal temperature event (furnace temperature 85℃), a normal decrease in moisture content event (moisture content 23%), a stable wind speed event (wind speed 3.15m / s), and a normal circulation speed event (circulation speed 0.8m / min) are generated. Arrange all these parameter events in the order of timestamps 10:00, 10:01, 10:02...12:00 to obtain the fault timing feature sequence. The fault timing feature sequence can clearly show the changes of various parameter events throughout the drying process.

[0051] As an optional embodiment of this application, in step S300, based on the normalized fault feature dataset, corresponding parameter events are generated according to data type, and then arranged in order according to the timestamps of each parameter event to construct a fault time-series feature sequence, including: S310 generates parameter events based on a normalized fault feature dataset, categorized by data type. These parameter events include temperature events, moisture content events, wind speed events, and cycle events.

[0052] It is understandable that generation based on data type involves filtering and analyzing the four main categories of parameters (temperature, moisture content, wind speed, and circulation) in the normalized fault feature dataset, and generating corresponding parameter events for each category. This ensures that the operational characteristics of each parameter category can be represented independently and accurately, avoiding confusion between events of different parameter types. Specifically, temperature event generation considers the normal threshold for temperature parameters (e.g., the normal threshold for furnace temperature is 82-86℃). If the temperature is within the threshold at a given moment, a normal temperature event is generated; otherwise, an abnormal temperature event is generated. Moisture content event generation considers the normal variation pattern of moisture content (e.g., a decrease of approximately 1% every 5 minutes). If the pattern is followed, a normal moisture content decrease event is generated; otherwise, an abnormal event is generated. Wind speed event generation considers the normal wind speed threshold (e.g., 3.1-3.3 m / s). If the threshold is met, a stable wind speed event is generated; otherwise, an abnormal wind speed event is generated. Circulation event generation considers the normal range of circulation parameters (e.g., circulation speed 0.78-0.82 m / min). If the range is met, a normal circulation event is generated; otherwise, an abnormal circulation event is generated. For example, in the normalized fault feature dataset, if the temperature parameter of 10:10 is 87℃ (exceeding the normal threshold of 82-86℃), then a high temperature event is generated. If the moisture content parameter is 22% (a 1% decrease from 23% in 10:05, which is within the normal range), then a normal moisture content decrease event is generated. If the wind speed parameter is 3.2m / s (within the normal threshold), then a stable wind speed event is generated. If the circulation speed is 0.81m / min (within the normal range), then a normal circulation speed event is generated. The generation of four types of parameter events is completed by classifying them according to data type.

[0053] S320 extracts the timestamps of temperature events, moisture content events, wind speed events, and cycle events to obtain a time feature data sequence.

[0054] As can be understood, a timestamp refers to the acquisition time information corresponding to each generated parameter event. Each parameter event is uniquely bound to a timestamp to characterize the occurrence time of the parameter event. A time feature data sequence refers to an ordered set of timestamps formed by organizing all extracted timestamps according to certain rules. The purpose of the time feature data sequence is to provide a time reference for the subsequent sorting of parameter events and the construction of fault timing feature sequences. For example, the events of 10:10 (higher temperature), 10:10 (normal decrease in moisture content), 10:10 (stable wind speed), and 10:10 (normal circulation speed) are all bound to the timestamp 10:10. The four types of parameter events generated at 10:15 are all bound to the timestamp 10:15. The ordered set of timestamps formed by extracting all these timestamps (10:10, 10:10, 10:10, 10:10, 10:15 ...

[0055] In one possible implementation, S320 extracts the timestamps of temperature events, moisture content events, wind speed events, and cycle events to obtain a time feature data sequence, including: S321 Extract the timestamps bound to the temperature event, moisture content event, wind speed event, and cycle event respectively to form an independent set of timestamps corresponding to the four types of events.

[0056] This can be understood as extracting timestamps separately for each of the four types of parameter events (temperature, moisture content, wind speed, and cycle time), without mixing timestamps from other event types. Each independent timestamp set refers to a separate set of timestamps corresponding to each type of parameter event. The four parameter events correspond to four independent timestamp sets, with each set containing timestamps representing the occurrence time of the parameter event. For example, the timestamp set for the temperature event is {10:00, 10:05, 10:10, 10:15…12:00}, with each timestamp representing the occurrence time of a temperature event. The timestamp set for the moisture content event is {10:00, 10:05, 10:10, 10:15…12:00}. The timestamp set for the wind speed event is {10:00, 10:05, 10:10, 10:15…12:00}. The timestamp set corresponding to the cyclic events is {10:00, 10:05, 10:10, 10:15...12:00}. The four sets are independent of each other and correspond to the time distribution of the four types of events.

[0057] S322, remove invalid timestamps from each independent timestamp set to obtain a valid timestamp set.

[0058] Invalid timestamps are understood to be meaningless and unable to reflect the actual occurrence time of a parameter event. They mainly fall into two categories: First, invalid timestamps caused by abnormal data acquisition (such as time discrepancies due to sensor malfunctions, e.g., data acquired at 10:00 is bound to a timestamp of 15:00). Second, timestamps that exceed the current batch drying cycle (e.g., if the current batch drying cycle is 10:00-12:00, timestamps of 9:50 or 12:10 are invalid). The set of valid timestamps refers to the set of timestamps that accurately reflect the occurrence time of the parameter event and are within the current drying cycle after removing invalid timestamps. For example, if the set of independent timestamps for a temperature event contains timestamps of 15:00 (time discrepancy) and 9:50 (exceeding the drying cycle of 10:00-12:00), both are invalid timestamps. After removing the invalid timestamps, the remaining set {10:00, 10:05, 10:10...12:00} is the set of valid timestamps corresponding to the temperature event. Similarly, invalid timestamps are removed from the independent timestamp sets of moisture content, wind speed, and cyclic events to obtain their respective valid timestamp sets.

[0059] S323, based on the set of valid timestamps, forms an ordered sequence of time feature data.

[0060] It can be understood that the effective timestamp set refers to integrating the effective timestamps corresponding to the four types of parameter events, removing duplicate timestamps (because four types of parameter data are collected at the same time, generating four types of parameter events, corresponding to the same timestamp), and then sorting them in chronological order to finally form a continuous and ordered time series. An ordered time feature data sequence refers to a time set arranged from earliest to latest, without repetition or invalid times, and is the core time benchmark for subsequent parameter event sorting and fault timing feature sequence construction. For example, after integrating the effective timestamps of the four types of parameter events and removing duplicate timestamps, the set is arranged in chronological order as {10:00, 10:01, 10:02…12:00}. The set {10:00, 10:01, 10:02…12:00} is the ordered time feature data sequence, and all subsequent parameter events will be arranged according to the chronological order of the time feature data sequence.

[0061] By adopting the above steps S321 to S323, it is helpful to effectively solve the problems of timestamp chaos, invalid time interference, and unclear time order, and provide a standardized and reliable time benchmark for subsequent parameter events to be sorted by time and the accurate construction of fault time sequence feature sequences, avoiding errors in time sequence construction due to time chaos.

[0062] S330 extracts the feature values ​​of temperature events, moisture content events, wind speed events, and cycle events, and combines them with the time feature data sequence to obtain the fault time sequence feature sequence.

[0063] As can be understood, feature values ​​refer to the specific parameter values ​​corresponding to each parameter event, and are the core data characterizing the specific state of the parameter event. Combining time-series feature data involves binding the feature values ​​of each type of parameter event with the corresponding timestamps in the time-series feature data, and then arranging all parameter events bound with feature values ​​and timestamps in chronological order according to the timestamps. Fault time-series feature sequences refer to the final continuous and ordered sequence containing time, parameter event type, and feature values. This clearly reflects the operating status and changing trends of various parameters at different time points, providing time-series feature data for subsequent in-depth feature mining. For example, the timestamp 10:00 in the time-series feature data corresponds to the following bound feature values: temperature event 84℃ (normalized to 0.92), moisture content event 24% (normalized to 0.92), wind speed event 3.2m / s (normalized to 0.5), and cycle event 0.8m / min (normalized to 0.5). The characteristic values ​​bound to the timestamp 10:05 are: temperature event 85℃ (normalized 0.96), moisture content event 23% (normalized 0.83), wind speed event 3.15m / s (normalized 0.42), and cycle event 0.8m / min (normalized 0.5). By arranging the bound data corresponding to all timestamps in chronological order, the fault time-series characteristic sequence can be obtained. This sequence visually shows the changes in the characteristic values ​​of various parameters over time from 10:00 to 12:00.

[0064] By adopting the above steps S310 to S330, it is helpful to solve the problems of disordered parameter events, lack of time correlation, and inability to reflect the fault evolution process. It realizes the time-series integration of parameter events, clearly depicts the changing pattern of the operating status of various parameters over time, and can especially capture the time-series evolution characteristics of early weak faults and gradual faults.

[0065] In one possible implementation, S330, the feature values ​​of temperature events, moisture content events, wind speed events, and cycle events are extracted and combined with the time feature data sequence to obtain the fault time sequence feature sequence, including: S331: Extract the feature values ​​corresponding to temperature events, moisture content events, wind speed events, and cycle events, and bind the feature values ​​with the timestamps of the corresponding parameter events to form binary association groups.

[0066] It can be understood that the corresponding feature values ​​refer to the specific parameter values ​​(normalized values ​​can be used for easier subsequent calculations) behind each temperature event, moisture content event, wind speed event, and cycle event. Each parameter event corresponds to a unique feature value. A binary association group refers to a binary set consisting of a timestamp and a feature value. One element is the timestamp (the time of the event), and the other element is the feature value of the parameter event corresponding to the timestamp. The binary association group clearly reflects the specific operating state of a certain type of parameter at a certain moment. For example, the feature value corresponding to the temperature event at 10:10 is 87℃ (normalized to 1.0). Binding the feature value 87℃ (normalized to 1.0) with the timestamp 10:10 forms a binary association group (10:10, 1.0). The feature value corresponding to the moisture content event at 10:10 is 22% (normalized to 0.75), which, after binding, forms a binary association group (10:10, 0.75). Similarly, the wind speed event of 10:10 (3.2 m / s, normalized to 0.5) forms (10:10, 0.5), and the cyclic event (0.81 m / min, normalized to 0.55) forms (10:10, 0.55). Each parameter event corresponds to an independent binary association group.

[0067] S332, sort all binary association groups according to their timestamps to obtain the standard association groups.

[0068] It can be understood that all binary association groups refer to all binary association groups generated by the four types of parameter events (each timestamp corresponds to 4 binary association groups, each corresponding to one of the four types of parameter events). Sorting by timestamp means using the timestamps within the binary association groups as a basis, arranging all association groups uniformly in chronological order, with the 4 binary association groups corresponding to the same timestamp grouped together to form an ordered set of association groups. A standard association group refers to the set of binary association groups formed after sorting, arranged in chronological order. The binary association groups corresponding to the four types of parameter events at each time point are arranged centrally, clearly showing the operational status of each type of parameter at the same moment. For example, after sorting all binary association groups by timestamp, the four binary association groups corresponding to 10:00 (10:00, 0.92), (10:00, 0.92), (10:00, 0.5), and (10:00, 0.5) are grouped together, and the four binary association groups corresponding to 10:05 are grouped together. Arranging them in this way, the ordered set formed is the standard association group.

[0069] S333, based on all standard association groups, constructs a fault timing feature sequence. This fault timing feature sequence clearly reflects the temporal distribution of each parameter event and the corresponding changes in feature values.

[0070] It is understandable that, based on all standard association groups, the sorted standard association groups are integrated, redundant information is removed, and the binary association groups of the four types of parameter events corresponding to each timestamp are retained to form a continuous and complete time series sequence. The fault time series feature sequence refers to the structured sequence that is finally integrated and contains time, parameter type, and feature value. The core feature of the fault time series feature sequence is that it is arranged in chronological order, which can intuitively reflect the operating status of various parameters at each time point, as well as the continuous change trend of the feature values ​​of various parameters over time, providing clear time series data support for subsequent in-depth feature mining. For example, after integrating the standard association group, the fault timing feature sequence is as follows: 10:00 [temperature (0.92), moisture content (0.92), wind speed (0.5), circulation (0.5)] to 10:05 [temperature (0.96), moisture content (0.83), wind speed (0.42), circulation (0.5)] to 10:10 [temperature (1.0), moisture content (0.75), wind speed (0.5), circulation (0.55)] to ... to 12:00 [temperature (0.88), moisture content (0.0), wind speed (0.5), circulation (0.5)]. The fault timing feature sequence can clearly show the changes in the characteristic values ​​of parameters such as temperature and moisture content over time during the entire drying process, as well as whether there are any abnormal fluctuations.

[0071] By adopting the above steps S331 to S333, it is helpful to solve the problems of parameter events being disconnected from time, feature values ​​lacking time markers, and failing to reflect the temporal change pattern. This makes the fault temporal feature sequence more organized and readable, clearly depicting the temporal evolution process of various parameters. It is convenient to subsequently mine the features of weak faults and gradual faults from the temporal dimension, avoid the omission of fault features due to temporal disorder, and further improve the accuracy of fault diagnosis.

[0072] S400 performs deep feature mining and matching diagnosis on the fault time sequence to determine the fault diagnosis conclusion of the batch circulating grain dryer. The fault diagnosis conclusion includes the specific fault type and fault optimization scheme.

[0073] Deep feature mining refers to using pre-defined algorithms (such as deep learning and machine learning algorithms) to deeply analyze the temporal information and parameter correlation information in the fault time-series feature sequence, extracting the fault features hidden in the sequence (including single-event abnormal features and multi-event correlated abnormal features). These features are the core basis for identifying fault types. Matching diagnosis refers to comparing and matching the mined fault features with standard fault features in a pre-defined fault feature library to determine whether the current equipment is operating with a fault and what kind of fault it is. Specific fault types refer to the actual types of faults existing in the equipment. Considering the operating characteristics of batch circulating grain dryers, common faults include abnormal temperature faults (such as excessively high furnace temperature and excessively low hot air outlet temperature), abnormal moisture content faults (such as excessively slow moisture content decrease and excessive moisture content fluctuation), abnormal wind speed faults (such as excessively low wind speed and unstable wind speed), and abnormal circulation faults (such as excessively fast circulation speed and insufficient number of circulations). Fault optimization solutions refer to the operable and implementable solutions proposed for the specific fault types identified, used to eliminate faults and restore normal equipment operation. For example, after performing deep feature mining on the fault time sequence, it was found that during the period of 10:10-10:30, the feature value of the temperature event was consistently higher than the normal threshold (greater than 1.0 after normalization), and the feature value of the wind speed event was consistently lower than the normal threshold (less than 0.3 after normalization). These two features were matched with the standard features of hot air circulation problems in the preset fault feature library. If the similarity reached 95%, the specific fault type was determined to be hot air circulation problems. The fault optimization solution was to check the operating status of the bottom parallel fans and clean the dust accumulated in the hot air ducts to ensure smooth hot air circulation.

[0074] As an optional embodiment of this application, in step S400, deep feature mining and matching diagnosis are performed on the fault timing feature sequence to determine the fault diagnosis conclusion of the batch circulating grain dryer, including: S410 performs deep feature mining on the fault timing feature sequence and integrates them to form an event feature vector. The event feature vector is a quantified feature vector used to characterize the overall operating status of the equipment.

[0075] Deep feature mining can be understood as an in-depth analysis of fault time-series feature sequences, extracting two core types of features: First, single-event features (the probability of an abnormal occurrence of a single parameter event, such as the probability of an abnormal temperature event or an abnormal wind speed event occurring in the entire sequence). Second, multi-event correlation features (the probability of concurrent anomalies among multiple parameter events, such as the probability of a high temperature event and a low wind speed event occurring simultaneously, or the probability of a slow decrease in moisture content event and a fast circulation speed event occurring simultaneously). An event feature vector refers to the quantification and integration of the extracted single-event features and multi-event correlation features into a fixed-dimensional vector with a unified numerical scale. Each element of the event feature vector corresponds to a quantized value of a fault feature. The core function of the event feature vector is to transform the complex time-series and correlation features in the fault time-series feature sequence into a calculable and comparable quantified vector, comprehensively representing the overall operating status (normal or abnormal) and potential faults of the equipment. For example, after deep feature mining, four single-event features (temperature anomaly probability 0.2, moisture content anomaly probability 0.1, wind speed anomaly probability 0.3, and cycle anomaly probability 0.05) and two multi-event correlation features (high temperature and low wind speed concurrent probability 0.15, and slow moisture content decrease and fast cycle speed concurrent probability 0.03) are extracted. These features are quantified and integrated to form an event feature vector [0.2, 0.1, 0.3, 0.05, 0.15, 0.03]. The event feature vector comprehensively reflects the current abnormal operating characteristics of the equipment.

[0076] S420: The event feature vector is compared with the standard fault vector in the preset fault feature library to obtain the fault feature coefficient. The fault feature coefficient is used to characterize the similarity between the currently collected feature and the preset standard fault feature.

[0077] It can be understood that the fault feature library refers to a pre-established database storing standard feature vectors corresponding to various common faults of batch circulating grain dryers. The database includes common fault types and standard fault vectors, with a corresponding relationship between them. Optionally, the database can be established by consulting relevant literature. The fault feature library contains typical feature vectors for each fault type and serves as the benchmark for fault matching and diagnosis. A standard fault vector refers to a representative feature vector set in the pre-defined fault feature library for each common fault (such as poor hot air circulation, circulation system failure, or temperature control system failure). The dimension of the standard fault vector is consistent with the event feature vector, facilitating similarity calculation. Similarity comparison refers to using a pre-defined similarity calculation algorithm (such as cosine similarity algorithm or Euclidean distance algorithm) to calculate the degree of similarity between the event feature vector and each standard fault vector. The fault feature coefficient refers to the result of the similarity calculation, typically ranging from [0,1]. The closer the coefficient is to 1, the higher the similarity between the current event feature vector and the standard fault vector, and the greater the probability of a fault in the current equipment. The closer the coefficient is to 0, the lower the similarity, and the smaller the probability of a fault. For example, in a preset fault feature library, the standard fault vector for poor hot air circulation is [0.25, 0.12, 0.35, 0.06, 0.18, 0.04]. The cosine similarity algorithm is used to calculate the similarity between the event feature vector [0.2, 0.1, 0.3, 0.05, 0.15, 0.03] and the standard fault vector. The cosine similarity calculation formula is as follows: Where A is the current event feature vector; B is the standard fault vector in the preset fault database; A×B is the dot product of the two vectors; These are the magnitudes of the two vectors, respectively. This is the fault characteristic coefficient. Substituting the two vectors into the cosine similarity calculation formula yields a fault characteristic coefficient of 0.99, indicating that the current operating characteristics of the equipment are highly similar to the standard characteristics of a fault with poor hot air circulation.

[0078] In one possible implementation, S420, the event feature vector is compared with the standard fault vector in a preset fault feature library to obtain fault feature coefficients, including: S421, obtain the standard fault vectors corresponding to common faults of batch circulating grain dryers in the preset fault feature library.

[0079] It is understandable that the preset fault feature library is a feature database containing various common faults of batch circulating grain dryers, established in advance through a large amount of experimental and equipment operation data accumulation. It covers abnormal temperature, abnormal moisture content, abnormal wind speed, abnormal circulation, and various compound faults (faults caused by the simultaneous abnormality of multiple parameters). Common faults refer to the types of faults that frequently occur in actual equipment operation, such as poor hot air circulation, fan failure (low wind speed), temperature sensor failure (abnormal temperature fluctuation), circulating motor failure (abnormal circulation speed), abnormal moisture content detection, etc. Standard fault vectors refer to fixed-dimensional vectors formed by extracting typical single-event features and multi-event correlation features for each common fault and quantifying them. Each common fault corresponds to a unique standard fault vector, and the vector dimension is completely consistent with the event feature vector to ensure similarity comparison. For example, the preset fault feature library contains 3 types of common faults and their corresponding standard fault vectors: 1. Poor hot air circulation: [0.25, 0.12, 0.35, 0.06, 0.18, 0.04]. 2. Fan failure: [0.05, 0.08, 0.8, 0.03, 0.02, 0.01]. 3. Abnormal moisture content detection: [0.03, 0.7, 0.05, 0.04, 0.01, 0.02]. These three types of standard fault vectors are obtained from the fault feature library and used for subsequent comparison with event feature vectors.

[0080] S422, calculate the similarity between the event feature vector and the standard fault vector in the preset fault feature library, and use the similarity as the fault feature coefficient.

[0081] Similarity calculation, as we understand it, refers to using suitable algorithms for vector comparison (such as cosine similarity, Euclidean distance, Pearson correlation coefficient, etc.) to calculate the degree of similarity between the currently generated event feature vector and each standard fault vector in a preset fault feature library. Essentially, it determines the degree of overlap between two vectors. Using similarity as a fault feature coefficient means directly using the calculated similarity result as the fault feature coefficient. The magnitude of the coefficient directly reflects the degree of matching between the current equipment operating characteristics and the corresponding standard fault characteristics, providing a quantitative basis for subsequent fault type determination. For example, if the event feature vector is [0.2, 0.1, 0.3, 0.05, 0.15, 0.03], the cosine similarity algorithm is used to calculate the similarity with three types of standard fault vectors: the similarity with the standard vector of poor hot air circulation is 0.99, the similarity with the standard vector of fan failure is 0.8, and the similarity with the standard vector of abnormal moisture content detection is 0.33. The corresponding fault feature coefficients are 0.99, 0.8, and 0.33, respectively. Among them, the fault type corresponding to 0.99 (poor hot air circulation) is the most likely fault in the current equipment.

[0082] By adopting the above steps S421 to S422, it is helpful to solve the problems of lack of standard basis for fault feature comparison and inability to quantify the comparison results, making fault matching diagnosis more objective and accurate, avoiding misdiagnosis and missed diagnosis caused by subjective judgment, and at the same time, the most likely fault type of the current equipment can be clearly determined by the quantification coefficient.

[0083] S430 determines the fault diagnosis conclusion based on the fault characteristic coefficients and the fault diagnosis coefficients in the preset fault characteristic library.

[0084] It can be understood that the fault feature coefficient refers to the similarity coefficient between the current event feature vector and each standard fault vector. The fault diagnosis coefficient in the preset fault feature library refers to the threshold coefficient (usually set to 0.7, but can be adjusted according to actual needs) set in advance in the fault feature library to determine whether a fault exists. When the fault feature coefficient corresponding to a certain standard fault vector is greater than or equal to the fault diagnosis coefficient, the equipment is determined to have a fault. When the fault feature coefficients corresponding to all standard fault vectors are less than the fault diagnosis coefficient, the equipment is determined to be operating normally. Determining the fault diagnosis conclusion refers to combining the comparison results of the fault feature coefficient and the fault diagnosis coefficient to determine whether the equipment has a fault, the specific fault type, and then generating a fault optimization plan based on the corresponding fault solutions in the preset fault feature library, ultimately forming a complete fault diagnosis conclusion. For example, if the preset fault diagnosis coefficient is 0.7, and the fault feature coefficients obtained in S422 are 0.99 (poor hot air circulation), 0.8 (fan fault), and 0.33 (abnormal moisture content detection), where 0.99 > 0.8 > 0.7 > 0.33, then the equipment is determined to have a fault of poor hot air circulation. Then, the optimization solution corresponding to the hot air circulation problem is retrieved from the preset fault feature library (check the operation status of the bottom fan, clean the dust in the hot air duct, and calibrate the wind speed sensor). Finally, the fault diagnosis conclusion is formed: the specific fault type is hot air circulation problem, and the fault optimization solution is to check the operation status of the bottom parallel fans, clean the dust in the hot air duct, calibrate the wind speed sensor, ensure smooth hot air flow, and restore the normal operation of the equipment.

[0085] By adopting the above steps S410 to S430, the accuracy, efficiency and practicality of fault diagnosis of batch circulating grain dryers can be improved, downtime due to faults can be reduced and equipment maintenance costs can be lowered.

[0086] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0087] Corresponding to the fault diagnosis method for batch circulating grain dryers described in the above embodiments, this application also provides a fault diagnosis device for batch circulating grain dryers, the various units of which can realize the various steps of the fault diagnosis method for batch circulating grain dryers. Figure 4 A structural block diagram of a fault diagnosis device for a batch circulating grain dryer provided in an embodiment of this application is shown. For ease of explanation, only the parts related to the embodiment of this application are shown.

[0088] Reference Figure 4 The device includes: The acquisition unit is used to acquire multi-dimensional operational data of the batch circulating grain dryer when it completes the drying of a batch of grain. This multi-dimensional operational data includes equipment temperature parameters, moisture content parameters, wind speed parameters, and circulation parameters.

[0089] The normalization unit is used to preprocess temperature, moisture content, wind speed, and circulation parameters to obtain a normalized fault feature dataset. This normalized fault feature dataset reflects the characteristic data set of the batch circulating grain dryer's operating status and potential faults.

[0090] The sequence unit is used to generate corresponding parameter events based on the data type of the normalized fault feature dataset, and then arrange them in order according to the timestamps of each parameter event to construct a fault time-series feature sequence. The parameter events include temperature events generated based on temperature parameters, moisture content events generated based on moisture content parameters, wind speed events generated based on wind speed parameters, and cycle events generated based on cycle parameters.

[0091] The diagnostic unit performs deep feature mining and matching diagnosis on the fault time-series characteristics to determine the fault diagnosis conclusion for the batch circulating grain dryer. The fault diagnosis conclusion includes the specific fault type and a fault optimization plan.

[0092] It should be noted that the information interaction and execution process between the above-mentioned units are based on the same concept as the method embodiments of this application. For details on their specific functions and technical effects, please refer to the method embodiments section, which will not be repeated here.

[0093] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units is used as an example. In practical applications, the above functions can be assigned to different functional units as needed, that is, the internal structure of the device can be divided into different functional units to complete all or part of the functions described above. The functional units in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units are only for easy differentiation and are not intended to limit the scope of protection of this application. The specific working process of the units in the above device can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0094] This application also provides a batch circulating grain dryer. Figure 5 This is a schematic diagram of a batch circulating grain dryer provided in one embodiment of this application. Figure 5 As shown, the batch circulating grain dryer 6 of this embodiment includes: at least one processor 60 ( Figure 5 Only one is shown in the image), at least one memory 61 ( Figure 5 (Only one is shown in the image) and a computer program 62 stored in the at least one memory 61 and executable on the at least one processor 60, wherein when the processor 60 executes the computer program 62, it causes the batch circulating grain dryer 6 to perform the steps in any of the above-described batch circulating grain dryer fault diagnosis method embodiments, or causes the batch circulating grain dryer 6 to perform the functions of the units in the above-described device embodiments.

[0095] Exemplarily, the computer program 62 may be divided into one or more units, which are stored in the memory 61 and executed by the processor 60 to complete this application. The one or more units may be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of the computer program 62 in the batch circulating grain dryer 6.

[0096] The batch circulating grain dryer 6 can be a cloud server, cloud host, commercial desktop computer, laptop computer, e-commerce smart terminal, tablet computer, etc. The batch circulating grain dryer includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the method as described in any of the foregoing aspects. The batch circulating grain dryer 6 may include, but is not limited to, a processor 60 and a memory 61. Those skilled in the art will understand that... Figure 5This is merely an example of a batch circulating grain dryer 6 and does not constitute a limitation on the batch circulating grain dryer 6. It may include more or fewer components than shown, or combine certain components, or different components, such as input / output devices, network access devices, buses, etc.

[0097] The processor 60 can be a Central Processing Unit (CPU), or it can be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor.

[0098] In some embodiments, the memory 61 may be an internal storage unit of the batch circulating grain dryer 6, such as a hard drive or memory of the batch circulating grain dryer 6. In other embodiments, the memory 61 may be an external storage device of the batch circulating grain dryer 6, such as a plug-in hard drive, smart media card (SMC), secure digital (SD) card, flash card, etc., equipped on the batch circulating grain dryer 6. Furthermore, the memory 61 may include both internal storage units and external storage devices of the batch circulating grain dryer 6. The memory 61 is used to store operating systems, applications, bootloaders, data, and other programs, such as the program code of computer programs. The memory 61 can also be used to temporarily store data that has been output or will be output.

[0099] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps in any of the above method embodiments.

[0100] This application provides a computer program product that, when run on a batch circulating grain dryer, enables the batch circulating grain dryer to perform the steps described in any of the above method embodiments.

[0101] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of this application can be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include at least: any entity or system capable of carrying the computer program code to a batch circulating grain dryer, a recording medium, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium, such as a USB flash drive, a portable hard drive, a magnetic disk, or an optical disk.

[0102] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0103] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0104] In the embodiments provided in this application, it should be understood that the disclosed batch circulating grain dryer fault diagnosis method, batch circulating grain dryer fault diagnosis device, and batch circulating grain dryer can be implemented in other ways. For example, the batch circulating grain dryer fault diagnosis device and batch circulating grain dryer embodiments described above are merely illustrative. For instance, the division of units is merely a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual coupling or direct coupling or communication connection may be through some interfaces; the indirect coupling or communication connection between units may be electrical, mechanical, or other forms.

[0105] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0106] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A fault diagnosis method for a batch circulating grain dryer, characterized in that, Applied to batch circulating grain dryers, the method includes: Acquire multi-dimensional operational data of a batch circulating grain dryer when it completes the drying of a batch of grain; wherein, the multi-dimensional operational data includes equipment temperature parameters, moisture content parameters, wind speed parameters, and circulation parameters; The temperature parameters, moisture content parameters, wind speed parameters, and circulation parameters are preprocessed to obtain a regularized fault feature dataset; wherein, the regularized fault feature dataset reflects the characteristic data set of the batch circulating grain dryer's operating status and potential faults; Based on the normalized fault feature dataset, corresponding parameter events are generated according to data types, and then arranged in order according to the timestamps of each parameter event to construct a fault time-series feature sequence; wherein, the parameter events include temperature events generated according to the temperature parameters, moisture content events generated according to the moisture content parameters, wind speed events generated according to the wind speed parameters, and cycle events generated according to the cycle parameters. Deep feature mining and matching diagnosis are performed on the fault time sequence to determine the fault diagnosis conclusion of the batch circulating grain dryer; wherein, the fault diagnosis conclusion includes the specific fault type and fault optimization scheme.

2. The fault diagnosis method for a batch circulating grain dryer according to claim 1, characterized in that, The preprocessing of the temperature parameters, moisture content parameters, wind speed parameters, and circulation parameters yields a normalized fault feature dataset, including: Based on the temperature parameters, moisture content parameters, wind speed parameters, and circulation parameters, a fault diagnosis standard dataset is obtained; The fault diagnosis standard dataset is numerically normalized to obtain the fault diagnosis normalized dataset. An outlier detection algorithm is used to identify and remove noisy data from the normalized fault diagnosis dataset to determine the normalized fault feature dataset.

3. The fault diagnosis method for a batch circulating grain dryer according to claim 2, characterized in that, The step of using an outlier detection algorithm to identify and remove noisy data from the fault diagnosis normalized dataset to determine the normalized fault feature dataset includes: Calculate the mean and standard deviation of each of the multidimensional running data in the fault diagnosis normalized dataset, and determine the noise identification threshold range; Each of the multidimensional running data points in the fault diagnosis normalized dataset that exceeds the noise identification threshold range is marked as a suspected noise point; The suspected noise points were verified using a sliding window filtering algorithm, and the suspected noise points caused by sensor anomalies were removed to obtain the missing dataset. The mean of adjacent multidimensional running data within a sliding window is used to fill the missing positions in the missing dataset to obtain a normalized fault feature dataset.

4. The fault diagnosis method for a batch circulating grain dryer according to claim 3, characterized in that, The process employs a sliding window filtering algorithm to verify the suspected noise points, removing those generated by sensor anomalies to obtain a missing dataset, including: Set a preset length sliding window; If only a single multidimensional running data point within the sliding window is the suspected noise point, the suspected noise point is determined to be interference noise data generated by sensor interference and is removed to obtain a blank dataset.

5. The fault diagnosis method for a batch circulating grain dryer according to claim 1, characterized in that, Based on the normalized fault feature dataset, corresponding parameter events are generated according to data type, and then arranged in order according to the timestamps of each parameter event to construct a fault time-series feature sequence, including: Based on the regularized fault feature dataset, parameter events are generated according to data type classification; wherein, the parameter events include temperature events, moisture content events, wind speed events, and cycle events; Extract the timestamps of the temperature event, moisture content event, wind speed event, and cycle event to obtain a time feature data sequence; The feature values ​​of the temperature event, moisture content event, wind speed event, and cycle event are extracted and combined with the time feature data sequence to obtain the fault time sequence feature sequence.

6. The fault diagnosis method for a batch circulating grain dryer according to claim 5, characterized in that, The extraction of timestamps from the temperature event, moisture content event, wind speed event, and cycle event yields a time feature data sequence, including: The timestamps bound to the temperature event, moisture content event, wind speed event, and cycle event are extracted respectively to form an independent set of timestamps corresponding to the four types of events; Invalid timestamps are removed from each of the independent timestamp sets to obtain a valid timestamp set; Based on the set of valid timestamps, an ordered sequence of time feature data is formed.

7. The fault diagnosis method for a batch circulating grain dryer according to claim 5, characterized in that, The extraction of feature values ​​from the temperature event, moisture content event, wind speed event, and cycle event, combined with the time feature data sequence, yields a fault time sequence feature sequence, including: Extract the feature values ​​corresponding to the temperature event, moisture content event, wind speed event, and cycle event, and bind the feature values ​​with the timestamps of the corresponding parameter events to form binary association groups; Sort all the binary association groups according to the chronological order of the timestamps to obtain the standard association group; Based on all the aforementioned standard association groups, a fault timing feature sequence is constructed; wherein, the fault timing feature sequence can clearly reflect the time distribution of each of the aforementioned parameter events and the corresponding feature value changes.

8. The fault diagnosis method for a batch circulating grain dryer according to claim 1, characterized in that, The process of performing deep feature mining and matching diagnosis on the fault time-series feature sequence to determine the fault diagnosis conclusion of the batch circulating grain dryer includes: Deep feature mining is performed on the fault time sequence feature sequence, and the results are integrated to form an event feature vector; wherein, the event feature vector is a quantitative feature vector used to characterize the overall operating status of the equipment; The event feature vector is compared with the standard fault vector in the preset fault feature library to obtain the fault feature coefficient; wherein, the fault feature coefficient is used to characterize the similarity between the currently collected feature and the preset standard fault feature. Based on the fault characteristic coefficients and the fault diagnosis coefficients in the preset fault characteristic library, the fault diagnosis conclusion is determined.

9. The fault diagnosis method for a batch circulating grain dryer according to claim 8, characterized in that, The event feature vector is compared with the standard fault vector in the preset fault feature library to obtain fault feature coefficients, including: Obtain the standard fault vectors corresponding to common faults of batch circulating grain dryers from the preset fault feature library; Calculate the similarity between the event feature vector and the standard fault vector in the preset fault feature library, and use the similarity as the fault feature coefficient.

10. A batch circulating grain dryer, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the method as described in any one of claims 1 to 9.

Images