A dynamic threshold adaptive adjustment method for dry bulk cargo terminal equipment
By dynamically adjusting the alarm thresholds of belt conveyors at dry bulk cargo terminals and combining adjustment strategies for various scenarios, the problems of false alarms and missed alarms under the fixed threshold method have been solved, achieving higher accuracy in anomaly detection and improved operation and maintenance efficiency.
Patent Information
- Application Number
- CN202611499760.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-09-24
- Publication Date
- 2026-11-17
AI Technical Summary
The existing alarm system for belt conveyor equipment at dry bulk terminals uses a fixed threshold method, which leads to frequent false alarms and missed alarms. It cannot adapt to changes in equipment operating status and individual differences, affecting operation and maintenance efficiency and equipment safety.
By dynamically adjusting alarm thresholds based on equipment operation data, historical maintenance data, and collaborative equipment data, and taking into account alarm authenticity, equipment load status, and maintenance impact, threshold adjustment strategies are adopted for various scenarios, including alarm accuracy feedback, equipment load conditions, maintenance status, and equipment quality differences.
It improves the accuracy of anomaly detection, reduces false alarms and missed alarms, enhances the adaptability of alarm strategies, and improves operation and maintenance efficiency and equipment security.
Smart Images

Figure CN123113382A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of operation and maintenance and online monitoring technology for dry bulk cargo terminal equipment, and more specifically, to a dynamic threshold adaptive adjustment method applicable to dry bulk cargo terminal equipment. Background Technology
[0002] Dry bulk cargo terminals are crucial infrastructure for loading, unloading, transshipment, and storage of bulk commodities such as coal and ore. Belt conveyor systems play a critical role in the continuous transport of materials, and their operational status directly impacts terminal efficiency and safety. Currently, belt conveyor systems at dry bulk cargo terminals are typically equipped with operational status monitoring and fault alarm systems. These systems collect operating parameters such as temperature, vibration, current, and speed during equipment operation and compare these parameters with preset alarm thresholds to determine if any abnormalities exist.
[0003] Most existing alarm systems for belt conveyors use fixed threshold alarms. This means that alarm thresholds are preset based on factory parameters, design standards, or the experience of maintenance personnel, and these fixed thresholds are used to determine anomalies during equipment operation. However, the actual operating environment and conditions of belt conveyors at dry bulk terminals are complex, with significant fluctuations in equipment load, continuous operating time, and operating parameters. Under prolonged, high-intensity, or high-load operating conditions, equipment operating parameters may exhibit periodic fluctuations related to normal operating conditions. Using fixed alarm thresholds can easily misinterpret these normal fluctuations as equipment anomalies, resulting in numerous false alarms. Furthermore, for some anomalous states that are gradual or vary individually, fixed alarm thresholds may fail to identify them in a timely manner, leading to missed alarms.
[0004] Furthermore, belt conveyor equipment at dry bulk terminals requires frequent maintenance, upkeep, and repair. After these operations, the equipment's operational stability and parameter characteristics may change. Existing fixed thresholds are typically not adjusted accordingly, leading to a mismatch between the original alarm thresholds and the actual operating status after maintenance, thus affecting the accuracy of alarm judgments. Simultaneously, belt conveyors of different models, manufacturers, years of operation, and historical operating conditions exhibit significant differences in operational stability and failure probabilities. Using a uniform alarm threshold fails to reflect the individual characteristics of different equipment, further reducing the adaptability of the alarm thresholds.
[0005] The aforementioned fixed threshold alarm method is prone to false alarms and missed alarms. On the one hand, this increases the workload of maintenance personnel in screening and handling alarm information, reducing maintenance efficiency. On the other hand, missed alarms may cause equipment anomalies to go undetected in a timely manner, allowing local faults to escalate, causing unplanned equipment downtime, affecting continuous port operations and resulting in significant economic losses. Furthermore, the persistent high frequency of false alarms may reduce maintenance personnel's sensitivity and trust in alarm information, thereby increasing the risk to safe equipment operation.
[0006] As dry bulk terminals evolve towards automation, digitalization, and intelligence, existing alarm methods based on fixed thresholds are no longer sufficient to meet the demands for accurate equipment monitoring and intelligent operation and maintenance under complex operating conditions. While existing technologies involve optimizing alarm thresholds based on specific operating conditions or adjusting alarm parameters in conjunction with equipment health assessments and remaining life predictions, they typically optimize thresholds for a single factor or scenario. This makes it difficult to simultaneously consider multiple influencing factors such as changes in equipment load, changes in operational status, individual equipment differences, and historical alarm feedback. Furthermore, they lack continuous evaluation of the effectiveness of threshold adjustments and adaptive correction mechanisms for strategies.
[0007] Therefore, how to dynamically and adaptively adjust the alarm threshold for different operating and maintenance scenarios of belt conveyor equipment in dry bulk cargo terminals, and continuously optimize the threshold adjustment strategy based on alarm feedback, in order to improve the accuracy of abnormal alarms and the threshold adaptability, has become an urgent technical problem to be solved. Summary of the Invention
[0008] To overcome a series of shortcomings in the existing technology, the purpose of this application is to provide a dynamic threshold adaptive adjustment method suitable for dry bulk cargo terminal equipment, which includes the following steps: Based on the equipment operation data, historical maintenance data, and collaborative equipment operation data associated with the belt conveyor at the dry bulk terminal, the threshold adjustment scenario corresponding to the current belt conveyor is determined. The corresponding threshold adjustment strategy is invoked based on the threshold adjustment scenario, and the alarm threshold adjustment parameters under the current threshold adjustment scenario are determined by combining the alarm authenticity, equipment load status, degree of operation and maintenance impact, or equipment health status. The alarm threshold of the belt conveyor is dynamically adjusted according to the alarm threshold adjustment parameters, and anomaly detection is performed based on the adjusted alarm threshold to obtain alarm result data.
[0009] In some embodiments, threshold adjustment scenarios include feedback adjustment scenarios based on alarm accuracy and threshold adaptability, dynamic adjustment scenarios based on equipment load conditions, compensation adjustment scenarios based on operation and maintenance status, and differentiated setting scenarios based on equipment quality differences, wherein: Based on the feedback adjustment scenario of alarm accuracy and threshold adaptability, the alarm accuracy change trend is analyzed according to historical alarm data and real-time alarm results. Combined with the adaptability relationship between the current alarm threshold and the equipment operating status, the optimization direction and adjustment range of the alarm threshold are determined. Based on the dynamic adjustment scenario of equipment load conditions, the impact of the current operating conditions on the alarm threshold is determined according to the changes in equipment load status, running time and key operating parameters, and the alarm threshold is dynamically corrected. Based on the compensation and adjustment scenario of operation and maintenance status, the impact of operation and maintenance operations on equipment operating characteristic parameters is assessed according to equipment maintenance and repair records and changes in equipment status after operation and maintenance, and the alarm threshold is compensated and corrected. Based on the differences in equipment quality, differentiated scenarios are set up. According to the equipment model, manufacturer, years of operation and historical operating status parameters, the individual characteristics of the equipment are determined, and corresponding alarm thresholds are configured based on the individual characteristics of the equipment.
[0010] In some embodiments, for feedback adjustment scenarios based on alarm accuracy and threshold adaptability, the method for dynamically adjusting the current alarm threshold of the belt conveyor is as follows: Based on the component types of the belt conveyor, historical fault statistics, and preset threshold adjustment rules, determine the first adjustment step size and the second adjustment step size. Obtain the alarm accuracy rate corresponding to the current statistical period, and compare the alarm accuracy rate with the preset alarm accuracy rate lower limit and the preset alarm accuracy rate upper limit respectively; When the alarm accuracy rate is less than the preset alarm accuracy rate lower limit, the current alarm threshold is increased according to the first adjustment step size. When the alarm accuracy rate is greater than or equal to the preset lower limit of alarm accuracy rate and less than or equal to the preset upper limit of alarm accuracy rate, the current alarm threshold remains unchanged. When the alarm accuracy rate is greater than the preset alarm accuracy rate upper limit, the current alarm threshold is reduced based on the second adjustment step size.
[0011] In some embodiments, for compensation adjustment scenarios based on operation and maintenance status, the method for dynamically adjusting the current alarm threshold of the belt conveyor is as follows: Obtain the current maintenance type and the degree of improvement in equipment operational stability of the belt conveyor; The corresponding threshold compensation coefficient is matched based on the type of operation and maintenance and the improvement of equipment operation stability, and the alarm threshold compensation amount is determined according to the threshold compensation coefficient. The current alarm threshold is adjusted based on the alarm threshold compensation amount to obtain the target alarm threshold; Obtain device alarm data after applying the target alarm threshold, and determine the alarm accuracy rate based on the device alarm data; Based on the comparison between the alarm accuracy rate and the preset accuracy threshold, the target alarm threshold is corrected to obtain the updated alarm threshold.
[0012] In some embodiments, the method for dynamically adjusting the current alarm threshold of the belt conveyor for a differentiated setting scenario based on equipment quality differences is as follows: Determine the corresponding classification adjustment parameters based on the equipment classification results; The alarm threshold adjustment step size is determined based on the classification adjustment parameters, and the current alarm threshold is adjusted based on the alarm threshold adjustment step size to obtain an initial adjusted alarm threshold that is compatible with the equipment category. Acquire equipment operation data, alarm result data, and fault data within a preset operating cycle after initial alarm threshold adjustment, and determine threshold adjustment evaluation parameters based on the equipment operation data, alarm result data, and fault data; The classification adjustment parameters are updated based on the threshold adjustment evaluation parameters to obtain the updated alarm threshold adjustment parameters. When the threshold adjustment evaluation parameters meet the preset compensation conditions, the current alarm threshold is compensated and adjusted based on the updated alarm threshold adjustment parameters.
[0013] In some embodiments, the method for obtaining alarm result data is as follows: The current equipment operation data of the belt conveyor is obtained, and the equipment operation data is divided into multiple windows based on a preset sliding time window. Compare the device operation data sample values in each window's operation data with the adjusted alarm threshold, and calculate the proportion of device operation data sample values in each window's operation data that exceed the alarm threshold. The proportion of device operation data sampling values exceeding the alarm threshold is compared with a preset early warning proportion threshold and a preset fault proportion threshold, wherein the preset early warning proportion threshold is less than the preset fault proportion threshold; Determine the alarm level of the corresponding window's running data based on the comparison results; Based on the alarm level corresponding to the operating data of each window, the alarm judgment data corresponding to the operating data of each window is determined, and the alarm result data of the belt conveyor is generated based on the alarm judgment data corresponding to the operating data of each window.
[0014] In some embodiments, it also includes: The effectiveness of alarm threshold adjustment is evaluated based on alarm result data to determine threshold optimization information; The threshold adjustment strategy is adaptively corrected based on the threshold optimization information, and the alarm threshold is continuously updated according to the corrected threshold adjustment strategy.
[0015] In some embodiments, the method for obtaining threshold optimization information is as follows: Obtain alarm result data for a preset evaluation period, and filter the alarm result data according to the scene identifier adjusted according to the corresponding threshold to obtain the target alarm result data; Based on the verification results of the target alarm results by the operation and maintenance personnel, the alarm precision, recall, F1 score and false alarm rate under the corresponding threshold adjustment scenario are obtained; The alarm precision, recall, F1 score, and false alarm rate are matched with the preset evaluation conditions of the corresponding threshold adjustment scenario to determine the evaluation results of the alarm threshold adjustment strategy. When the false alarm rate exceeds the preset upper limit or the recall rate is lower than the preset lower limit, the current alarm threshold adjustment strategy is determined to not meet the evaluation conditions, and threshold optimization information is generated.
[0016] In some embodiments, the method for adaptively correcting the threshold adjustment strategy based on threshold optimization information is as follows: Obtain the threshold optimization information corresponding to the current threshold adjustment scenario, and parse the threshold optimization information to determine the adjustment direction deviation, adjustment magnitude deviation, and scenario judgment deviation; Based on the adjustment direction deviation, adjustment magnitude deviation, and scenario judgment deviation, the adjustment step size parameter, threshold floating parameter, and threshold baseline parameter in the threshold adjustment strategy corresponding to the current threshold adjustment scenario are adaptively corrected to obtain the corrected threshold adjustment strategy parameters. The modified threshold adjustment strategy parameters are constrained and verified to determine whether each parameter meets the corresponding preset parameter constraint conditions. When any parameter exceeds the preset correction range or the single adjustment range limit, the corresponding parameter is subject to boundary constraints to obtain the threshold adjustment strategy parameters that meet the constraint conditions. Based on the threshold adjustment strategy parameters that meet the constraints, the threshold adjustment strategy corresponding to the current threshold adjustment scenario is updated to form an adaptive threshold adjustment strategy suitable for the current threshold adjustment scenario.
[0017] In some embodiments, the method for continuously updating the alarm threshold according to the modified threshold adjustment strategy is as follows: Set the alarm threshold update cycle, which includes a first update cycle, a second update cycle, and a third update cycle; When the first update cycle is reached, the current alarm threshold is updated based on the adjustment result of the sensitivity coefficient k value. When the second update cycle is reached, the threshold adjustment strategy is corrected based on the threshold optimization information, and the alarm threshold adjustment process in the subsequent first update cycle is executed based on the corrected threshold adjustment strategy. When the third update cycle is reached, the equipment quality difference classification results are updated based on the output results of the incrementally trained equipment health assessment model and equipment remaining life prediction model, and the corresponding threshold benchmark parameters are adjusted according to the updated equipment quality difference classification results. The current alarm threshold is continuously updated based on the updated threshold adjustment strategy, equipment quality difference classification results, and threshold benchmark parameters.
[0018] Compared with the prior art, this application has the following beneficial effects: This application identifies different operating and maintenance scenarios of belt conveyor equipment and adaptively adjusts the alarm threshold to achieve dynamic matching between the alarm threshold and equipment operating conditions, individual differences, and health status, thereby reducing false alarms and missed alarms, improving the accuracy of equipment anomaly detection, and enhancing the continuous adaptability of the alarm strategy. Attached Figure Description
[0019] Figure 1 This is a flowchart illustrating the overall process of a dynamic threshold adaptive adjustment method for dry bulk cargo terminal equipment according to the present invention. Figure 2 This is a flowchart illustrating the processing of the feedback adjustment scenario based on alarm accuracy and threshold adaptability in this invention. Figure 3 This is a flowchart illustrating the processing of a dynamic adjustment scenario based on equipment load conditions in this invention. Figure 4 This is a flowchart illustrating the processing of compensation and adjustment scenarios based on operation and maintenance status in this invention. Figure 5 This is a flowchart illustrating the differentiated scenario setting and adaptive optimization closed-loop processing based on equipment quality differences in this invention. Detailed Implementation
[0020] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of the embodiments of this invention will be described in more detail below with reference to the accompanying drawings. In the drawings, the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The described embodiments are some embodiments of this invention, but not all embodiments.
[0021] Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0022] The embodiments and directional terms described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, and should not be construed as limiting the present invention.
[0023] In a broad embodiment of the present invention, a dynamic threshold adaptive adjustment method applicable to dry bulk cargo terminal equipment includes the following steps: Based on the equipment operation data, historical maintenance data, and collaborative equipment operation data associated with the belt conveyor at the dry bulk terminal, the threshold adjustment scenario corresponding to the current belt conveyor is determined. The corresponding threshold adjustment strategy is invoked based on the threshold adjustment scenario, and the alarm threshold adjustment parameters for the current scenario are determined by combining the alarm authenticity, device load status, degree of operational and maintenance impact, or device health status. The alarm threshold of the belt conveyor is dynamically adjusted according to the alarm threshold adjustment parameters, and anomaly detection is performed based on the adjusted alarm threshold to obtain alarm result data. The effectiveness of alarm threshold adjustment is evaluated based on alarm result data to determine threshold optimization information; The threshold adjustment strategy is adaptively corrected based on the threshold optimization information, and the alarm threshold is continuously updated according to the corrected threshold adjustment strategy.
[0024] In some embodiments, threshold adjustment scenarios include feedback adjustment scenarios based on alarm accuracy and threshold adaptability, dynamic adjustment scenarios based on equipment load conditions, compensation adjustment scenarios based on operation and maintenance status, and differentiated setting scenarios based on equipment quality differences. Based on the feedback adjustment scenario of alarm accuracy and threshold adaptability, it is used to adjust the leniency of the current alarm threshold according to the historical and real-time changes in alarm accuracy of the belt conveyor. Based on the dynamic adjustment scenario of equipment load conditions, it is used to temporarily adjust the current alarm threshold according to the continuous high load and high intensity operation of the belt conveyor. The compensation adjustment scenario based on operation and maintenance status is used to compensate and adjust the alarm threshold by combining the operation and maintenance work order information corresponding to equipment maintenance and repair and the changes in equipment status after operation and maintenance. The differentiated setting scenario based on equipment quality differences is used to differentiate the alarm thresholds corresponding to different equipment quality levels according to equipment runtime, equipment manufacturer, and historical operating status parameters.
[0025] In some embodiments, for feedback adjustment scenarios based on alarm accuracy and threshold adaptability, the method for determining the alarm threshold adjustment parameter is as follows: If the current threshold adjustment scenario is determined to be a feedback adjustment scenario based on alarm accuracy and threshold adaptability, obtain the equipment operation parameters of the belt conveyor that triggered the alarm within the time window corresponding to the alarm trigger time, the collaborative operation parameters of the associated collaborative equipment, and obtain historical normal operation data and historical alarm data. An alarm authenticity judgment dataset is constructed based on equipment operating parameters, collaborative operating parameters, historical normal operation data, and historical alarm data. The alarm authenticity judgment dataset is then classified and calculated using a logistic regression algorithm to obtain the true alarm probability corresponding to each alarm event. The authenticity judgment result of the corresponding alarm event is determined by matching the actual alarm probability with the preset alarm authenticity judgment conditions. Based on the authenticity judgment results of each alarm event within the preset statistical period, the number of real alarms and the total number of alarm events are obtained, and the alarm accuracy rate corresponding to the current statistical period is determined according to the number of real alarms and the total number of alarm events. The alarm accuracy rate is matched with the preset alarm accuracy rate evaluation conditions, and the corresponding alarm threshold adjustment parameters are determined based on the matching results.
[0026] In some embodiments, for feedback adjustment scenarios based on alarm accuracy and threshold adaptability, the method for dynamically adjusting the current alarm threshold of the belt conveyor is as follows: Based on the component type of the belt conveyor, historical fault statistics, and preset threshold adjustment rules, determine the corresponding first adjustment step size and second adjustment step size; Obtain the alarm accuracy rate corresponding to the current statistical period, and compare the alarm accuracy rate with the preset alarm accuracy rate lower limit and the preset alarm accuracy rate upper limit respectively; When the alarm accuracy rate is less than the preset alarm accuracy rate lower limit, the current alarm threshold is increased according to the first adjustment step size. When the alarm accuracy rate is greater than or equal to the preset lower limit of alarm accuracy rate and less than or equal to the preset upper limit of alarm accuracy rate, the current alarm threshold remains unchanged; When the alarm accuracy rate is greater than the preset alarm accuracy rate upper limit, the current alarm threshold is reduced according to the second adjustment step.
[0027] In some embodiments, the method for determining the dynamic adjustment scenario based on equipment load conditions is as follows: Acquire load condition data during the operation of the belt conveyor. The load condition data includes the continuous running time of the equipment, the actual operating load rate of the equipment, and the density parameters of the conveyed material. Based on the preset load status determination rules and combined with time series analysis methods, feature extraction is performed on the load condition data to obtain the load status characteristics corresponding to the current belt conveyor. The load status characteristics are matched with the preset high load judgment threshold to determine the load status judgment result corresponding to the current belt conveyor. Based on the load status determination results, determine whether the load status corresponding to the continuous running time of the equipment, the actual operating load rate of the equipment, and the density parameters of the conveyed material meets the preset high load determination conditions. When at least one of the preset high-load judgment conditions is met, the current belt conveyor is determined to be in a high-load operation scenario.
[0028] In some embodiments, for scenarios involving dynamic adjustment based on equipment load conditions, the method for determining the alarm threshold adjustment parameter is as follows: When the current threshold adjustment scenario is determined to be a dynamic adjustment scenario based on the equipment load condition, the current equipment operating status parameters are obtained, and the current alarm threshold is adjusted based on the preset first temporary adjustment step size to obtain the initial alarm threshold adjustment parameters. Based on the initial alarm threshold adjustment parameters, the operating status of the belt conveyor is monitored to obtain the corresponding operating parameter change information. The equipment operating status is determined based on the change information of operating parameters. When the equipment operating status is continuously in the preset abnormal fluctuation range and meets the continuous operating conditions, a preset second temporary adjustment step is added on the first temporary adjustment step to update the alarm threshold adjustment parameters. Store the updated alarm threshold adjustment parameters and corresponding device operating status information, and generate threshold adjustment records for high-load scenarios; Continuously acquire the operating load data of the belt conveyor and determine whether the current equipment status meets the preset threshold recovery conditions based on the operating load data; When the equipment operating load is lower than the preset recovery load ratio and the duration reaches the preset recovery time, the alarm threshold adjustment parameter will be restored to the baseline alarm threshold before adjustment. When the high-load scenario determination conditions are met again during the device recovery process, the high-load scenario identification is re-executed, and the duration corresponding to the threshold recovery condition is re-accumulated.
[0029] In some embodiments, for compensation adjustment scenarios based on operation and maintenance status, the method for determining alarm threshold adjustment parameters is as follows: When the current threshold adjustment scenario is determined to be a dynamic adjustment scenario based on equipment load conditions, the corresponding maintenance work order information of the belt conveyor is obtained, and the corresponding maintenance type is determined based on the maintenance work order information. Based on the maintenance type, obtain the equipment operating parameters, alarm records and baseline alarm threshold parameters of the belt conveyor within a preset time period before maintenance, as well as the equipment operating status parameters within a preset time period after maintenance. Compare and analyze the equipment operating parameters, alarm records, and baseline alarm threshold parameters before and after maintenance to determine the changes in equipment status before and after maintenance. Stability assessment is conducted based on changes in equipment status to determine the extent to which equipment operational stability can be improved. Determine the corresponding threshold compensation coefficient based on the improvement in equipment operational stability; The baseline alarm threshold parameters are compensated based on the operation and maintenance type and the threshold compensation coefficient to determine the corresponding alarm threshold adjustment parameters.
[0030] In some embodiments, the method for determining the change in device state is as follows: Determine the data collection period before maintenance based on the corresponding maintenance type of the belt conveyor equipment; The operating parameters, alarm records, and threshold benchmark parameters of the belt conveyor equipment during the data collection period before maintenance are obtained to form the status data before maintenance. Obtain the operating status parameters of the belt conveyor within a preset time period after maintenance, and generate post-maintenance status data. Perform correlation analysis on the pre-operation and maintenance status data and the post-operation and maintenance status data to extract the corresponding operational status feature parameters; The differences in equipment status before and after maintenance are calculated based on the characteristic parameters of the operating status to determine the amount of change in equipment status.
[0031] In some embodiments, for compensation adjustment scenarios based on operation and maintenance status, the method for dynamically adjusting the current alarm threshold of the belt conveyor is as follows: Obtain the corresponding maintenance type and the degree of improvement in equipment operational stability for belt conveyor equipment; The corresponding threshold compensation coefficient is matched based on the type of operation and maintenance and the improvement of equipment operation stability, and the alarm threshold compensation amount is determined according to the threshold compensation coefficient. The current alarm threshold is adjusted based on the alarm threshold compensation amount to obtain the target alarm threshold; Obtain device alarm data after applying the target alarm threshold, and calculate the corresponding alarm accuracy based on the device alarm data; The alarm accuracy rate is compared with the preset accuracy threshold, and the target alarm threshold is corrected based on the comparison result to obtain the updated alarm threshold.
[0032] In some embodiments, for differentiated setting scenarios based on differences in device quality, the method for determining alarm threshold adjustment parameters is as follows: When the current threshold adjustment scenario is determined to be a differentiated setting scenario based on equipment quality differences, the equipment attribute information and operating status information of the belt conveyor are obtained. The equipment attribute information includes equipment manufacturer information and running time, and the operating status information includes historical operating data and historical alarm data. Based on equipment attribute information and operating status information, feature analysis is performed on belt conveyor equipment, and the remaining operating time of the equipment is determined by combining the preset remaining life prediction model. The belt conveyor equipment is classified according to its remaining operating time, equipment attribute information, and operating status information to obtain the equipment batch classification results; Based on the equipment batch classification results, obtain the historical alarm data of the corresponding equipment batch, and determine the corresponding threshold benchmark parameters and threshold adjustment range based on the historical alarm data; Based on the threshold baseline parameters and threshold adjustment range corresponding to the equipment batch classification results, determine the alarm threshold adjustment parameters corresponding to the current belt conveyor.
[0033] In some embodiments, the method for classifying belt conveyor equipment is as follows: Obtain the equipment operating parameters and equipment attribute parameters of the belt conveyor. The equipment operating parameters include the equipment running time, and the equipment attribute parameters include the manufacturer information. Based on the equipment's operating parameters and attribute parameters, historical operating data of the same brand and type of equipment is retrieved and input into a preset equipment remaining life prediction model to obtain the prediction result of the equipment's remaining sustainable operating time. The equipment category corresponding to the belt conveyor is determined by matching the predicted remaining time of sustainable operation of the equipment with the preset classification threshold. Equipment classification results are generated based on equipment category. The equipment classification results include high-quality equipment, ordinary equipment, monitoring equipment, and early warning equipment.
[0034] In some embodiments, the method for dynamically adjusting the current alarm threshold of the belt conveyor for a differentiated setting scenario based on equipment quality differences is as follows: Determine the corresponding classification adjustment parameters based on the equipment classification results; The alarm threshold adjustment step size is determined based on the classification adjustment parameters, and the current alarm threshold is adjusted based on the alarm threshold adjustment step size to obtain the initial adjusted alarm threshold after classification adaptation. Acquire equipment operation data, alarm data, and fault data within the operating cycle corresponding to the initial adjustment alarm threshold, and determine the threshold adjustment evaluation parameters based on the equipment operation data, alarm data, and fault data; The classification adjustment parameters are updated based on the threshold adjustment evaluation parameters to obtain the updated alarm threshold adjustment parameters. When the threshold adjustment evaluation parameters meet the preset compensation conditions, the current alarm threshold is compensated and adjusted based on the updated alarm threshold adjustment parameters.
[0035] It should be noted that the sensitivity coefficient k is not used to replace the threshold adjustment strategies corresponding to the four types of threshold adjustment scenarios mentioned above, but rather serves as a global fine-tuning parameter shared by all four types of threshold adjustment scenarios. For any threshold adjustment scenario, the scenario-based alarm threshold is first obtained according to the threshold adjustment strategy corresponding to that scenario. Then, the sensitivity coefficient k is used to uniformly fine-tune the scenario-based alarm threshold to obtain the final effective alarm threshold. Therefore, the four types of threshold adjustment scenarios are used to determine the scenario-based basic adjustment results of the alarm threshold, and the sensitivity coefficient k is used to perform secondary fine-tuning of the scenario-based basic adjustment results based on cross-cycle alarm performance feedback.
[0036] In some embodiments, it also includes: Obtain feedback information from maintenance personnel regarding alarm result data input, and determine the corresponding false alarm rate and missed alarm events based on the feedback information; According to the preset optimization cycle, the alarm result data of each belt conveyor are statistically analyzed to obtain the corresponding alarm performance indicators, including accuracy, recall, F1 value and false alarm rate. Based on the matching relationship between alarm performance indicators and preset evaluation thresholds, the adjustment direction and adjustment amount of the current sensitivity coefficient k value are determined, and a sensitivity coefficient k value adjustment instruction is generated according to the constrained adaptive update strategy based on alarm performance feedback. The current sensitivity coefficient k value is updated according to the sensitivity coefficient k value adjustment command, and the updated sensitivity coefficient k value is obtained.
[0037] In some embodiments, the method for updating the current sensitivity coefficient k value is as follows: When the false alarm rate is higher than the preset false alarm rate target value for a continuous preset optimization period, the sensitivity coefficient k is increased by the preset first adjustment amount. When the false alarm rate is less than the preset lower limit of the false alarm rate for a continuous preset evaluation period, the sensitivity coefficient k is reduced according to the preset second adjustment amount. When a missed event is detected, the sensitivity coefficient k is reduced by a preset third adjustment amount, where the third adjustment amount is greater than the second adjustment amount. When no missed events are detected within the preset time range and the false alarm rate is less than the preset stability evaluation threshold, the sensitivity coefficient k value is restored to the initial setting value.
[0038] In some embodiments, the constrained adaptive update strategy based on alarm performance feedback uses alarm precision, alarm recall, F1 score, and false alarm rate as alarm performance evaluation indicators, and uses a preset upper limit for false alarm rate, a preset lower limit for alarm recall, and a preset safe adjustment range for the sensitivity coefficient k as constraints. Within each preset optimization cycle, the current alarm performance indicator is compared with the corresponding preset evaluation threshold. Based on the comparison result, the corresponding update method is determined from the following update methods: increasing the sensitivity coefficient k, decreasing the sensitivity coefficient k, and restoring the sensitivity coefficient k to its initial set value. The corresponding adjustment amount is determined based on a preset first adjustment amount, a preset second adjustment amount, or a preset third adjustment amount. A safety constraint verification is performed on the determined sensitivity coefficient k adjustment amount. When the sensitivity coefficient k or the magnitude of a single adjustment exceeds the corresponding preset safe adjustment range, boundary constraints are applied. Through the above-mentioned constrained adaptive update based on alarm performance feedback, false alarms are suppressed while the risk of missed alarms is limited, achieving a dynamic balance between alarm precision and alarm recall.
[0039] In some embodiments, it also includes: Obtain the false alarm rate and missed alarm events generated by the belt conveyor, and associate them with the corresponding equipment operation data, alarm result data and load condition data to build a special sample library; New samples are detected in a special sample library. When the number of new samples reaches a preset incremental threshold, an incremental training task is triggered. According to the preset incremental training cycle, new samples from the special sample library are called to incrementally train the pre-established equipment health assessment model and equipment remaining life prediction model in order to update the corresponding model parameters. The equipment health assessment model and equipment remaining life prediction model after parameter updates are validated, and the target state assessment model is determined based on the validation results. Based on the target state assessment model, the health status of the belt conveyor is assessed and the remaining operating time is predicted. The equipment status classification information is determined based on the health status assessment results and the remaining operating time prediction results.
[0040] In some embodiments, the method for incrementally training the equipment health assessment model is as follows: Retrieve new false alarm case data and false alarm case data from the special sample library, and input the new false alarm case data and false alarm case data into the currently effective equipment health assessment model; Based on the currently effective equipment health assessment model, incremental training is performed on newly added false alarm case data and false alarm case data, and the model parameters are updated according to the preset first learning rate and preset first iteration number; Obtain a preset verification sample and input the preset verification sample into the device health assessment model after incremental training to obtain the recognition accuracy for false alarm cases and false alarm cases; The recognition accuracy is compared with a preset accuracy threshold, and the comparison result determines whether to update the currently effective device health assessment model using the incrementally trained model parameters. When the recognition accuracy reaches the preset accuracy threshold, the model parameters after incremental training are used to update the currently effective device health assessment model; when the recognition accuracy does not reach the preset accuracy threshold, the currently effective device health assessment model remains unchanged, and the corresponding newly added false alarm case data and missed alarm case data are retained to the next incremental training cycle.
[0041] In some embodiments, the method for incrementally training the equipment remaining life prediction model is as follows: Acquire new equipment degradation sequence data from a special sample library and input the new equipment degradation sequence data into the currently effective equipment remaining life prediction model to construct incremental training samples; The equipment remaining life prediction model is fine-tuned based on incremental training samples, and the model weights of the equipment remaining life prediction model are updated according to a preset second learning rate and a preset second iteration number. The second learning rate is less than the first learning rate in the incremental training process of the equipment health assessment model, and the second iteration number is greater than the first iteration number in the incremental training process of the equipment health assessment model. Based on the equipment degradation sequence data corresponding to the false alarm case data and the missed alarm case data, the remaining life prediction model of the equipment after incremental training is predicted and validated, and the remaining life prediction deviations corresponding to the false alarm cases and the missed alarm cases are determined respectively. The remaining lifetime prediction deviation corresponding to the missed cases is compared with the preset first deviation threshold, and the remaining lifetime prediction deviation corresponding to the false cases is compared with the preset second deviation threshold. When the remaining lifetime prediction deviation corresponding to a missed case is less than the first deviation threshold and the remaining lifetime prediction deviation corresponding to a false case is less than the second deviation threshold, the model weight of the device remaining lifetime prediction model after incremental training is used to update the currently effective device remaining lifetime prediction model. When any remaining lifetime prediction deviation does not meet the corresponding deviation threshold condition, the currently effective remaining lifetime prediction model of the equipment remains unchanged, and the corresponding newly added equipment degradation sequence data is retained to the next incremental training cycle.
[0042] In some embodiments, the method for obtaining alarm result data is as follows: The system acquires the current operating data of the belt conveyor and divides the operating data into multiple windows based on a preset sliding time window. Compare the device operation data sample values in each window's operation data with the adjusted alarm threshold, and calculate the proportion of device operation data sample values in each window's operation data that exceed the alarm threshold; The proportion of device operation data sampling values exceeding the alarm threshold is compared with a preset early warning proportion threshold and a preset fault proportion threshold, wherein the preset early warning proportion threshold is less than the preset fault proportion threshold; The alarm level of the corresponding window's operating data is determined based on the comparison results. Specifically, when the proportion of the device operating data sampled values exceeding the alarm threshold reaches the preset warning proportion threshold, warning-level alarm result data is generated; when the proportion of the device operating data sampled values exceeding the alarm threshold reaches the preset fault proportion threshold, fault-level alarm result data is generated. Based on the alarm level corresponding to the operating data of each window, the alarm judgment data corresponding to the operating data of each window is determined, and the alarm result data of the belt conveyor is generated based on the alarm judgment data corresponding to the operating data of each window.
[0043] In some embodiments, alarm result data includes alarm occurrence time, characteristic parameter type triggering the alarm, alarm level, corresponding threshold adjustment scenario identifier, equipment operation data snapshot at the time of alarm triggering, and alarm threshold adjustment parameters corresponding to the time of alarm triggering. The alarm result data is processed hierarchically according to the alarm level. For warning-level alarm result data, it is sent to the operation and maintenance management terminal and a corresponding reminder message is generated, while also being stored in the alarm result dataset. For fault-level alarm result data, it is sent to the operation and maintenance management terminal and a corresponding equipment shutdown suggestion is generated, while simultaneously pushing the equipment shutdown suggestion. The alarm result data is classified and stored according to the belt conveyor equipment number, threshold adjustment scenario type, and alarm level to construct a structured alarm result dataset. Based on the structured alarm result dataset, the alarm threshold adjustment effect is evaluated and threshold optimization information is determined. Simultaneously, the alarm triggering process is traced and analyzed according to the equipment dimension and the threshold adjustment scenario dimension.
[0044] In some embodiments, the method for obtaining threshold optimization information is as follows: Obtain alarm result data for a preset evaluation period, and filter the alarm result data according to the scene identifier adjusted according to the corresponding threshold to obtain the target alarm result data; Based on the verification results of the target alarm results by the operation and maintenance personnel, the alarm precision, recall, F1 score and false alarm rate under the corresponding threshold adjustment scenario are obtained; The alarm precision, recall, F1 score, and false alarm rate are matched with the preset evaluation conditions of the corresponding threshold adjustment scenario to determine the evaluation results of the alarm threshold adjustment strategy. When the false alarm rate exceeds the preset upper limit or the recall rate is lower than the preset lower limit, the current alarm threshold adjustment strategy is determined to not meet the evaluation conditions, and threshold optimization information is generated.
[0045] In some embodiments, the threshold optimization information includes adjustment direction deviation information, adjustment magnitude deviation information, and scenario judgment deviation information, and establishes corresponding relationships with feedback adjustment scenarios based on alarm accuracy and threshold adaptability, dynamic adjustment scenarios based on equipment load conditions, compensation adjustment scenarios based on operation and maintenance status, and differentiated setting scenarios based on equipment quality differences, respectively. The adjustment direction deviation information is used to characterize the deviation between the alarm threshold adjustment direction and the target adjustment direction determined based on the alarm performance evaluation results in the corresponding scenario. The adjustment magnitude deviation information is used to characterize the deviation between the adjustment step size, threshold floating coefficient or threshold reference parameter and the target adjustment parameter in the corresponding scenario. The scenario judgment deviation information is used to characterize the matching deviation between the corresponding scenario judgment conditions and the actual operating state of the belt conveyor. Threshold optimization information is generated based on the deviation information corresponding to each threshold adjustment scenario, and is used for subsequent optimization and updating of alarm threshold adjustment strategies.
[0046] In some embodiments, the method for adaptively correcting the threshold adjustment strategy based on threshold optimization information is as follows: Obtain the threshold optimization information corresponding to the current threshold adjustment scenario, and parse out the adjustment direction deviation, adjustment magnitude deviation, and scenario judgment deviation; Based on the aforementioned adjustment direction deviation and adjustment magnitude deviation, the adjustment step size parameter, floating coefficient parameter, and threshold reference parameter in the threshold adjustment strategy under the corresponding threshold adjustment scenario are corrected. Based on the scene determination deviation, the determination parameters of the corresponding threshold adjustment scene are corrected, wherein the determination parameters include duration threshold parameters, proportion threshold parameters and density threshold parameters. When the threshold adjustment scenario is a feedback adjustment scenario based on alarm accuracy and threshold adaptability, the sliding window time parameter in the alarm accuracy statistics process is corrected according to the threshold optimization information. The modified threshold adjustment strategy parameters are constrained and verified. When the modification result of any parameter exceeds the corresponding preset modification range, the modification result is restricted to the corresponding preset modification range, and the single modification magnitude is restricted from exceeding the preset single modification ratio. The threshold adjustment strategy for the corresponding threshold adjustment scenario is updated based on the correction results that satisfy the constraints.
[0047] In some embodiments, the step of continuously updating the alarm threshold according to the modified threshold adjustment strategy includes: An alarm threshold update cycle is set, which includes a first update cycle, a second update cycle, and a third update cycle. The first update cycle corresponds to the sensitivity coefficient k value adjustment process, the second update cycle corresponds to the threshold adjustment strategy correction process, and the third update cycle corresponds to the equipment health assessment model and the equipment remaining life prediction model update process. The first update cycle is shorter than the second update cycle, and the second update cycle is shorter than the third update cycle. When the first update cycle is reached, the current alarm threshold is updated based on the adjustment result of the sensitivity coefficient k value. When the second update cycle is reached, the adjustment step size parameter, floating coefficient parameter and threshold base parameter in the threshold adjustment strategy are corrected according to the threshold optimization information, and the alarm threshold adjustment process in the subsequent first update cycle is updated based on the corrected threshold adjustment strategy. When the third update cycle is reached, the equipment quality difference classification results and the corresponding threshold benchmark parameters are updated based on the output results of the incrementally trained equipment health assessment model and equipment remaining life prediction model. The current alarm threshold is continuously updated based on the updated threshold adjustment strategy, equipment quality difference classification results, and threshold benchmark parameters.
[0048] The preferred embodiments of the present invention are described below with reference to the accompanying drawings, and the present invention will be further described in detail.
[0049] This embodiment takes dry bulk cargo terminal belt conveyor equipment (belt conveyor, stacker-reclaimer matching belt conveyor, rollers and other components) as the application scenario, and systematically describes the technical solution of the present invention around the links of data acquisition, scene recognition, parameter calculation, threshold adjustment, result output, feedback evaluation and adaptive correction.
[0050] like Figure 1 As shown, the dynamic threshold adaptive adjustment method for dry bulk cargo terminal equipment provided by the present invention includes the following steps in its overall implementation process: Step S1: Data Acquisition. Continuously collect equipment operation data (vibration, temperature, current, speed, etc.), historical operation and maintenance data (maintenance and repair work orders and processing records) of the belt conveyor, as well as the operation data of the associated equipment (such as the simultaneous window operation parameters of the belt conveyor's peripheral idlers and the stacker-reclaimer body), as the basic data source for scene recognition and parameter calculation; Step S2: Scene Recognition. Based on the multi-source data obtained in Step S1, determine the current threshold adjustment scene of the belt conveyor. The scenes include feedback adjustment scenes based on alarm accuracy and threshold adaptability, dynamic adjustment scenes based on equipment load conditions, compensation adjustment scenes based on operation and maintenance status, and differentiated setting scenes based on equipment quality differences. These four types of scenes can be triggered individually or evaluated in a coordinated manner when multiple judgment conditions are met. Step S3: Parameter Calculation. The threshold adjustment strategy corresponding to the scenario identified in Step S2 is invoked. Combined with information such as alarm authenticity, device load status, degree of operational impact, or device health status, the alarm threshold adjustment parameters for that scenario (including adjustment direction, adjustment step size, compensation coefficient, or threshold benchmark, etc.) are calculated. Step S4: Threshold Adjustment and Anomaly Detection. Based on the alarm threshold adjustment parameters obtained in Step S3, the current alarm threshold of the belt conveyor is dynamically adjusted, and anomaly detection is performed on the real-time collected equipment operation data based on the adjusted alarm threshold, outputting alarm result data; Step S5: Performance Evaluation. The alarm results data output in Step S4 are statistically analyzed and verified. Evaluation indicators such as alarm precision, recall, F1 score, and false alarm rate are calculated, and threshold optimization information is determined accordingly. Step S6: Adaptive Correction and Continuous Update. Based on the threshold optimization information obtained in Step S5, the adjustment step size parameter, threshold floating parameter, and threshold baseline parameter in the threshold adjustment strategy are adaptively corrected, and the alarm threshold is continuously updated according to a preset multi-level update cycle, forming a closed-loop adjustment mechanism of "scene recognition - data acquisition - parameter calculation - threshold adjustment - effect evaluation - adaptive correction".
[0051] In step S2, based on the equipment operation data, historical maintenance data, and collaborative equipment operation data of the belt conveyor, judgment conditions are set for the four types of threshold adjustment scenarios: For feedback adjustment scenarios based on alarm accuracy and threshold adaptability, the alarm accuracy is analyzed based on the historical alarm data and real-time alarm review results of the device within a preset statistical period (7 days in this embodiment). The scenario is triggered when the alarm accuracy deviates from the current threshold adaptability range. For dynamic adjustment scenarios based on equipment load conditions, the scenario is triggered when any parameter reaches the high load judgment condition, based on parameters such as continuous operating time of the equipment, actual operating load rate and conveyed material density. For compensation adjustment scenarios based on operation and maintenance status, the scenario is triggered when the equipment has just completed maintenance, repair or other operation and maintenance operations, based on the operation and maintenance work order information issued by the production management system. For differentiated setting scenarios based on equipment quality differences, the scenario is triggered when the equipment batch classification results change or when the equipment is first connected to the system, based on the equipment manufacturer, runtime, and historical operating status parameters.
[0052] When the same device meets the trigger conditions of multiple scenarios at the same time, the linkage evaluation is carried out in the priority order of "operation and maintenance status compensation - load condition adjustment - quality differentiation setting - alarm accuracy feedback adjustment" to avoid conflicts between threshold adjustments in multiple scenarios.
[0053] like Figure 2 As shown, the processing flow for this scenario includes four stages: data acquisition, calculation of actual alarm probability, statistics of alarm accuracy, and adjustment of threshold levels, as detailed below.
[0054] 1) Data Acquisition and Calculation of Actual Alarm Probability. When an alarm is triggered on the belt conveyor, the equipment operating parameters and the collaborative operating parameters of associated equipment within the corresponding time window are collected. An alarm authenticity judgment dataset is constructed by combining historical normal operation data and historical alarm data. A logistic regression algorithm is used to calculate the actual alarm probability corresponding to this alarm. The calculation method is as follows: (1) In the formula, The true alarm probability output by the logistic regression model, with a value range of 0 to 1, is used to characterize the possibility that the alarm belongs to a real equipment abnormality event. The logistic regression linear combination value is calculated using the following formula: (2) In the formula, These are the bias parameters for the logistic regression model; For the first The regression weights corresponding to each feature ( ); The first time within the time window corresponding to the alarm trigger time The features (including equipment vibration amplitude, temperature, current fluctuation, and deviation of operating parameters of cooperating equipment, etc.) are derived from the equipment operation data and cooperating equipment operation data collected in step S1. The bias parameters and weight parameters mentioned above are obtained by training the previously accumulated training sample set through maximum likelihood estimation. The model can only be put into use when the accuracy of the model judgment is not less than 90% after training. The alarm is compared with a preset alarm authenticity threshold (0.5 in this embodiment). If the alarm value is greater than the threshold, it is determined to be a real alarm; otherwise, it is determined to be a false alarm. The determination result, along with the corresponding device operation data, is stored in the alarm result dataset as the data source for the next step of alarm accuracy statistics.
[0055] 2) Alarm Accuracy Statistics. A sliding time window is used to statistically analyze alarm events within a preset statistical period, and the alarm accuracy rate corresponding to the current statistical period is calculated. The calculation method is as follows: (3) In the formula, For the first Alarm accuracy rate corresponding to each statistical period (a 7-day sliding window is used in this embodiment); This represents the number of events that have been verified and confirmed as genuine alarms by operations and maintenance personnel within the statistical period. This represents the total number of alarm events within the statistical period; both are obtained from the alarm result dataset.
[0056] 3) Threshold adjustment by level. Based on the component types and historical fault statistics of the belt conveyor, pre-determine the preset lower limit and upper limit of alarm accuracy. Compare the alarm accuracy with the two proportional thresholds, and adjust the current alarm threshold by level according to the following formula: ,when (4) ,when (5) ,when (6) In the formula, The alarm threshold before adjustment. The adjusted alarm thresholds are both critical values of operating parameters that trigger alarms (such as vibration velocity, temperature, etc.). The lower limit of alarm accuracy is preset (60% in this embodiment). To preset the upper limit of alarm accuracy (in this embodiment, it is set to 85%), and The values of both are derived from historical fault statistics corresponding to the component type of the belt conveyor equipment, reflecting the reasonable range of alarm accuracy of the component under normal working conditions. This is the first adjustment step size (ranging from 5% to 10%). The second adjustment step size (ranging from 5% to 8%) is preset based on component type and historical fault statistics. When alarm accuracy... When the alarm accuracy rate is below the preset lower limit, it indicates that the current alarm threshold setting is too low, too sensitive, and has too many false alarms. The threshold is adjusted upward by the first adjustment step to relax the alarm conditions. When the alarm accuracy rate is between the preset lower limit and the preset upper limit, it indicates that the threshold is compatible with the actual operating state of the equipment, and the threshold is kept unchanged. When the alarm accuracy rate is above the preset upper limit, it indicates that the current alarm threshold setting is too high, the sensitivity is insufficient, and there is a risk of missed alarms. The threshold is adjusted downward by the second adjustment step to tighten the alarm conditions and move towards the alarm value limit, thereby achieving dynamic adaptation between the threshold and the alarm accuracy rate.
[0057] Specific data example: Taking a belt conveyor vibration alarm as an example, the baseline threshold is 1.2 mm / s. A training sample set of 5000 data points was built, and the accuracy of the logistic regression model trained was 92%. Using a 7-day sliding window, when the alarm accuracy rate is 55% for a consecutive week (lower than the preset lower limit of 60%), the threshold is increased by 8% in the first adjustment step, resulting in an adjusted threshold of 1.2 × (1 + 8%) = 1.296 mm / s. When the alarm accuracy rate rises back to the 60%–85% range, the threshold remains unchanged at 1.296 mm / s. If the alarm accuracy rate then rises to 88% for a consecutive week (higher than the preset upper limit of 85%), the threshold is decreased by 6% in the second adjustment step, resulting in 1.296 × (1 - 6%) ≈ 1.218 mm / s. This achieves dynamic tracking and adjustment of the threshold as the alarm accuracy rate changes.
[0058] like Figure 3 As shown, the processing flow for this scenario includes five stages: load condition data acquisition and scenario determination, temporary threshold adjustment, dynamic fine-tuning, and threshold recovery, as detailed below.
[0059] 1) Load Condition Data Acquisition and Scenario Determination. Real-time data acquisition of the belt conveyor's continuous operating time, actual operating load rate, and conveyed material density parameters is performed. A determination is made according to preset load condition judgment rules: If any one of the following three conditions is met—continuous operating time reaching a preset reference duration (10 hours in this example), actual operating load reaching a preset percentage of rated load (80% in this example), or conveyed material density reaching a preset reference density (1.8 t / m³ in this example)—the equipment is determined to be in a high-load operating scenario. When two or more conditions are met simultaneously, a comprehensive load condition characteristic index is calculated using time series analysis to quantify the load severity, serving as the basis for determining the subsequent dynamic fine-tuning range. (7) In the formula, The load condition comprehensive characteristic index is used to represent the comprehensive load level of the equipment under its current operating condition relative to its rated operating condition. The larger the value, the more severe the load. This refers to the continuous operating time of the equipment (in hours). The preset reference runtime is 10 hours. This represents the actual operating load of the equipment. The rated load of the equipment; To convey the actual density of the material, For reference material density (1.8t / m³) 3 ); To correspond to the weighting coefficients, which are calibrated based on historical fault statistics of belt conveyor equipment, all three are set to 1 / 3 in this embodiment. When the threshold is not less than the preset linkage evaluation threshold (1.2 in this embodiment), it is determined to be a heavy high-load condition with multiple parameters superimposed, and a larger secondary adjustment step size is adopted in the subsequent dynamic fine-tuning stage; otherwise, a conventional secondary adjustment step size is adopted, thereby realizing differentiated linkage between scene judgment and adjustment range.
[0060] 2) Temporary threshold adjustment and dynamic fine-tuning. After successful scene recognition, the temporary threshold adjustment mechanism is activated: (8) If, after adjustment, the equipment operating parameters continue to fluctuate abnormally and the equipment can still continue to operate, a second adjustment is made based on the above formula: (9) In the formula, This serves as the baseline alarm threshold before being triggered in high-load scenarios; This is the first temporary adjustment step size; This is the second temporary adjustment step size; The first temporary alarm threshold is obtained by adjusting the baseline alarm threshold once based on the first temporary adjustment step size; The second temporary alarm threshold is obtained by adjusting the first temporary alarm threshold a second time based on the second temporary adjustment step size.
[0061] 3) Automatic threshold recovery. The system continuously collects equipment operating load data. When the recovery conditions shown in the following formula are met, the alarm threshold is automatically restored to the baseline value: (10) In the formula, The preset recovery load ratio is (in this embodiment, it is 45% of the rated load). The cumulative duration for which the operating load remains below the recovery load ratio. The preset recovery time is 10 minutes in this embodiment. When the high load judgment condition is met again during the device recovery process, the high load scene recognition process is re-executed, and the data is re-accumulated. To avoid misjudging recovery conditions.
[0062] Specific data example: Taking the temperature alarm of the belt conveyor supporting a certain stacker-reclaimer as an example, the reference threshold is 80°C. When it is detected that the equipment operates continuously for 10 hours (reaching the preset reference duration of 10 hours), the operating load reaches 85% of the rated load (exceeding the preset ratio of 80%), and the actual density of the conveyed material is 3.6t / m 3 (exceeding the preset reference density of 1.8t / m 3 ), when the three conditions of operation duration, load rate and material density are met at the same time, a linkage evaluation is triggered, and the comprehensive characteristic index of the load state is calculated according to formula (7) ≈1.28, which is not less than the linkage evaluation threshold of 1.2, and it is determined as a severe high-load working condition with multi-parameter superposition; first, the threshold is lowered by the first temporary adjustment step of 6%, obtaining 80×(1-6%)=75.2°C; after adjustment, it is monitored that the equipment temperature keeps fluctuating between 74 and 75°C without obvious fault. In view of the determined severe high-load working condition, the threshold is further lowered by the second temporary adjustment step of 8%, obtaining 75.2×(1-8%)≈69.2°C; when the equipment operation ends and the operating load drops to 45% of the rated load and lasts for 10 minutes, the threshold is automatically restored to the reference value of 80°C.
[0063] As shown Figure 4 in the figure, the processing procedure of this scenario includes four links: operation and maintenance data collection, operation and maintenance effect evaluation, threshold compensation adjustment, and subsequent review and correction, which are specifically as follows.
[0064] 1) Operation and maintenance data collection. By docking with the data interface of the terminal production management system, the operating parameters, alarm records and threshold reference within the preset time period before the operation and maintenance of the belt conveyor equipment (3 to 6 months before operation and maintenance in this embodiment) and the operating state parameters within the preset time period after the operation and maintenance (15 days after operation and maintenance in this embodiment) are acquired, forming two sets of state data before and after operation and maintenance.
[0065] 2) Operation and maintenance effect evaluation. Taking the dispersion degree of the equipment operating parameters before and after operation and maintenance as the measurement index of operation stability, linear regression analysis is used to calculate the improvement range of equipment operation stability before and after operation and maintenance: (11) In the formula, is the improvement range of equipment operation stability, which is used to represent the improvement degree of operation and maintenance operations on the volatility of equipment operating parameters; is the standard deviation of the equipment operating parameters (such as vibration and temperature) before operation and maintenance, is the standard deviation of the equipment operating parameters after operation and maintenance, both of which are obtained by performing linear regression fitting on the state data before and after operation and maintenance and calculating the residual standard deviation.
[0066] The corresponding threshold compensation coefficient is matched according to the operation and maintenance type and the improvement range of operation stability The specific rules are as follows: When the maintenance type is major repair and At that time, threshold compensation coefficient Take 10% to 15%; When the maintenance type is routine maintenance, the threshold compensation coefficient Take 3% to 5%; When the maintenance type is temporary repair or routine repair, the threshold compensation coefficient Take 1% to 3%; Threshold compensation coefficient The specific value within the corresponding interval is the same as They are positively correlated.
[0067] 3) Threshold compensation adjustment. Based on the threshold compensation coefficient. Compensate for the baseline alarm threshold: (12) 4) Subsequent review and correction. After the threshold compensation is adjusted, the alarm accuracy rate is calculated every 15 days. If the accuracy rate is higher than the preset review threshold (85% in this embodiment), it indicates that the compensation range is too large. The compensation threshold is then adjusted downward according to the preset correction step size (3% in this embodiment). If the accuracy rate is consistently lower than the review threshold, the current compensation threshold is maintained. Periodic reviews ensure that the threshold continuously matches the actual operating status of the equipment after maintenance or repair.
[0068] Specific data example: Taking the vibration alarm of a belt conveyor roller as an example, the roller has completed bearing replacement, and the maintenance type is determined to be major maintenance. Through linear regression analysis, the vibration standard deviations before and after maintenance are 0.35 mm / s and 0.228 mm / s, respectively. According to formula (11), the improvement in equipment operation stability is calculated to be ΔS≈35%. According to the above compensation coefficient matching rules, the compensation coefficient value range corresponding to major maintenance of this equipment is 10%~15%, and the threshold compensation coefficient μ=12% is determined by combining the stability improvement degree of ΔS=35%. The baseline threshold of roller vibration alarm is 1.0 mm / s, so the target alarm threshold after compensation is 1.0×(1+12%)=1.12 mm / s.
[0069] Figure 5 The processing flow shown covers five stages: equipment remaining life prediction, batch classification, threshold benchmark matching, sensitivity coefficient update, and incremental model training. The first three stages are the core content of this scenario, as detailed below; the latter two stages (sensitivity coefficient k-value feedback adaptive update and incremental training of the health assessment model and remaining life prediction model) serve as an adaptive optimization closed loop shared by the four scenarios.
[0070] 1) Equipment Batch Classification. Based on equipment manufacturer information, runtime, and a historical operation database of similar equipment of the same brand, the system inputs a pre-defined equipment remaining life prediction model (1D-CNN-LSTM) to obtain the predicted remaining operating time of the equipment. The equipment batches are then classified according to a pre-defined classification threshold. (13) In the formula, The remaining time of sustainable operation of the equipment is predicted (expressed as a percentage of remaining lifespan), which is used to represent the relative remaining proportion of the equipment in its current health state from the end of its entire life cycle; The input model is a sequence of equipment degradation features (including historical operating parameters, alarm records, and maintenance records). This is the mapping function for the 1D-CNN-LSTM remaining lifetime prediction model, where convolutional layers are used to extract local degradation features, and long short-term memory layers are used to capture the temporal dependencies of degradation trends. According to... Based on the comparison results with the preset classification threshold, the equipment is classified as high-quality equipment. >85%), general equipment (60% < ≤85%), monitoring equipment (30% < ≤60%) and early warning equipment ( Category 4 (≤30%).
[0071] 2) Classification threshold benchmark matching. Based on the equipment batch classification results, determine the corresponding classification adjustment parameters and adjust the current alarm threshold accordingly. (14) In the formula, Initial adjustment of alarm thresholds to adapt to equipment batches; This is a batch adjustment factor used to represent the degree of relaxation or tightening of the relative benchmark threshold for equipment of different quality grades: High-quality equipment Use +8% to +12%, 0% for ordinary equipment, -3% to -6% for monitoring equipment, and -6% to -12% for early warning equipment. The specific values are determined based on the historical alarm data of the same batch of equipment.
[0072] 3) Dynamic Optimization. Every six months, the alarm accuracy and failure rate of each batch of equipment are statistically analyzed. Equipment operation data, alarm result data, and failure data within a preset operating cycle after the initial alarm threshold adjustment are obtained. Based on this, the threshold adjustment evaluation parameters are determined. If the failure rate of a certain batch of equipment increases significantly (more than the preset proportion relative to the previous statistical cycle), the step size is further reduced by 3% on the basis of the original adjustment range. The classification adjustment parameters are updated to obtain the updated alarm threshold adjustment parameters. When the preset compensation conditions are met, the current alarm threshold is compensated and adjusted. At the same time, the historical alarm data of this batch of equipment will be continuously used for the feedback adaptive update of the sensitivity coefficient k value and the incremental training of the remaining life prediction model, forming a data linkage between Scenario 4 and the adaptive optimization closed loop.
[0073] Specific data example: In a batch of belt conveyor equipment, the RUL of equipment with a running time of 1 year was predicted using a 1D-CNN-LSTM model. pred =90%, classified as high-quality equipment, and the factory vibration alarm threshold will be increased from 1.0mm / s to 1.1mm / s in 10% increments; equipment with a service life of 3 years has a RUL (Rating of Infrared Detection). pred =70%, classified as ordinary equipment, with the threshold remaining unchanged at 1.0 mm / s; equipment with a running time of 5 years RUL pred =40%, classified as regulated equipment, with a reduction in speed to 0.95mm / s in 5% increments; equipment with an operating time of 8 years has a RUL pred =25%, classified as early warning equipment, and the threshold was lowered to 0.9 mm / s in 10% increments; after six months, statistics showed that the failure rate of the early warning equipment batch increased, so the threshold was lowered by an additional 3% increment, and the threshold was adjusted to 0.9×(1-3%)≈0.873 mm / s to improve the monitoring sensitivity of this batch of equipment.
[0074] After adjusting the threshold, anomaly detection is performed on the currently collected equipment operation data of the belt conveyor based on the adjusted alarm threshold, generating alarm result data. Specifically, the equipment operation data is divided into multiple windows based on a preset sliding time window. The sampled values in each window are compared with the adjusted alarm threshold, and the proportion of sampled values exceeding the alarm threshold in each window is calculated. (15) In the formula, For the first The proportion of sampled values exceeding the alarm threshold in the data generated by each window; This represents the number of sampling points within the window that exceed the alarm threshold. This represents the total number of sampling points within this window. Compared with the preset warning ratio threshold (In this embodiment, 30% is used) and a preset fault ratio threshold. (In this embodiment, 60% is used) ) for comparison: when Generate early warning level alarm result data in real time; when Fault-level alarm result data is generated in real time; when two conditions are met simultaneously in the same window, the fault-level alarm result data shall prevail. The alarm result data includes the alarm occurrence time, the characteristic parameter type that triggered the alarm, the alarm level, the corresponding threshold adjustment scenario identifier, the snapshot of the equipment operation data when the alarm was triggered, and the corresponding alarm threshold adjustment parameters when the alarm was triggered; the warning-level alarm result data is sent to the operation and maintenance management terminal and a reminder message is generated, and the fault-level alarm result data is sent to the operation and maintenance management terminal and a device shutdown suggestion is generated. At the same time, the data is stored in categories according to the device number, scenario type, and alarm level to build a structured alarm result dataset for subsequent effect evaluation and traceability analysis.
[0075] Acquire alarm result data within a preset evaluation period, filter by scene identifier adjustment according to threshold, and combine the verification results of the target alarm result data by the operation and maintenance personnel to obtain the number of four types of events: true alarms (TP), false alarms (FP), true no alarms (TN), and missed alarms (FN). Calculate the four evaluation indicators—alarm precision, recall, F1 score, and false alarm rate—using the following formula: (16) (17) (18) (19) In the formula, Alarm accuracy rate is used to represent the proportion of true alarms in the alarm results. The alarm recall rate, in physical terms, is the proportion of all real abnormal events that are successfully identified. It is the harmonic mean of precision and recall, used to comprehensively evaluate the overall discrimination effect of alarm threshold; The false alarm rate, in physical terms, is the proportion of all events without abnormalities that are mistakenly identified as alarms. The above four indicators are matched with preset evaluation conditions for the corresponding threshold adjustment scenarios. When the false alarm rate exceeds the preset upper limit (20% in this embodiment) or the recall rate is lower than the preset lower limit (90% in this embodiment), the current threshold adjustment strategy is determined to not meet the evaluation conditions, and threshold optimization information is generated. This threshold optimization information includes three categories: adjustment direction deviation, adjustment magnitude deviation, and scenario judgment deviation, which respectively characterize the deviation between the alarm threshold adjustment direction, adjustment step size (or compensation coefficient, benchmark parameter), and scenario judgment conditions and the actual operating state of the equipment.
[0076] The threshold optimization information corresponding to the current threshold adjustment scenario is obtained and analyzed to obtain three types of deviations: adjustment direction deviation, adjustment magnitude deviation, and scenario judgment deviation. Based on this, the adjustment step size parameter, threshold floating parameter, and threshold baseline parameter in the threshold adjustment strategy for this scenario are adaptively corrected. (20) In the formula, The strategy parameters before correction (adjustment step size parameter, threshold floating parameter, or threshold baseline parameter). These are the revised strategy parameters; For the corresponding deviation (adjustment direction deviation, adjustment magnitude deviation, or scene judgment deviation). A correction coefficient (0.3–0.5 in this embodiment) is used to control the magnitude of a single correction and prevent drastic fluctuations in the strategy parameters. The corrected strategy parameters are then constrained to determine if they meet the corresponding preset parameter constraints. When a parameter exceeds the preset correction range or the single adjustment magnitude limit, boundary constraints are applied to the parameter to obtain threshold adjustment strategy parameters that meet the constraints. Based on this, the threshold adjustment strategy corresponding to the current threshold adjustment scenario is updated, forming an adaptive threshold adjustment strategy suitable for that scenario.
[0077] like Figure 5 As shown in the sensitivity coefficient update step, during the threshold adjustment process of each threshold adjustment scenario, a sensitivity coefficient k is introduced to uniformly correct the alarm threshold. This coefficient is not specific to any particular scenario, but rather a global fine-tuning layer superimposed on the calculation results of the four scenario-based adjustment strategies. (twenty one) In the formula, The alarm threshold is calculated using a scenario-based adjustment strategy. This is the final alarm threshold after the sensitivity coefficient is added; This is the sensitivity coefficient, used to represent the overall fine-tuning ratio of alarm sensitivity based on the scenario-based threshold. The initial value is set to 0.
[0078] Alarm performance metrics for each device are calculated weekly according to a preset optimization cycle, including alarm accuracy, alarm recall, F1 score, and false alarm rate. The sensitivity coefficient k is adjusted according to a constrained adaptive update strategy based on alarm performance feedback. The original method of judging based on three conditions—high false alarm rate, low false alarm rate, and detected missed events—could conflict when the false alarm rate simultaneously falls into the low and hysteresis recovery ranges. Therefore, within each preset optimization cycle, the priority order is changed to "missed alarm priority—high false alarm rate—hysteresis recovery—low false alarm rate," and only the highest priority condition is executed. The specific update rules are as follows: (twenty two) (twenty three) (twenty four) (25) In the formula, Let be the sensitivity coefficient for the t-th optimization cycle. This is the updated sensitivity coefficient; , , , These are the adjustment amounts corresponding to high false alarm rate (0.1 in this embodiment), low false alarm rate (0.05 in this embodiment), detected missed event (0.2 in this embodiment), and hysteresis recovery (0.02 in this embodiment), respectively. All four satisfy... This is used to reflect that the handling of missed alarms has a higher priority than that of false alarms, and to avoid the value of k oscillating repeatedly around the critical value of the false alarm rate; the recovery condition shown in equation (24) is only applicable when (That is, it was triggered when the policy was previously in a relaxed state) The time is not restored, so that the tightening caused by the underreporting can only be released gradually by the continuous decline of the false alarm rate according to formula (25), and will not be pulled back to the initial value all at once by the short-term low false alarm rate. The preset false alarm rate target value is (20% in this embodiment). The threshold for determining a low false alarm rate is set at 5% in this embodiment. The hysteresis recovery threshold (3% in this embodiment) is set, and all three conditions are met. ,exist and A hysteresis dead zone is formed between them, which corresponds to the triggering conditions for recovery judgment in equations (23), (25) and (24), respectively, to avoid frequent increases and decreases in the value of k when the false alarm rate fluctuates near the critical value.
[0079] Safe adjustment range of sensitivity coefficient k value , Labeling is done according to equipment category: take historical alarm data of this type of equipment that has been reviewed and labeled by maintenance personnel for at least 6 months, and re-determine the historical alarm events according to formula (21); where k>0 (relaxed), the existing review labels can be directly reused to calculate the recall rate and false alarm rate, k<0 (tightened) will generate new alarm events that were not triggered before, lack review labels, and the real alarm probability is used. They are included in the statistics only after being verified as pseudo-labels or in conjunction with fault repair work orders. The upper bound candidate value is determined by a constraint that the alarm recall rate is not less than 90%, and the lower bound candidate value is determined by a constraint that the false alarm rate does not exceed 20%. (26) (27) In the formula, Recall(k), (k) are the alarm recall rate and false alarm rate functions calculated after the threshold is re-evaluated according to equation (21) with a given k value. , Each component is contracted inward by a preset margin (0.02 in this embodiment) and rounded down by a preset step size (0.05 in this embodiment) to obtain... , ; , The slope of the curve showing the false alarm rate as a function of k at the current operating point is used to determine the range within which the change in the false alarm rate caused by a single adjustment falls (1-2 percentage points). Derived from the aforementioned multiple relationship.
[0080] Boundary constraints are applied to both the updated sensitivity coefficient k value and the final alarm threshold after superposition. (28) (29) In the formula, , The boundary of the safety adjustment range obtained from the aforementioned calibration, The current adjustment amount is determined according to formulas (22)-(25); , To prevent the final alarm threshold from exceeding the safe range due to the superposition of extreme values of the adjustment range of various threshold adjustment scenarios, and in combination with the lower and upper limits of the alarm threshold set within the allowable operating range of the equipment, the system integrates multiple adjustments.
[0081] It should be noted that the statistical period (7 days) for feedback adjustment scenarios based on alarm accuracy and threshold adaptability is longer than the update period (1 week) for the sensitivity coefficient k value. This is used to correct mid-term systematic deviations at the component type level, and the adjustment results serve as the benchmark for k value adjustment. When the scenario has triggered an adjustment within a certain period, the adjustment amount of the k-value in that period is reduced by a preset attenuation coefficient (0.5 in this embodiment) to avoid overlapping overshoots in the same direction. Furthermore, after the threshold adjustment strategy is corrected in the second update period, The range of values changes accordingly, therefore, after each second update cycle, the aforementioned method is used to adjust the values. , , , , Recalibrate; update when the deviation between the calibration result and the current parameters exceeds the preset verification threshold (20% in this embodiment), otherwise keep it unchanged.
[0082] The sensitivity coefficient k value update process aims to improve alarm accuracy while maintaining alarm recall. It uses a preset upper limit for false alarm rate, a preset lower limit for alarm recall, and the aforementioned calibrated safe adjustment range of the k value as constraints. When the sensitivity coefficient k value or its single adjustment exceeds the safe adjustment range, boundary constraints are applied according to equation (28). This method achieves a dynamic balance between alarm accuracy and alarm recall, while also ensuring that the false alarm rate does not exceed 20% and the missed alarm rate does not exceed 10%.
[0083] like Figure 5 As shown in the sample library update and incremental training process, false positive (FP) and false negative (FN) case data, along with their original data, preprocessed features, and operating condition information, are stored in a special sample library. Every 100 new cases trigger a sample library update. Incremental training is performed on the equipment health assessment model (XGBoost) and the equipment remaining life prediction model (1D-CNN-LSTM) every quarter. The output results of the trained models are fed back to the equipment batch classification process to update the equipment quality difference classification results.
[0084] Incremental training of the health assessment model: Based on the currently effective model, newly added FP and FN case data from a special sample library are loaded. The learning rate is set to 0.05, and the number of iterations is set to 50. After training, the model is validated using preset validation samples to obtain the recognition accuracy for FP and FN cases. (30) In the formula, To verify the recognition accuracy on the samples; To verify the number of correctly identified FP and FN cases in the sample, To verify the total number of samples. When If the current health assessment model is updated, the parameters of the incrementally trained model are used; otherwise, the original model is kept unchanged, and the corresponding case data is retained to the next incremental training cycle.
[0085] Incremental training of the remaining life expectancy prediction model: Newly added degraded sequence data from a special sample library was loaded using a fine-tuning method. The learning rate was set to 0.0005 (less than the learning rate of the health assessment model), and the number of iterations was set to 100 (more than the number of iterations of the health assessment model to ensure sufficient convergence of the fine-tuning). After training, prediction verification was performed on the degraded sequence data corresponding to missed and false positive cases, and the remaining life expectancy prediction bias was calculated. (31) In the formula, The remaining lifetime prediction bias is taken as the absolute value to ensure that the bias is always non-negative. The smaller the value, the higher the model prediction accuracy. To verify the actual remaining lifespan corresponding to the sample (obtained by inverse calculation from the actual scrapping or overhaul time of the equipment), when the prediction deviation corresponding to the missed case is less than the preset first deviation threshold (3% in this embodiment) and the prediction deviation corresponding to the false positive case is less than the preset second deviation threshold (5% in this embodiment), the currently effective remaining lifespan prediction model is updated using the model weights after incremental training; otherwise, the original model remains unchanged, and the corresponding data is retained to the next incremental training cycle.
[0086] The alarm threshold update cycle is set, including the first update cycle, the second update cycle, and the third update cycle, which increase sequentially. These correspond to the sensitivity coefficient k value adjustment process, the threshold adjustment strategy correction process, and the model update process, respectively. When the first update cycle is reached (one week in this embodiment), the current alarm threshold is updated according to the sensitivity coefficient adjustment formula based on the adjustment result of the current sensitivity coefficient k value. When the second update cycle is reached (one month in this embodiment), the threshold adjustment strategy is corrected according to the threshold optimization information and the adaptive correction formula, and the alarm threshold adjustment process in the subsequent first update cycle is executed based on the corrected strategy. When the third update cycle is reached (one quarter in this embodiment), the equipment quality difference classification results are updated based on the output results of the incrementally trained health assessment model and the remaining life prediction model, and the corresponding threshold benchmark parameters are adjusted according to the classification threshold matching formula. By integrating the updated threshold adjustment strategy, equipment quality difference classification results, and threshold benchmark parameters, the current alarm threshold is continuously updated, forming a multi-level adaptive adjustment system from weekly sensitivity fine-tuning and monthly strategy correction to quarterly model iteration, ensuring that the alarm threshold can continuously track changes in the actual operating status of the equipment and the operation and maintenance scenario.
[0087] Through the above implementation methods, the four threshold adjustment scenarios constructed by this invention can cover the full operating conditions and full lifecycle maintenance scenarios of dry bulk cargo terminal belt conveyor equipment, from normal operation, high-load operation, pre- and post-maintenance, to different quality batches. Each scenario provides a clear evaluation model or parameter calculation method, and illustrates the physical meaning, value source, and calculation process of the parameters with specific data examples. The two-layer adaptive optimization closed loop composed of sensitivity coefficient k-value feedback adaptive update and model incremental training can continuously optimize the alarm threshold adjustment strategy and prediction model accuracy based on the feedback of alarm result data. As verified by the above embodiments, this invention can control the false alarm rate to within 20% and the false alarm rate to within 10%, significantly improving the adaptability and accuracy of the alarm system compared with the fixed threshold mode, reducing the workload of maintenance personnel and the risk of unplanned downtime, and is suitable for the intelligent maintenance needs of belt conveyor equipment in dry bulk cargo terminals of different sizes.
[0088] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them. Although the present invention 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; and these 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 the present invention.
Claims
1. A dynamic threshold adaptive adjustment method suitable for dry bulk cargo terminal equipment, characterized in that, Includes the following steps: Based on the equipment operation data, historical maintenance data, and collaborative equipment operation data associated with the belt conveyor at the dry bulk terminal, the threshold adjustment scenario corresponding to the current belt conveyor is determined. The corresponding threshold adjustment strategy is invoked based on the threshold adjustment scenario, and the alarm threshold adjustment parameters under the current threshold adjustment scenario are determined by combining the alarm authenticity, equipment load status, degree of operation and maintenance impact, or equipment health status. The alarm threshold of the belt conveyor is dynamically adjusted according to the alarm threshold adjustment parameters, and anomaly detection is performed based on the adjusted alarm threshold to obtain alarm result data.
2. The dynamic threshold adaptive adjustment method according to claim 1, characterized in that, Threshold adjustment scenarios include feedback adjustment scenarios based on alarm accuracy and threshold adaptability, dynamic adjustment scenarios based on equipment load conditions, compensation adjustment scenarios based on operation and maintenance status, and differentiated setting scenarios based on equipment quality differences, among which: Based on the feedback adjustment scenario of alarm accuracy and threshold adaptability, the alarm accuracy change trend is analyzed according to historical alarm data and real-time alarm results. Combined with the adaptability relationship between the current alarm threshold and the equipment operating status, the optimization direction and adjustment range of the alarm threshold are determined. Based on the dynamic adjustment scenario of equipment load conditions, the impact of the current operating conditions on the alarm threshold is determined according to the changes in equipment load status, running time and key operating parameters, and the alarm threshold is dynamically corrected. Based on the compensation and adjustment scenario of operation and maintenance status, the impact of operation and maintenance operations on equipment operating characteristic parameters is assessed according to equipment maintenance and repair records and changes in equipment status after operation and maintenance, and the alarm threshold is compensated and corrected. Based on the differences in equipment quality, differentiated scenarios are set up. According to the equipment model, manufacturer, years of operation and historical operating status parameters, the individual characteristics of the equipment are determined, and corresponding alarm thresholds are configured based on the individual characteristics of the equipment.
3. The dynamic threshold adaptive adjustment method according to claim 2, characterized in that, For feedback adjustment scenarios based on alarm accuracy and threshold adaptability, the method for dynamically adjusting the current alarm threshold of the belt conveyor is as follows: Based on the component types of the belt conveyor, historical fault statistics, and preset threshold adjustment rules, determine the first adjustment step size and the second adjustment step size. Obtain the alarm accuracy rate corresponding to the current statistical period, and compare the alarm accuracy rate with the preset alarm accuracy rate lower limit and the preset alarm accuracy rate upper limit respectively; When the alarm accuracy rate is less than the preset alarm accuracy rate lower limit, the current alarm threshold is increased according to the first adjustment step size. When the alarm accuracy rate is greater than or equal to the preset lower limit of alarm accuracy rate and less than or equal to the preset upper limit of alarm accuracy rate, the current alarm threshold remains unchanged. When the alarm accuracy rate is greater than the preset alarm accuracy rate upper limit, the current alarm threshold is reduced based on the second adjustment step size.
4. The dynamic threshold adaptive adjustment method according to claim 2, characterized in that, For compensation adjustment scenarios based on operation and maintenance status, the method for dynamically adjusting the current alarm threshold of the belt conveyor is as follows: Obtain the current maintenance type and the degree of improvement in equipment operational stability of the belt conveyor; The corresponding threshold compensation coefficient is matched based on the type of operation and maintenance and the improvement of equipment operation stability, and the alarm threshold compensation amount is determined according to the threshold compensation coefficient. The current alarm threshold is adjusted based on the alarm threshold compensation amount to obtain the target alarm threshold; Obtain device alarm data after applying the target alarm threshold, and determine the alarm accuracy rate based on the device alarm data; Based on the comparison between the alarm accuracy rate and the preset accuracy threshold, the target alarm threshold is corrected to obtain the updated alarm threshold.
5. The dynamic threshold adaptive adjustment method according to claim 2, characterized in that, For scenarios with differentiated settings based on equipment quality variations, the method for dynamically adjusting the current alarm threshold of belt conveyors is as follows: Determine the corresponding classification adjustment parameters based on the equipment classification results; The alarm threshold adjustment step size is determined based on the classification adjustment parameters, and the current alarm threshold is adjusted based on the alarm threshold adjustment step size to obtain an initial adjusted alarm threshold that is compatible with the equipment category. Acquire equipment operation data, alarm result data, and fault data within a preset operating cycle after initial alarm threshold adjustment, and determine threshold adjustment evaluation parameters based on the equipment operation data, alarm result data, and fault data; The classification adjustment parameters are updated based on the threshold adjustment evaluation parameters to obtain the updated alarm threshold adjustment parameters. When the threshold adjustment evaluation parameters meet the preset compensation conditions, the current alarm threshold is compensated and adjusted based on the updated alarm threshold adjustment parameters.
6. The dynamic threshold adaptive adjustment method according to claim 2, characterized in that, The method for obtaining alarm result data is as follows: The current equipment operation data of the belt conveyor is obtained, and the equipment operation data is divided into multiple windows based on a preset sliding time window. Compare the device operation data sample values in each window's operation data with the adjusted alarm threshold, and calculate the proportion of device operation data sample values in each window's operation data that exceed the alarm threshold. The proportion of device operation data sampling values exceeding the alarm threshold is compared with a preset early warning proportion threshold and a preset fault proportion threshold, wherein the preset early warning proportion threshold is less than the preset fault proportion threshold; Determine the alarm level of the corresponding window's running data based on the comparison results; Based on the alarm level corresponding to the operating data of each window, the alarm judgment data corresponding to the operating data of each window is determined, and the alarm result data of the belt conveyor is generated based on the alarm judgment data corresponding to the operating data of each window.
7. The dynamic threshold adaptive adjustment method according to any one of claims 1-6, characterized in that, Also includes: The effectiveness of alarm threshold adjustment is evaluated based on alarm result data to determine threshold optimization information; The threshold adjustment strategy is adaptively corrected based on the threshold optimization information, and the alarm threshold is continuously updated according to the corrected threshold adjustment strategy.
8. The dynamic threshold adaptive adjustment method according to claim 7, characterized in that, The method for obtaining threshold optimization information is as follows: Obtain alarm result data for a preset evaluation period, and filter the alarm result data according to the scene identifier adjusted according to the corresponding threshold to obtain the target alarm result data; Based on the verification results of the target alarm results by the operation and maintenance personnel, the alarm precision, recall, F1 score and false alarm rate under the corresponding threshold adjustment scenario are obtained; The alarm precision, recall, F1 score, and false alarm rate are matched with the preset evaluation conditions of the corresponding threshold adjustment scenario to determine the evaluation results of the alarm threshold adjustment strategy. When the false alarm rate exceeds the preset upper limit or the recall rate is lower than the preset lower limit, the current alarm threshold adjustment strategy is determined to not meet the evaluation conditions, and threshold optimization information is generated.
9. The dynamic threshold adaptive adjustment method according to claim 7, characterized in that, The method for adaptively correcting the threshold adjustment strategy based on threshold optimization information is as follows: Obtain the threshold optimization information corresponding to the current threshold adjustment scenario, and parse the threshold optimization information to determine the adjustment direction deviation, adjustment magnitude deviation, and scenario judgment deviation; Based on the adjustment direction deviation, adjustment magnitude deviation, and scenario judgment deviation, the adjustment step size parameter, threshold floating parameter, and threshold baseline parameter in the threshold adjustment strategy corresponding to the current threshold adjustment scenario are adaptively corrected to obtain the corrected threshold adjustment strategy parameters. The modified threshold adjustment strategy parameters are constrained and verified to determine whether each parameter meets the corresponding preset parameter constraint conditions. When any parameter exceeds the preset correction range or the single adjustment range limit, the corresponding parameter is subject to boundary constraints to obtain the threshold adjustment strategy parameters that meet the constraint conditions. Based on the threshold adjustment strategy parameters that meet the constraints, the threshold adjustment strategy corresponding to the current threshold adjustment scenario is updated to form an adaptive threshold adjustment strategy suitable for the current threshold adjustment scenario.
10. The dynamic threshold adaptive adjustment method according to claim 7, characterized in that, The method for continuously updating the alarm threshold based on the revised threshold adjustment strategy is as follows: Set the alarm threshold update cycle, which includes a first update cycle, a second update cycle, and a third update cycle; When the first update cycle is reached, the current alarm threshold is updated based on the adjustment result of the sensitivity coefficient k value. When the second update cycle is reached, the threshold adjustment strategy is corrected based on the threshold optimization information, and the alarm threshold adjustment process in the subsequent first update cycle is executed based on the corrected threshold adjustment strategy. When the third update cycle is reached, the equipment quality difference classification results are updated based on the output results of the incrementally trained equipment health assessment model and equipment remaining life prediction model, and the corresponding threshold benchmark parameters are adjusted according to the updated equipment quality difference classification results. The current alarm threshold is continuously updated based on the updated threshold adjustment strategy, equipment quality difference classification results, and threshold benchmark parameters.
