Operation monitoring management system for laboratory ventilation equipment
By classifying the operating parameters of laboratory ventilation equipment into states and comparing their trajectories, abnormal transfer paths of the equipment can be identified, solving the problem that existing technologies cannot identify the overall operating status of the equipment, and realizing in-depth anomaly detection and monitoring of equipment operating modes.
Patent Information
- Application Number
- CN202512046153.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-31
- Publication Date
- 2026-02-24
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
Existing methods for monitoring the operation of laboratory ventilation equipment cannot effectively identify the overall operating status of the equipment, nor can they detect illogical transition sequences between legitimate states, leading to the omission of hidden faults.
The data acquisition module acquires the operating parameters of the ventilation equipment, the state division module divides the parameter sequence into discrete operating states, the real-time state matching module maps the current state, the state trajectory construction module traces back the historical state, and the anomaly detection module compares the state trajectory with the standard transfer map to identify abnormal transfer paths.
It enables quantitative identification and in-depth anomaly detection of the overall operating mode of equipment, and can discover hidden anomalies that violate normal operating patterns, thereby improving the accuracy and comprehensiveness of monitoring.
Smart Images
Figure CN121557573A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of intelligent monitoring technology for laboratory equipment, specifically to an operation monitoring and management system for laboratory ventilation equipment. Background Technology
[0002] Currently, the monitoring of laboratory ventilation equipment largely relies on threshold-based alarm systems or simple time-series data analysis. These traditional methods trigger alarms by setting fixed upper and lower limits for operating parameters, or by performing trend statistics on historical data to identify significant deviations. These methods can only capture momentary exceedances or slow drifts of parameters, and their monitoring logic remains at the level of judging isolated data points.
[0003] Existing technical solutions have shortcomings. They cannot effectively identify and define the overall "operating state" of equipment. The operation of ventilation equipment is a complex pattern formed by the interaction of multiple parameters. Simple threshold judgments fragment this process, leading to the omission of many operating conditions where parameters are within the normal range but the combination pattern is abnormal. These methods lack monitoring of the temporal logic of equipment behavior. Equipment anomalies may manifest as illogical and unexpected transition sequences between a series of legal states, and traditional methods lack the ability to model and verify the transition relationships between states, failing to detect such hidden faults. A technology is needed that can abstract continuous sequences of operating parameters into representative discrete states and perform deep anomaly detection on the dynamic process of equipment operation based on a normal state transition logic model. Summary of the Invention
[0004] The purpose of this invention is to provide an operation monitoring and management system for laboratory ventilation equipment to solve the problems mentioned in the background art.
[0005] To achieve the above objectives, the present invention provides an operation monitoring and management system for laboratory ventilation equipment, the system comprising: The data acquisition module is used to collect the operating parameters of the ventilation equipment within a continuous time interval and integrate the time-series records of each operating parameter into an initial state sequence. The state partitioning module is used to perform state partitioning on the initial state sequence to generate a series of discrete operating states, each of which corresponds to a standardized state description vector. The real-time status matching module is used to continuously collect real-time operating parameters and map them into current status description vectors during the monitoring process. Based on the matching degree between the current status description vector and each of the standardized status description vectors, the module determines the current belonging status of the real-time operating parameters from the series of discrete operating states. The state trajectory construction module is used to backtrack and retrieve historical state of a preset time length based on the current state of ownership, thereby forming a short-term state trajectory. An anomaly detection module is used to compare the short-term state trajectory with a pre-generated standard state transition map to identify whether there are abnormal transition paths in the short-term state trajectory that deviate from the standard state transition map.
[0006] Preferably, the state partitioning module generates a series of discrete operating states, each of which corresponds to a standardized state description vector, specifically including: In the initial state sequence, multiple parameter cluster centers are identified based on the numerical distribution density of the operating parameters; The equipment operating conditions represented by multiple time-series records belonging to the same parameter aggregation center are merged and defined as an independent discrete operating state. Calculate the average value of each operating parameter in all time-series records belonging to the discrete operating state, and arrange the average values in a preset order to form the standardized state description vector characterizing the discrete operating state.
[0007] Preferably, the real-time state matching module determines the current state of the real-time operating parameter based on the matching degree between the current state description vector and each of the standardized state description vectors, including: Calculate the spatial distance between the current state description vector and each of the normalized state description vectors; Select the standardized state description vector with the smallest spatial distance and mark the corresponding discrete operating state as a candidate state; If the spatial distance between the current state description vector and the standardized state description vector corresponding to the candidate state is less than a preset proximity threshold, then the candidate state is determined to be the current belonging state. If the spatial distance is greater than or equal to the proximity threshold, a new unconfirmed state is created for the current state description vector.
[0008] Preferably, after creating a new unconfirmed state for the current state description vector, the method further includes: During subsequent continuous monitoring cycles, operating parameters will continue to be collected and new status description vectors will be generated. Within a preset observation window, the frequency of each newly generated state description vector is determined as the new state to be confirmed. If the frequency exceeds a preset frequency threshold, the new unconfirmed state will be formally incorporated into the newly added discrete operating state, and the corresponding standardized state description vector will be calculated for the newly added discrete operating state. The newly added discrete operating states and their standardized state description vectors are updated to the series of discrete operating states.
[0009] Preferably, the construction process of the standard state transition map pre-generated by the anomaly detection module includes: During specific historical periods of normal equipment operation, the process of its state changes is recorded to form a set of historical state trajectories; Analyze the set of historical state trajectories and count the number of times a transition occurs from any discrete running state to another discrete running state; The state transition path that occurs more than a preset threshold number of times is drawn as a directed connection edge in the standard state transition graph; The starting and ending states connected by the connecting edges, along with the weights of the connecting edges, together constitute the standard state transition graph.
[0010] Preferably, the anomaly detection module identifies whether there are abnormal transition paths in the short-term state trajectory that deviate from the standard state transition map, including: In the short-term state trajectory, each continuous state change is extracted in chronological order; Check whether there is a corresponding directed connection edge between the starting state and the ending state in the standard state transition graph for each consecutive state change. If there is no corresponding directed connection edge in the standard state transition graph, then the subsequent state change is marked as a suspected abnormal transition; The number of suspected abnormal transitions in the short-term state trajectory is accumulated. If the accumulated number exceeds the trajectory abnormality threshold, the short-term state trajectory as a whole is determined to be an abnormal state trajectory.
[0011] Preferably, after accumulating the number of suspected abnormal transitions in the short-term state trajectory, the method further includes: When the short-term state trajectory is determined to be an abnormal state trajectory as a whole, all suspected abnormal transitions in the abnormal state trajectory are extracted. For each of the suspected abnormal transitions, calculate the numerical differences between the standardized state description vector corresponding to its starting state and the standardized state description vector corresponding to its ending state in each dimension of the operating parameters. Running parameters whose numerical differences exceed the preset difference threshold of the corresponding dimension are marked as key mutation parameters of the suspected abnormal transfer; An anomaly report is generated, which includes the time range of the anomaly state trajectory, the start and end status of each suspected anomaly transition, and the key variation parameters corresponding to each suspected anomaly transition.
[0012] Preferably, the data acquisition module acquires the operating parameters of the ventilation equipment over a continuous time interval, including: The system simultaneously collects the fan speed in the ventilation equipment, the real-time pressure difference in the ventilation duct, the pressure drop across the filter, and the temperature and humidity readings from the environmental sensors. The rotational speed, real-time pressure difference, pressure drop, and temperature and humidity readings are synchronously sampled at a fixed sampling period to form a timestamp-aligned multidimensional parameter time series.
[0013] Preferably, the data acquisition module integrates the time-series records of each operating parameter into an initial state sequence, including: Data cleaning is performed on the multidimensional parameter time series to remove invalid data points that are significantly beyond the physical range; The cleaned multidimensional parameter time series is subjected to smoothing filtering to generate a smoothed parameter series; On the smoothed parameter sequence, data segments of fixed length are extracted at equal time intervals, and the average value of all operating parameters in each data segment is taken as a time series record.
[0014] Preferably, after comparing the short-term state trajectory with the pre-generated standard state transition map, the anomaly detection module further includes: Based on the comparison results, the proximity threshold used to determine the current affiliation status of the real-time operating parameters is dynamically adjusted. If an abnormal state trajectory is detected, the value of the adjacent threshold will be temporarily reduced in the next monitoring cycle. If no abnormal state trajectory is detected in several consecutive monitoring cycles, the value of the adjacent threshold is gradually increased until it is restored to the default value.
[0015] Compared with the prior art, the beneficial effects of the present invention are: The system continuously collects multidimensional operating parameter sequences from ventilation equipment, processes them through state partitioning, and forms a series of discrete operating states with clear boundaries and representativeness. Each state corresponds to a comprehensive, standardized state description vector. In real-time monitoring, the system also maps the current operating parameter to a description vector and accurately determines its discrete operating state by calculating the matching degree between this vector and all preset standardized state description vectors. This method overcomes the limitation of traditional threshold monitoring, which can only determine whether a single parameter exceeds its limit. It achieves quantitative identification and classification of the overall operating mode of the equipment, elevating the assessment of operating conditions from scattered parameter points to an integrated state level, providing accurate and standardized input for in-depth analysis based on state sequences.
[0016] Based on the currently identified state, the system automatically traces back and stitches together a short-term state trajectory reflecting the recent evolution of equipment behavior, and compares this trajectory with a pre-generated standard state transition map. This standard map defines all possible, logically consistent transition relationships between discrete operating states under normal conditions. Through comparison, the system can detect whether there are state transition sequences in the short-term state trajectory that are not permitted by the standard map or have extremely low probability. The core of this detection mechanism lies in focusing on the dynamic logic of state changes rather than static values; therefore, it can effectively discover hidden anomalies where each individual state parameter is normal, but the state transition sequence violates the normal operating rules of the equipment. Attached Figure Description
[0017] Figure 1 This is a schematic diagram of the working principle of the operation monitoring and management system for laboratory ventilation equipment described in this invention. Figure 2 A flowchart for generating discrete operating states and standardized state description vectors; Figure 3 A flowchart for determining the current attribution status; Figure 4 A heat map showing the correlation between operating parameters of laboratory ventilation equipment; Figure 5 This is a graph showing the trend of changes in the dynamic proximity threshold and the number of consecutive normal cycles. Detailed Implementation
[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. 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.
[0019] Please see Figure 1 This invention provides an operation monitoring and management system for laboratory ventilation equipment. The system includes: a data acquisition module, a state division module, a real-time state matching module, a state trajectory construction module, and an anomaly detection module operating collaboratively. Specific implementation details are as follows: The data acquisition module collects the operating parameters of the ventilation equipment within a continuous time interval and integrates the time-series records of each operating parameter into an initial state sequence. The state division module performs state division on the initial state sequence, generating a series of discrete operating states. Each discrete operating state corresponds to a standardized state description vector. The real-time state matching module continuously collects real-time operating parameters during the monitoring process and maps them to the current state description vector. Based on the matching degree between the current state description vector and each of the standardized state description vectors, the current state of the real-time operating parameter is determined from the series of discrete operating states. The state trajectory construction module uses the current state as a base point and backtracks to retrieve historical states of a preset time length to form a short-term state trajectory. The anomaly detection module compares the short-term state trajectory with a pre-generated standard state transition map to identify whether there are abnormal transition paths in the short-term state trajectory that deviate from the standard state transition map.
[0020] Example 1: See Figure 2 In specific implementation, the state partitioning module processes the initial state sequence generated by the data acquisition module. This initial state sequence consists of a large number of time-series records containing multi-dimensional operating parameters arranged chronologically. Specifically, the state partitioning module uses a clustering algorithm to analyze the distribution of all operating parameters in the numerical space of the initial state sequence. The clustering algorithm measures the numerical differences between time-series records based on Euclidean distance and identifies regions where record points are relatively clustered in the parameter space according to a preset density threshold. In specific implementations, a noisy density-based clustering method can be used, the core idea of which is to guide the clustering process by assessing the probability that each time-series record will become a parameter cluster center. In some embodiments, the state partitioning module merges multiple time-series records identified by the clustering algorithm that belong to the same parameter cluster center into a group. These time-series records are numerically close to each other, reflecting the steady-state or quasi-steady-state operating conditions of the ventilation equipment under specific load, wind resistance, and environmental conditions. Each such group represents an independent discrete operating state.
[0021] In practical implementation, to quantitatively assess the likelihood of each time-series record serving as a parameter clustering center, the system employs an evaluation index based on local density and relative distance. The calculation method for this index is as follows:
[0022] Where: symbol Indicates that for the first Cluster center evaluation value of each time-series record, symbol Indicates the first A time series record is the center, and the radius is [missing information]. The total number of other time-series records contained in the neighborhood of the symbol. Indicates the first Each time series record is updated to its most recent one with a higher time series. Euclidean distance between time-series records of values, sign Represents all time-series records The maximum value among all time-series records. In practice, this is achieved by calculating the maximum value among all time-series records. Values can be selected. Time-series records with values greater than a preset threshold are used as candidate parameter cluster centers; these points typically also have high local density. and a large relative distance In specific implementation, parameters The values need to be adjusted and set according to the numerical range and number of dimensions of the initial state sequence.
[0023] In practical implementation, each defined discrete operating state needs to be quantitatively represented by a standardized numerical vector. For each discrete operating state, the state partitioning module calculates the average value of each operating parameter across all time-series records belonging to that state. In practice, operating parameters may include fan speed, pipeline pressure difference, filter pressure drop, and ambient temperature and humidity. The calculation process involves summing the values of all time-series records for that parameter dimension and then dividing by the total number of time-series records. After calculating the average values of all operating parameters, the state partitioning module arranges these average values into a one-dimensional vector according to a pre-defined order. This order remains unchanged throughout the system's lifecycle to ensure consistency. The resulting vector is the standardized state description vector representing the discrete operating state, which serves as the benchmark for comparison with real-time data in subsequent real-time state matching.
[0024] Example 2: See Figure 3In practical implementation, the real-time state matching module determines the current state of the real-time operating parameter based on the matching degree between the current state description vector and each standardized state description vector. In practice, the matching degree is quantified by calculating the spatial distance between vectors. The real-time state matching module sequentially calculates the Euclidean distance between the current state description vector and each standardized state description vector defined in the system. After calculation, the real-time state matching module selects the smallest distance value from all calculated distance values and marks the discrete operating state to which the standardized state description vector corresponding to this smallest distance value belongs as a candidate state. In practice, the real-time state matching module has a preset proximity threshold. It compares the spatial distance between the current state description vector and the standardized state description vector corresponding to the candidate state with the proximity threshold. If the comparison result shows that the spatial distance is less than the proximity threshold, the real-time state matching module determines the candidate state as the current state of the real-time operating parameter. In practice, if the comparison result shows that the spatial distance is greater than or equal to the value of the nearest threshold, the real-time state matching module considers that the current state description vector is significantly different from the standardized state description vectors of all existing discrete operating states. At this time, the real-time state matching module creates a new unconfirmed state for the current state description vector.
[0025] In practice, after creating a new pending-confirmation state for the current state description vector, the system enters the observation and confirmation process for the new state. In subsequent continuous monitoring cycles, the system continues to collect operating parameters of the ventilation equipment through the data acquisition module, and the real-time state matching module generates new state description vectors. In practice, the system sets a fixed-length observation window. Within this observation window, the real-time state matching module counts the number of times each newly generated state description vector is determined to belong to the aforementioned new pending-confirmation state. In some embodiments, the real-time state matching module uses a proportional parameter to characterize the frequency of new pending-confirmation states occurring within the observation window, calculated as follows:
[0026] Where: symbol This indicates the proportion of new pending confirmation states appearing within the observation window; the symbol is... This represents the total number of times a newly generated state description vector is determined to belong to a new unconfirmed state within the observation window's time frame. The symbol is... This represents the total number of state description vectors generated by the system within the observation window's time frame. In practical implementation, the real-time state matching module will use the calculated matching ratio. It is compared with a preset frequency threshold, which can optionally be a specific proportional value. It can be understood that if the matching proportion... If the frequency exceeds a preset threshold, it indicates that the operating mode representing the new unconfirmed state continuously and repeatedly occurs within the observation window. Based on this, the system determines that the operating mode is not accidental noise or a transient anomaly. In specific implementations, when the above conditions are met, the system formally incorporates this new unconfirmed state as a newly added discrete operating state. In some embodiments, the system calculates the corresponding standardized state description vector for the newly added discrete operating state. The calculation method is to obtain all state description vectors determined to belong to this state within the observation window, and then calculate the arithmetic mean of these vectors on each parameter dimension, thereby forming the standardized state description vector of the newly added discrete operating state. In specific implementations, the state division module or system management module updates the newly added discrete operating state and its standardized state description vector to a series of discrete operating state sets maintained by the system. Subsequent real-time state matching will be based on the updated state set. Optionally, if the matching ratio is high after a complete observation window ends, the system will consider the following: If the preset frequency threshold is not exceeded, the system will abandon this new pending confirmation status and will no longer track it.
[0027] Example 3: In a specific implementation, the construction process of the standard state transition map pre-generated by the anomaly detection module is based on historical data of normal equipment operation. In this implementation, during a specific historical period when the ventilation equipment is confirmed to be operating normally, the system continuously records the discrete operating states output by the real-time state matching module and records the state change process as a state sequence in chronological order. A large number of such state sequences together constitute a historical state trajectory set. In this implementation, the analysis process traverses each state trajectory in the historical state trajectory set, counting the number of times a state transitions from one discrete operating state to another. For example, counting the number of times a state transitions from "S1" to "S2", and from "S1" to "S3", until all possible state pairs are covered. In this implementation, the system presets a threshold number to filter out frequently occurring state transitions that can be considered normal patterns. In this implementation, for any two discrete operating states "Sa" and "Sb", the condition for determining whether to draw a directed connection edge from "Sa" to "Sb" in the standard state transition map is as follows:
[0028] Where: symbol This represents the total number of times a discrete running state "Sa" transitions to a discrete running state "Sb" within the historical state trajectory set. The symbol is... This represents a preset threshold number of occurrences. In practice, this threshold is set only if the state transition from "Sa" to "Sb" occurs a certain number of times. Greater than or equal to the number of times threshold Only when this occurs will the system establish a directed connection edge in the standard state transition graph, starting at "Sa" and ending at "Sb". Optionally, the directed connection edge can be assigned a weight value, which can be related to the number of occurrences. Positive correlation.
[0029] In specific implementation, the anomaly detection module identifies whether there are abnormal transition paths deviating from the standard state transition graph in the short-term state trajectory based on graph comparison. Specifically, for a short-term state trajectory submitted by the state trajectory construction module, the anomaly detection module extracts each consecutive state change in chronological order. Each consecutive state change includes a starting state and an ending state. For each extracted consecutive state change, the anomaly detection module queries the pre-generated standard state transition graph to check for a directed connection edge that starts from the current starting state and ends at the current ending state. In some embodiments, if no corresponding directed connection edge exists in the standard state transition graph, it indicates that this state transition has not occurred or rarely occurs in the high-frequency mode during normal device operation, and the anomaly detection module marks this consecutive state change as a suspected abnormal transition. In specific implementation, the anomaly detection module accumulates the number of occurrences of all suspected abnormal transitions marked after traversing the entire short-term state trajectory. In specific implementation, the system presets a trajectory anomaly threshold. If the accumulated number of suspected abnormal transitions exceeds the trajectory anomaly threshold, the anomaly detection module determines the entire short-term state trajectory as an abnormal state trajectory.
[0030] In practical implementation, when the short-term state trajectory is determined to be an abnormal trajectory overall, the system further analyzes the abnormal details. Specifically, the anomaly detection module extracts all records marked as potentially abnormal transitions from the short-term state trajectories already identified as abnormal. For each potentially abnormal transition, the anomaly detection module obtains the standardized state description vector corresponding to its starting state and the standardized state description vector corresponding to its ending state. The anomaly detection module calculates the numerical difference between the two standardized state description vectors across various operational parameter dimensions; that is, it subtracts the value of the starting state vector in the same dimension from the value of the ending state vector in a certain dimension, and takes the absolute value to obtain the numerical difference for that dimension. In some embodiments, the system presets a difference threshold for each operational parameter dimension, and the anomaly detection module compares the calculated numerical difference for each dimension with the preset difference threshold for that dimension. For operational parameters whose numerical differences exceed the preset difference threshold for their corresponding dimension, the anomaly detection module marks them as key variation parameters for that potentially abnormal transition. Optionally, the final anomaly detection module generates a structured anomaly report, which includes the start and end times of the anomaly trajectory, the specific start and end state identifiers of each suspected anomaly transition, and a list of names of one or more key variation parameters corresponding to each suspected anomaly transition.
[0031] Example 4: In specific implementation, the process of the data acquisition module collecting the operating parameters of the ventilation equipment within a continuous time interval involves the synchronous acquisition of multiple source signals. The data acquisition module, through the equipment controller interface and sensor network, synchronously collects the real-time speed of the fan, the real-time pressure difference between specific measuring points in the ventilation duct, the pressure drop between the inlet and outlet of the filter, and the temperature and humidity readings from environmental sensors placed within the ventilation environment. In specific implementation, the data acquisition module synchronously samples all the above operating parameters at a fixed sampling period, which can be set to one second or several seconds. At each sampling moment, the data acquisition module marks the speed, real-time pressure difference, pressure drop, temperature, and humidity readings read from different sensors with a unified timestamp. After continuous sampling, multiple parameter sequences with strictly aligned timestamps and arranged in chronological order are formed. These sequences together constitute a multidimensional parameter time series. Refer to Table 1 for the composition of the multidimensional parameter time series.
[0032] Table 1: Multidimensional Parameter Time Series Fragment Table
[0033] In practical implementation, the process by which the data acquisition module integrates the time-series records of various operating parameters into an initial state sequence includes data cleaning and normalization steps. The data acquisition module cleans the multi-dimensional parameter time series, setting upper and lower limits for the effective range based on the physical range of each type of operating parameter. For any parameter value at a given timestamp, if it exceeds its corresponding effective range, this data point is determined to be invalid and discarded. In practical implementation, the logic for determining the validity of data points can be expressed through the following formal conditions:
[0034] Where: symbol Indicates the first The operating parameter in the first A validity flag for each timestamp data point, with a value of 1 indicating validity and a value of 0 indicating invalidity. Indicates the first The operating parameter in the first The original collected values of each timestamp, symbol and They represent the first The effective lower and upper limits of each operating parameter. In specific implementation, the data acquisition module will set all... Data points with a value of 0 are removed from the multidimensional parameter time series. In some embodiments, the cleaned multidimensional parameter time series is subjected to smoothing filtering using a moving median filtering algorithm to eliminate random noise interference and generate a smoothed parameter series. In a specific implementation, the moving median filtering algorithm is based on defining a fixed-size sliding window to process the multidimensional parameter time series. The width of the sliding window can be set to 5 sampling points. For each data point in the time series, the algorithm selects the values of all adjacent sampling points within the sliding window centered on that point, sorts these values in ascending or descending order, and then takes the median value of the sorted sequence as the new smoothed value of that data point. Optionally, the window width of the moving average can be set to 5 sampling points. In a specific implementation, data segments of fixed time length are extracted from the smoothed parameter series at equal time intervals. The fixed time length can be five minutes or ten minutes. The average value of all operating parameters within each data segment is calculated as a time series record. When calculating the average value, for each operating parameter, the sum of the values of that parameter at all sampling times within the data segment is divided by the number of sampling points.
[0035] See Figure 4In the correlation analysis of operating parameters, the heatmap uses the correlation coefficient as the core indicator to quantify the linear correlation between key operating parameters of ventilation equipment (fan speed, real-time pipeline pressure difference, filter pressure drop, ambient temperature, and ambient humidity). In practice, the correlation between each parameter is visually presented through a color gradient (red-white-blue): red represents a strong positive correlation (correlation coefficient = 1.00 indicates perfect correlation), blue represents a strong negative correlation, and white corresponds to a weak correlation or no correlation. The analysis results show that the real-time pipeline pressure difference and filter pressure drop exhibit a weak positive correlation (correlation coefficient 0.18), reflecting a certain correlation between pipeline pressure changes and filter resistance. The absolute values of the correlation coefficients between environmental parameters (temperature and humidity) and equipment operating parameters (fan speed and pipeline pressure difference) are all below 0.1, indicating that environmental factors have a weak linear impact on the current operating status of the equipment. Regarding parameter configuration, the correlation coefficient calculation is based on a multi-dimensional parameter time series generated by the data acquisition module (effective data after cleaning and smoothing) to ensure the statistical reliability of the indicators.
[0036] Example 5: In a specific implementation, after the anomaly detection module compares the short-term state trajectory with the pre-generated standard state transition map, the system dynamically adjusts the proximity threshold parameter based on the comparison results. The proximity threshold is a key judgment parameter used by the real-time state matching module to determine the current state of real-time operating parameters. In this implementation, the goal of dynamic adjustment is to enable the system's state matching sensitivity to adapt to changes in equipment operating conditions. When the system frequently detects abnormal state trajectories, lowering the proximity threshold makes the state matching conditions more stringent, making it easier to classify real-time data into new unconfirmed states and facilitating early detection of abnormal patterns. In this implementation, if the system does not detect abnormal state trajectories within several consecutive monitoring cycles, the proximity threshold is gradually increased to moderately relax the state matching conditions. This helps to classify minor fluctuations in equipment operating conditions within the normal range into known discrete operating states, avoiding excessive interference from unconfirmed states. In some embodiments, the dynamic adjustment of the proximity threshold follows a feedback rule based on recent comparison results. The system records the comparison conclusion of each short-term state trajectory and maintains a sliding window that records recent anomaly occurrences.
[0037] In practical implementation, if an abnormal state trajectory is identified, the system temporarily lowers the neighbor threshold value in the next monitoring cycle. This lowering is achieved by multiplying a reduction factor less than 1 by the current neighbor threshold value. The reduced neighbor threshold takes effect immediately and is used for subsequent real-time state description vector matching calculations. In this implementation, temporarily lowering the neighbor threshold makes the system more sensitive to differences in state descriptions. The real-time state description vector needs to be closer to the existing standardized state description vector to be matched. Any relatively large deviation will lead to the creation of new unconfirmed states, increasing the probability of the system capturing potential abnormal patterns or new operating conditions. Optionally, the specific value of the reduction factor can be fine-tuned according to the severity of the identified abnormal state trajectory. In practical implementation, if no abnormal state trajectory is identified in several consecutive monitoring cycles, the system gradually increases the neighbor threshold value. This increase is achieved by multiplying a recovery factor greater than 1 by the current neighbor threshold value. After each adjustment, the neighbor threshold value increases, and the tolerance range for state matching also expands accordingly. In some embodiments, the gradual increase process has an upper limit, that is, the neighboring threshold value cannot exceed a preset default value during the increase process. When the neighboring threshold is restored to the default value through multiple adjustments, the adjustment process is paused and the system enters the stable monitoring phase.
[0038] In practice, the dynamic adjustment process of the proximity threshold can be described by a formalized update rule. This rule is triggered after each anomaly detection of the short-term state trajectory is completed. The expression of the update rule is as follows:
[0039] Where: symbol This indicates the adjusted proximity threshold to be used in the next monitoring cycle. (Symbol) Indicates the nearest threshold currently in use during the monitoring period, symbol This represents the system's preset default neighbor threshold, function. This indicates that the smaller of the two values is used to ensure that the adjusted threshold does not exceed the default value. (Symbol) It is a binary indicator function; if an abnormal state trajectory is detected in the current monitoring cycle, then... The value is 1 if it is positive and 0 otherwise. (Symbol) This is an indicator related to a continuous normal cycle. If no anomalies are detected in the current monitoring cycle, and the cycle has been normal for several consecutive cycles, then... It can be set to a constant greater than 1, otherwise it is 1. (Symbol) It is the reduction factor when the anomaly occurs, a positive number less than 1, with the sign... It is the recovery coefficient under continuous normal conditions, and is a positive number greater than or equal to 1. Optionally, The value can increase stepwise as the number of consecutive normal cycles increases.
[0040] See Figure 5 During the dynamic proximity threshold adjustment process of ventilation equipment operation monitoring, the correlation and change trends of the proximity threshold, default threshold, abnormal cycle marker, and number of consecutive normal cycles are visually presented. Specifically, the blue curve in the figure represents the dynamic proximity threshold, and its fluctuation is strongly correlated with the abnormal cycle (marked by a red star): when an abnormal cycle is identified, the dynamic proximity threshold will temporarily decrease (such as around 01-0112 or 01-0212), reflecting the role of the reduction coefficient when an anomaly occurs; while in the continuous anomaly-free phase, the dynamic proximity threshold gradually rises, eventually approaching the default threshold shown by the red dashed line, and the rise process is constrained by "not exceeding the default value" (which can be verified by the dynamic proximity threshold repeatedly approaching the default threshold in the figure). The green dashed line (right vertical axis) represents the number of consecutive normal cycles, and its value increases with the accumulation of anomaly-free cycles and is reset to 0 when an abnormal cycle occurs. This change echoes the adjustment logic of the dynamic proximity threshold: the increase in the number of consecutive normal cycles triggers the effect of the recovery coefficient, pushing the dynamic proximity threshold back to the default value; while the occurrence of an abnormal cycle interrupts this process, driving the threshold to temporarily decrease. The fluctuation characteristics of the parameters in the figure clearly reflect the dynamic adjustment mechanism of "tightening the matching conditions when there are abnormalities and relaxing the tolerance when there are continuous normal conditions", which provides visual support for sensitive monitoring and stable matching of the operating status of ventilation equipment.
[0041] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.
[0042] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.
Claims
1. A monitoring and management system for the operation of laboratory ventilation equipment, characterized in that, The system includes: The data acquisition module is used to collect the operating parameters of the ventilation equipment within a continuous time interval and integrate the time-series records of each operating parameter into an initial state sequence. The state partitioning module is used to perform state partitioning on the initial state sequence to generate a series of discrete operating states, each of which corresponds to a standardized state description vector. The real-time status matching module is used to continuously collect real-time operating parameters and map them into current status description vectors during the monitoring process. Based on the matching degree between the current status description vector and each of the standardized status description vectors, the module determines the current belonging status of the real-time operating parameters from the series of discrete operating states. The state trajectory construction module is used to backtrack and retrieve historical state of a preset time length based on the current state of ownership, thereby forming a short-term state trajectory. An anomaly detection module is used to compare the short-term state trajectory with a pre-generated standard state transition map to identify whether there are abnormal transition paths in the short-term state trajectory that deviate from the standard state transition map.
2. The operation monitoring and management system for laboratory ventilation equipment according to claim 1, characterized in that, The state partitioning module generates a series of discrete operating states, each of which corresponds to a standardized state description vector, specifically including: In the initial state sequence, multiple parameter cluster centers are identified based on the numerical distribution density of the operating parameters; The equipment operating conditions represented by multiple time-series records belonging to the same parameter aggregation center are merged and defined as an independent discrete operating state. Calculate the average value of each operating parameter in all time-series records belonging to the discrete operating state, and arrange the average values in a preset order to form the standardized state description vector characterizing the discrete operating state.
3. The operation monitoring and management system for laboratory ventilation equipment according to claim 2, characterized in that, The real-time state matching module determines the current state of the real-time operating parameters based on the matching degree between the current state description vector and each of the standardized state description vectors, including: Calculate the spatial distance between the current state description vector and each of the normalized state description vectors; Select the standardized state description vector with the smallest spatial distance and mark the corresponding discrete operating state as a candidate state; If the spatial distance between the current state description vector and the standardized state description vector corresponding to the candidate state is less than a preset proximity threshold, then the candidate state is determined to be the current belonging state. If the spatial distance is greater than or equal to the proximity threshold, a new unconfirmed state is created for the current state description vector.
4. The operation monitoring and management system for laboratory ventilation equipment according to claim 3, characterized in that, After creating a new pending state for the current state description vector, the process further includes: During subsequent continuous monitoring cycles, operating parameters will continue to be collected and new status description vectors will be generated. Within a preset observation window, the frequency of each newly generated state description vector is determined as the new state to be confirmed. If the frequency exceeds a preset frequency threshold, the new unconfirmed state will be formally incorporated into the newly added discrete operating state, and the corresponding standardized state description vector will be calculated for the newly added discrete operating state. The newly added discrete operating states and their standardized state description vectors are updated to the series of discrete operating states.
5. The operation monitoring and management system for laboratory ventilation equipment according to claim 1, characterized in that, The construction process of the pre-generated standard state transition map by the anomaly detection module includes: During specific historical periods of normal equipment operation, the process of its state changes is recorded to form a set of historical state trajectories; Analyze the set of historical state trajectories and count the number of times a transition occurs from any discrete running state to another discrete running state; The state transition path that occurs more than a preset threshold number of times is drawn as a directed connection edge in the standard state transition graph; The starting and ending states connected by the connecting edges, along with the weights of the connecting edges, together constitute the standard state transition graph.
6. The operation monitoring and management system for laboratory ventilation equipment according to claim 5, characterized in that, The anomaly detection module identifies whether there are abnormal transition paths in the short-term state trajectory that deviate from the standard state transition map, including: In the short-term state trajectory, each continuous state change is extracted in chronological order; Check whether there is a corresponding directed connection edge between the starting state and the ending state in the standard state transition graph for each consecutive state change. If there is no corresponding directed connection edge in the standard state transition graph, then the subsequent state change is marked as a suspected abnormal transition; The number of suspected abnormal transitions in the short-term state trajectory is accumulated. If the accumulated number exceeds the trajectory abnormality threshold, the short-term state trajectory as a whole is determined to be an abnormal state trajectory.
7. The operation monitoring and management system for laboratory ventilation equipment according to claim 6, characterized in that, After accumulating the number of suspected abnormal transitions in the short-term state trajectory, the method further includes: When the short-term state trajectory is determined to be an abnormal state trajectory as a whole, all suspected abnormal transitions in the abnormal state trajectory are extracted. For each of the suspected abnormal transitions, calculate the numerical differences between the standardized state description vector corresponding to its starting state and the standardized state description vector corresponding to its ending state in each dimension of the operating parameters. Running parameters whose numerical differences exceed the preset difference threshold of the corresponding dimension are marked as key mutation parameters of the suspected abnormal transfer; An anomaly report is generated, which includes the time range of the anomaly state trajectory, the start and end status of each suspected anomaly transition, and the key variation parameters corresponding to each suspected anomaly transition.
8. The operation monitoring and management system for laboratory ventilation equipment according to claim 1, characterized in that, The data acquisition module collects the operating parameters of the ventilation equipment over a continuous time interval, including: The system simultaneously collects the fan speed in the ventilation equipment, the real-time pressure difference in the ventilation duct, the pressure drop across the filter, and the temperature and humidity readings from the environmental sensors. The rotational speed, real-time pressure difference, pressure drop, and temperature and humidity readings are synchronously sampled at a fixed sampling period to form a timestamp-aligned multidimensional parameter time series.
9. A monitoring and management system for the operation of laboratory ventilation equipment according to claim 8, characterized in that, The data acquisition module integrates the time-series records of each operating parameter into an initial state sequence, including: Data cleaning is performed on the multidimensional parameter time series to remove invalid data points that are significantly beyond the physical range; The cleaned multidimensional parameter time series is subjected to smoothing filtering to generate a smoothed parameter series; On the smoothed parameter sequence, data segments of fixed length are extracted at equal time intervals, and the average value of all operating parameters in each data segment is taken as a time series record.
10. A monitoring and management system for the operation of laboratory ventilation equipment according to claim 1, characterized in that, The anomaly detection module, after comparing the short-term state trajectory with the pre-generated standard state transition map, also includes: Based on the comparison results, the proximity threshold used to determine the current affiliation status of the real-time operating parameters is dynamically adjusted. If an abnormal state trajectory is detected, the value of the adjacent threshold will be temporarily reduced in the next monitoring cycle. If no abnormal state trajectory is detected in several consecutive monitoring cycles, the value of the adjacent threshold is gradually increased until it is restored to the default value.