A time sequence monitoring data anomaly identification and perception method for a biological fermentation process

Abstract

This invention discloses a method for anomaly identification and sensing in time-series monitoring data of a bio-fermentation process, belonging to the field of anomaly identification technology. It addresses the problem of reduced final product concentration and fermentation process stability. By performing stage-sensing processing on the fermentation process, the method obtains the stage switching intervals of the synthesis pathway. Based on this, the fermentation process is divided into different metabolic stages. A correlation analysis window is constructed within each metabolic stage, and the Pearson correlation coefficient is used to quantify the synchronous correlation relationship of multi-source fermentation data. Simultaneously, a structural aggregation index of the fermentation data is constructed by combining time-lag information, and a baseline fluctuation range is set. Anomalies are defined as states that deviate from the normal correlation topology of the current metabolic stage. This method effectively distinguishes between stage switching of the synthesis pathway and metabolic imbalance anomalies, improving the accuracy and stability of anomaly sensing without interfering with normal metabolic transitions.

Description

Technical Field

[0001] This invention relates to the field of anomaly recognition technology, and more specifically, to a method for anomaly recognition and sensing of time-series monitoring data in a biological fermentation process. Background Technology

[0002] Protocatechuic acid, as an important aromatic organic acid, has wide applications in pharmaceutical intermediates, food antioxidants, and fine chemicals. In current industrial production, protocatechuic acid is mostly produced by microbial fermentation with glucose, vanillic acid, or p-hydroxybenzoic acid as precursor substrates. Existing fermentation process monitoring and anomaly identification technologies are mostly based on univariate static thresholds, empirical upper and lower limits, or moving average drift judgment methods to independently monitor parameters such as DO, pH, and substrate concentration.

[0003] The existing technology has the following shortcomings:

[0004] Currently, existing technologies mostly employ independent monitoring methods based on single-parameter thresholds or simple statistical features. These methods do not fully consider the dynamic correlation structure and time-delay evolution characteristics of multi-source fermentation parameters in different metabolic stages of the protocatechuic acid synthesis pathway, which arise from the redistribution of metabolic flux. This makes it difficult to effectively distinguish the temporal similarities between normal metabolic transition fluctuations and abnormal conditions such as metabolic imbalances caused by the switching of synthesis pathway stages. Consequently, frequent misjudgments of abnormalities during the switching of synthesis stages lead to unnecessary process interventions, disrupting the protocatechuic acid accumulation window and reducing the concentration of the final product and the stability of the fermentation process. Therefore, this paper proposes a time-series monitoring data anomaly identification and sensing method for biofermentation processes.

[0005] The information disclosed in the background section is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0006] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide a method for identifying and sensing anomalies in time-series monitoring data of a bio-fermentation process. This method utilizes a time-series data segmentation modeling mechanism based on metabolic stage perception, combines multi-parameter synchronous correlation analysis and time delay intensity quantification with a sliding time window, and employs a structural deviation discrimination strategy centered on the correlation structure aggregation index and its benchmark tensor to address the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides the following technical solution: a method for identifying and sensing anomalies in time-series monitoring data of a bio-fermentation process, comprising the following steps:

[0008] Step S1: During the fermentation of protocatechuic acid, monitor the fermentation data of protocatechuic acid and perform stage sensing processing to obtain the stage switching intervals in the fermentation process of protocatechuic acid.

[0009] Step S2: Divide the protocatechuic acid fermentation process into different metabolic stages based on the stage switching interval, set up a correlation analysis window in each metabolic stage, calculate the synchronous correlation coefficient of fermentation data in the correlation analysis window, and statistically analyze the time lag information of fermentation data.

[0010] Step S3: Generate the time lag strength of fermentation data based on the time lag information, analyze the structural aggregation index of fermentation data in combination with the synchronization correlation coefficient, perform cluster analysis on the structural aggregation index, and obtain the baseline tensor of fermentation data based on the clustering results;

[0011] Step S4: Set the baseline fluctuation range for fermentation data, use the baseline tensor to evaluate the structural deviation rate of fermentation data at each metabolic stage, analyze the metabolic state of each metabolic stage in combination with the baseline fluctuation range, and determine whether to output a sensing abnormality alarm signal based on the metabolic state.

[0012] In a preferred embodiment, in step S1, during the protocatechuic acid fermentation process, fermentation data during the protocatechuic acid fermentation process is continuously collected by an online monitoring device connected to the fermentation reactor.

[0013] Fermentation data include dissolved oxygen concentration, fermentation broth pH, ​​respiratory quotient, and fermentation substrate concentration. Based on the fermentation substrate concentration, the rate of change of fermentation substrate concentration over time is calculated to obtain the fermentation substrate consumption rate.

[0014] In a preferred embodiment, in step S1, after obtaining the fermentation data, the fermentation data undergoes stage-aware processing in the time series dimension:

[0015] Centered on the current sampling time, a local sensing window with a length of several sampling periods is constructed. The first-order rate of change of dissolved oxygen concentration, fermentation broth pH value and respiratory quotient within the local sensing window are calculated to obtain the rate of change of dissolved oxygen concentration, the rate of change of fermentation broth pH value and the rate of change of respiratory quotient.

[0016] When the rate of change of dissolved oxygen concentration, the rate of change of pH value of fermentation broth, and the rate of change of respiratory quotient meet the condition that the proportion of the same sign is not lower than the preset consistency ratio threshold within the local sensing window, it is determined that the fermentation data within the local sensing window are consistent in the direction of change.

[0017] In a preferred embodiment, in step S1, the average fermentation substrate consumption rate is calculated within the same local sensing window, and the difference between the average fermentation substrate consumption rate and the average fermentation substrate consumption rate corresponding to the previous local sensing window is calculated.

[0018] When the absolute value of the difference is greater than the preset threshold for metabolic change, it is determined that the fermentation substrate consumption level has been significantly adjusted.

[0019] When both the consistency of change direction and the significant change in fermentation substrate consumption rate are met within the same local perception window, it is determined that the distribution of microbial metabolic flux during protocatechuic acid fermentation has changed within the corresponding time period.

[0020] The time ranges of multiple consecutive local perception windows that meet the judgment conditions are merged to obtain the corresponding time interval, and the time interval is marked as the stage switching interval.

[0021] In a preferred embodiment, in step S2, the complete time series of the protocatechuic acid fermentation process is segmented using the stage switching interval as the time boundary, and the protocatechuic acid fermentation process is divided into three metabolic stages.

[0022] Multiple correlation analysis windows are set within each metabolic stage. The correlation analysis window is a time interval that slides along the time axis within the corresponding metabolic stage according to a preset window length.

[0023] Within each correlation analysis window, the Pearson correlation coefficient calculation method is used to calculate the synchronous correlation coefficient of the fermentation data during the protocatechuic acid fermentation process;

[0024] Statistical analysis of fermentation data time lag information is performed within the same correlation analysis window;

[0025] Within a preset time lag range, the Pearson coefficients of discrete time series of different fermentation data under different lag conditions are calculated, and the time lag at which the Pearson coefficient reaches its maximum value is taken as the time lag information.

[0026] In a preferred embodiment, in step S3, the average value of each time lag quantity is taken as the time lag intensity within the correlation analysis window.

[0027] The product of the time lag strength and the synchronization correlation coefficient after standardization is used as the time-series correlation strength factor.

[0028] The product of the temporal correlation strength factor and the preset structure adjustment coefficient is used as the structure aggregation index of the fermentation data.

[0029] In a preferred embodiment, in step S3, cluster analysis is performed on the structural aggregation indices formed by each association analysis window within each metabolic stage;

[0030] Each association analysis window is used as a clustering sample, and the structural aggregation index corresponding to the association analysis window is used as the sample feature of the sample.

[0031] The aggregation index distance is calculated based on the structural aggregation index of any two clustered samples. If the aggregation index distance is less than the preset aggregation distance threshold, the two clustered samples are determined to be structurally similar candidate windows.

[0032] Clusters are obtained by aggregating candidate windows based on their temporal dimension;

[0033] Count the number of cluster samples contained in each cluster, and select the structural benchmark cluster based on the number of cluster samples;

[0034] The combination of structural aggregation indices contained in the structural benchmark cluster is used as the benchmark tensor of fermentation data in the current metabolic stage.

[0035] In a preferred embodiment, in step S4, a baseline fluctuation range for fermentation data is preset. The baseline fluctuation range is used to define a reasonable fluctuation boundary for the structural deviation rate without metabolic imbalance.

[0036] The average value of the baseline tensor is taken as the baseline mean, and the absolute value of the difference between the aggregation index of each structure in the metabolic stage and the baseline mean is taken to obtain the structural deviation.

[0037] The result of standardization after taking the maximum value of each structural deviation is used as the structural deviation rate of the fermentation data in the metabolic stage.

[0038] In a preferred embodiment, in step S4, the structural deviation rate is compared with the baseline fluctuation range to analyze the metabolic state during the metabolic phase.

[0039] If the structural deviation rate falls within the baseline fluctuation range, the metabolic stage is determined to be in a normal metabolic transition state.

[0040] Otherwise, the metabolic phase is determined to be in a state of metabolic imbalance, and a sensory abnormality alarm signal is generated.

[0041] The technical effects and advantages of this invention are as follows:

[0042] This invention performs stage-sensing processing on the fermentation process to obtain the stage switching intervals of the synthetic pathway, and divides the fermentation process into different metabolic stages based on this. Within each metabolic stage, a correlation analysis window is constructed, and the Pearson correlation coefficient is used to quantify the synchronous correlation relationship of multi-source fermentation data. Simultaneously, a structural aggregation index of the fermentation data is constructed by combining time-lag information, and a baseline fluctuation range is set. Anomalies are defined as states that deviate from the normal correlation topology of the current metabolic stage. This enables effective differentiation between synthetic pathway stage switching and metabolic imbalance anomalies, improving the accuracy and stability of anomaly sensing without interfering with normal metabolic transitions, and providing reliable data support for the refined monitoring and process optimization of protocatechuic acid fermentation. Attached Figure Description

[0043] Figure 1 This is a flowchart illustrating the implementation of a time-series monitoring data anomaly identification and sensing method for a bio-fermentation process according to the present invention.

[0044] Figure 2 This is a schematic diagram illustrating the steps of a time-series monitoring data anomaly identification and sensing method for a bio-fermentation process according to the present invention. Detailed Implementation

[0045] 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.

[0046] This invention employs stage-sensing processing of the fermentation process to obtain the stage switching intervals of the synthetic pathway, and then divides the fermentation process into different metabolic stages. Within each metabolic stage, a correlation analysis window is constructed, and the Pearson correlation coefficient is used to quantify the synchronous correlation relationships of multi-source fermentation data. Simultaneously, a structural aggregation index for the fermentation data is constructed by combining time-lag information, and a baseline fluctuation range is set. Anomalies are defined as states that deviate from the normal correlation topology of the current metabolic stage, thereby effectively distinguishing between synthetic pathway stage switching and metabolic imbalance anomalies.

[0047] Example 1, such as Figures 1 to 2 As shown, a method for identifying and sensing anomalies in time-series monitoring data of a bio-fermentation process includes the following steps:

[0048] Step S1: During the fermentation of protocatechuic acid, monitor the fermentation data of protocatechuic acid and perform stage sensing processing to obtain the stage switching intervals in the fermentation process of protocatechuic acid.

[0049] Step S2: Divide the protocatechuic acid fermentation process into different metabolic stages based on the stage switching interval, set up a correlation analysis window in each metabolic stage, calculate the synchronous correlation coefficient of fermentation data in the correlation analysis window, and statistically analyze the time lag information of fermentation data.

[0050] Step S3: Generate the time lag strength of fermentation data based on the time lag information, analyze the structural aggregation index of fermentation data in combination with the synchronization correlation coefficient, perform cluster analysis on the structural aggregation index, and obtain the baseline tensor of fermentation data based on the clustering results;

[0051] Step S4: Set the baseline fluctuation range for fermentation data, use the baseline tensor to evaluate the structural deviation rate of fermentation data at each metabolic stage, analyze the metabolic state of each metabolic stage in combination with the baseline fluctuation range, and determine whether to output a sensing abnormality alarm signal based on the metabolic state.

[0052] The specific implementation is as follows:

[0053] In step S1, protocatechuic acid, as an important aromatic organic acid, has wide applications in multiple fields. In industrial production, it is synthesized by microbial fermentation using glucose, vanillic acid, or p-hydroxybenzoic acid as precursor substrates. The fermentation process involves a long metabolic pathway, significant stage switching, and high coupling of regulatory parameters. When switching from the cell growth stage to the protocatechuic acid synthesis stage, the rapid consumption of aromatic intermediate metabolites and the reconstruction of coenzyme balance will cause synchronous fluctuations in process parameters such as dissolved oxygen, respiratory quotient, and fermentation broth pH, ​​resulting in significant structural changes in the temporal morphology of fermentation data. If normal metabolic stage transitions are misjudged as metabolic imbalances, unnecessary process interventions will be triggered, disrupting the effective accumulation process of protocatechuic acid synthesis. Therefore, it is necessary to identify anomalies in the monitoring data that distinguishes and senses metabolic stage switching and abnormal states during protocatechuic acid fermentation.

[0054] During the protocatechuic acid fermentation process, fermentation data is continuously collected through an online monitoring device connected to the fermentation reactor.

[0055] Fermentation data included dissolved oxygen concentration, fermentation broth pH, ​​respiratory quotient, and fermentation substrate concentration. Dissolved oxygen concentration reflects the oxygen utilization status of microorganisms per unit time; fermentation broth pH characterizes the generation and accumulation of acidic metabolites in the fermentation system; the respiratory quotient characterizes changes in respiratory metabolism type during microbial carbon source metabolism; and fermentation substrate concentration reflects the consumption of precursor substrates during protocatechuic acid synthesis. Based on the collected fermentation substrate concentrations, the rate of change of fermentation substrate concentration over time was calculated to obtain the fermentation substrate consumption rate, which characterizes the changing trend of metabolic flux during protocatechuic acid fermentation.

[0056] After obtaining the fermentation data, stage-aware processing is performed on the fermentation data in the time series dimension. Specifically, a local sensing window with a length of several sampling periods is constructed with the current sampling time as the center. The first-order rate of change of dissolved oxygen concentration, fermentation broth pH value, and respiratory quotient within the local sensing window are calculated to obtain the rate of change of dissolved oxygen concentration, the rate of change of fermentation broth pH value, and the rate of change of respiratory quotient. The first-order rate of change is the ratio of the difference between the corresponding fermentation data at adjacent sampling times to the sampling time interval, which is used to characterize the direction and intensity of change of each fermentation data in the current time period.

[0057] Furthermore, within the local sensing window, the consistency of the signs of the rate of change in dissolved oxygen concentration, the rate of change in pH of the fermentation broth, and the rate of change in respiratory quotient is determined. When the proportion of the same signs of the rate of change in dissolved oxygen concentration, the rate of change in pH of the fermentation broth, and the rate of change in respiratory quotient within the local sensing window is not lower than a preset consistency ratio threshold, it is determined that the fermentation data within the local sensing window are consistent in the direction of change, reflecting the synergy of metabolic regulation direction during protocatechuic acid fermentation.

[0058] It should be noted that the preset consistency ratio threshold is a parameter used to quantify the degree of consistency in the direction of change of dissolved oxygen concentration, fermentation broth pH, ​​and respiratory quotient within a local sensing window. Its value is statistically set based on historical stable operating data during protocatechuic acid fermentation. Specifically, in multiple historical fermentation batches confirmed to be in the same metabolic stage and without metabolic imbalance, the proportion of sampling points with consistent signs of dissolved oxygen concentration change, fermentation broth pH change, and respiratory quotient change within each local sensing window is statistically analyzed, and the lower quantile of this proportion is used as the consistency ratio threshold.

[0059] Simultaneously, the mean fermentation substrate consumption rate is calculated within the same local sensing window, and the difference is calculated with the mean fermentation substrate consumption rate corresponding to the previous local sensing window. When the absolute value of the difference is greater than the preset metabolic change threshold, it is determined that the fermentation substrate consumption level has been significantly adjusted.

[0060] It should be noted that the preset metabolic change amplitude threshold is a quantitative parameter used to determine whether the change in the average fermentation substrate consumption rate between adjacent local sensing windows reaches the level of metabolic flux redistribution. Its value is set by analyzing the phased statistical characteristics of substrate consumption rate during protocatechuic acid fermentation. Specifically, in historical fermentation data, the absolute value distribution of the difference between the average substrate consumption rate of adjacent local sensing windows within the same metabolic stage and the distribution interval of the corresponding difference when a metabolic stage switch has been confirmed are calculated respectively, and the boundary value between the two distributions is selected as the metabolic change amplitude threshold.

[0061] When both the consistency of change direction and the significant change in fermentation substrate consumption rate are met simultaneously within the same local perception window, it is determined that the distribution of microbial metabolic flux during protocatechuic acid fermentation has changed within the corresponding time period.

[0062] The time ranges of multiple consecutive local sensing windows that all meet the above judgment conditions are merged to obtain the corresponding time intervals. The time intervals are marked as stage switching intervals. The stage switching intervals characterize the time range of the transition from one metabolic dominant state to another metabolic dominant state during the protocatechuic acid fermentation process, reflecting the reconstruction process of the microbial metabolic network during the adjustment of the protocatechuic acid synthesis pathway.

[0063] By identifying the transition intervals between metabolic stages, we can provide a basis for stage boundaries for subsequent correlation analysis and anomaly identification based on metabolic stages, thereby avoiding the mixed analysis of fermentation data under different metabolic states.

[0064] In step S2, the complete time series of the protocatechuic acid fermentation process is segmented using the stage switching interval as the time boundary, thereby dividing the protocatechuic acid fermentation process into three metabolic stages. Each metabolic stage corresponds to a period of relatively stable metabolic behavior before and after the stage switching interval, which is used to characterize the fermentation operation state in which the distribution of microbial metabolic flux in the protocatechuic acid synthesis process is basically consistent within the corresponding time period.

[0065] After completing the metabolic stage division, multiple correlation analysis windows are set within each metabolic stage. The correlation analysis window is a time interval that slides along the time axis within the corresponding metabolic stage according to a preset window length.

[0066] Within each correlation analysis window, the synchronous correlation coefficient of fermentation data during the protocatechuic acid fermentation process is calculated. The Pearson correlation coefficient is used as a unified quantitative calculation method for the synchronous correlation coefficient. Specifically, for any two types of fermentation data, including dissolved oxygen concentration, fermentation broth pH, ​​respiratory quotient, and fermentation substrate consumption rate, discrete time series of these data within the same correlation analysis window are extracted. Based on the discrete time series, the Pearson correlation coefficient of the two types of fermentation data within the correlation analysis window is calculated as the synchronous correlation coefficient. The synchronous correlation coefficient characterizes the degree of linear correlation between the numerical changes of different fermentation data in the current metabolic stage at the same time scale. Its magnitude reflects the strength of the synergistic changes between fermentation parameters, and the sign is used to characterize the consistency or oppositeity of the direction of change. This provides a quantitative basis for characterizing the synchronous coupling characteristics of the metabolic network during the protocatechuic acid fermentation process.

[0067] Simultaneously, the time lag information of fermentation data is statistically analyzed within the same correlation analysis window to quantify the dynamic response sequence relationship between different fermentation parameters. Specifically, within a preset time lag range, discrete time series corresponding to any two types of fermentation data are constructed with time lags, and the corresponding Pearson correlation coefficient is calculated for each time lag within the time lag range.

[0068] By traversing all time lags within the time lag range, the time lag that maximizes the absolute value of the Pearson correlation coefficient is selected as the time lag information of the two types of fermentation data within the current correlation analysis window. By statistically analyzing the time lag information of the pairwise combinations between dissolved oxygen concentration, fermentation broth pH, ​​respiratory quotient, and fermentation substrate consumption rate, the dynamic response structure of each fermentation data in the current metabolic stage is obtained. This provides an objective and quantitative time-series characteristic basis for subsequently constructing the correlation topology of fermentation data and identifying metabolic stage switching and abnormal states.

[0069] In step S3, within the correlation analysis window, the average value of each time lag is taken as the time lag intensity, which is used to reflect the overall level of the successive response intervals between different fermentation data.

[0070] The product of the time lag strength and the synchronization correlation coefficient after standardization is used as the time-series correlation strength factor.

[0071] The product of the temporal correlation strength factor and the preset structure regulation coefficient is used as the structure aggregation index of fermentation data, which reflects the structural synergy and stability of fermentation data in the current metabolic stage.

[0072] Within each metabolic stage, cluster analysis is performed on the structural aggregation indices formed by each correlation analysis window;

[0073] Specifically, each association analysis window is taken as a clustering sample, and the structural aggregation index corresponding to the association analysis window is taken as the sample feature of the sample;

[0074] Cluster analysis was performed on the structural aggregation index sequence formed by multiple consecutive association analysis windows within the same metabolic stage.

[0075] For any two clustered samples, calculate the absolute value of the difference in the structural aggregation index of the two clustered samples respectively, and use it as the aggregation index distance. If the aggregation index distance is less than the preset aggregation distance threshold, then the two clustered samples are determined to be structurally similar candidate windows.

[0076] If structurally similar candidate windows are adjacent on the time axis or within a preset range of adjacent windows, then the structurally similar candidate windows will be merged into the same cluster.

[0077] After clustering is completed, the number of cluster samples contained in each cluster during the current metabolic stage is counted, and the cluster corresponding to the maximum number of cluster samples is taken as the structural baseline cluster.

[0078] The correlation analysis windows contained in the structural baseline cluster are combined into a baseline window set. Based on the baseline window set, the structural aggregation indexes in each correlation analysis window are combined into a baseline tensor of the fermentation data in the current metabolic stage.

[0079] This step constructs a structural aggregation index based on time lag strength and synchronization correlation coefficient, and performs cluster analysis on the structural aggregation index sequence within the same metabolic stage. This enables the quantitative solidification of normal correlation structures in fermentation data, avoids misjudging normal metabolic transitions during the switching of synthetic pathway stages as abnormalities, effectively improves the ability of anomaly detection to distinguish changes in metabolic structure, and ensures the process continuity and end-product accumulation stability during the feeding and control strategy switching stages.

[0080] It should be noted that the standardization methods include, but are not limited to, standard linear transformation based on interval scaling, statistical Z-Score standardization, or normalization based on nonlinear mapping functions. The application methods of standardization will not be elaborated here. The preset structure adjustment coefficient can be set according to the fluctuation range of multiple parameters during fermentation or the structural distribution characteristics of historical operating data. The preset polymerization distance threshold can be set according to the historical fluctuation range of the structural polymerization index within the current metabolic stage, the stage stability requirements, or the engineering requirements for the sensitivity of the fermentation process to anomalies. The preset adjacent window range can be set according to the time length of the correlation analysis window, the dynamic change rate of the fermentation process, and the adjustment frequency of the control strategy.

[0081] In step S4, a baseline fluctuation range for fermentation data is preset. The baseline fluctuation range refers to the range of natural fluctuations in the structural deviation rate of fermentation data in the metabolic stage along the time axis under normal metabolic transition conditions. It is used to limit the reasonable fluctuation boundary of the structural deviation rate when metabolic imbalance does not occur.

[0082] The average value of the baseline tensor is taken as the baseline mean, and the absolute value of the difference between the aggregation index of each structure in the metabolic stage and the baseline mean is taken to obtain the structural deviation.

[0083] The result of standardization after taking the maximum value of each structural deviation is used as the structural deviation rate of fermentation data in the metabolic stage, which reflects the most severe deviation of the fermentation data associated structure from the normal structural baseline in this metabolic stage.

[0084] By comparing the structural deviation rate with the baseline fluctuation range, the metabolic state during the metabolic phase can be analyzed.

[0085] If the structural deviation rate falls within the baseline fluctuation range, the metabolic stage is determined to be in a normal metabolic transition state.

[0086] Otherwise, the metabolic phase is determined to be in a state of metabolic imbalance, and a sensory abnormality alarm signal is generated.

[0087] When the structural deviation rate falls within the baseline fluctuation range, it indicates that although there is structural fluctuation in this metabolic stage, the fluctuation is a normal response within the stage caused by ventilation, feeding or metabolic flux adjustment, and the metabolic stage is in a normal metabolic transition state.

[0088] When the structural deviation rate exceeds or falls below the baseline fluctuation range, it indicates that a structural abnormality has occurred in this metabolic stage that cannot be explained by the normal regulatory response. This indicates that the metabolic stage is in a state of metabolic imbalance and generates a sensing abnormality alarm signal.

[0089] It should be noted that the baseline fluctuation range of the preset fermentation data can be set based on the statistical distribution of structural deviation rate, duration of metabolic phase, sampling cycle, and control response characteristics of the fermentation process in historical normal operating batches.

[0090] This step introduces a benchmark fluctuation range to constrain the structural deviation rate, avoiding misjudging normal structural fluctuations caused by stage switching or control response as abnormalities. While preserving the natural fluctuation characteristics of the time axis, it effectively distinguishes between metabolic transitions and metabolic imbalances, improves the accuracy of anomaly detection, and ensures the process continuity and stable accumulation of final products during the feeding and ventilation strategy switching stages.

[0091] Finally, it should be noted that in this paper, relational terms such as first and second are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations.

[0092] 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 a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0093] In this document, the singular forms “a,” “an,” and “the” may also include the plural forms unless the context clearly indicates otherwise. It should also be understood that terms such as “comprising / including” or “having” specify the presence of the stated features, integrals, steps, operations, components, parts, or combinations thereof, but do not preclude the possibility of the presence or addition of one or more other features, integrals, steps, operations, components, parts, or combinations thereof. Meanwhile, the term “and / or” as used in this specification includes any and all combinations of the associated listed items.

[0094] The various embodiments in this specification are described in a progressive manner. Each embodiment focuses on the differences from other embodiments. The various embodiments can be combined as needed, and the same or similar parts can be referred to each other.

[0095] The above description of the disclosed embodiments will enable those skilled in the art to make or use various modifications to these embodiments. It will be readily apparent to those skilled in the art that the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for identifying and sensing anomalies in time-series monitoring data of a bio-fermentation process, characterized in that: Includes the following steps: Step S1: During the fermentation of protocatechuic acid, monitor the fermentation data of protocatechuic acid and perform stage sensing processing to obtain the stage switching intervals in the fermentation process of protocatechuic acid. Step S2: Divide the protocatechuic acid fermentation process into different metabolic stages based on the stage switching interval, set up a correlation analysis window in each metabolic stage, calculate the synchronous correlation coefficient of fermentation data in the correlation analysis window, and statistically analyze the time lag information of fermentation data. Step S3: Generate the time lag strength of fermentation data based on the time lag information, analyze the structural aggregation index of fermentation data in combination with the synchronization correlation coefficient, perform cluster analysis on the structural aggregation index, and obtain the baseline tensor of fermentation data based on the clustering results; Step S4: Set the baseline fluctuation range for fermentation data, use the baseline tensor to evaluate the structural deviation rate of fermentation data at each metabolic stage, analyze the metabolic state of each metabolic stage in combination with the baseline fluctuation range, and determine whether to output a sensing abnormality alarm signal based on the metabolic state.

2. The method for identifying and sensing anomalies in time-series monitoring data of a bio-fermentation process according to claim 1, characterized in that: In step S1, during the protocatechuic acid fermentation process, fermentation data during the protocatechuic acid fermentation process is continuously collected through an online monitoring device connected to the fermentation reactor. Fermentation data include dissolved oxygen concentration, fermentation broth pH, ​​respiratory quotient, and fermentation substrate concentration. Based on the fermentation substrate concentration, the rate of change of fermentation substrate concentration over time is calculated to obtain the fermentation substrate consumption rate.

3. The method for identifying and sensing anomalies in time-series monitoring data of a bio-fermentation process according to claim 2, characterized in that: In step S1, after obtaining the fermentation data, the fermentation data undergoes stage-aware processing in the time series dimension: Centered on the current sampling time, a local sensing window with a length of several sampling periods is constructed. The first-order rate of change of dissolved oxygen concentration, fermentation broth pH value and respiratory quotient within the local sensing window are calculated to obtain the rate of change of dissolved oxygen concentration, the rate of change of fermentation broth pH value and the rate of change of respiratory quotient. When the rate of change of dissolved oxygen concentration, the rate of change of pH value of fermentation broth, and the rate of change of respiratory quotient meet the condition that the proportion of the same sign is not lower than the preset consistency ratio threshold within the local sensing window, it is determined that the fermentation data within the local sensing window are consistent in the direction of change.

4. The method for anomaly identification and sensing of time-series monitoring data in a bio-fermentation process according to claim 3, characterized in that: In step S1, the average fermentation substrate consumption rate is calculated within the same local sensing window, and the difference between the average fermentation substrate consumption rate and the average fermentation substrate consumption rate corresponding to the previous local sensing window is calculated. When the absolute value of the difference is greater than the preset threshold for metabolic change, it is determined that the fermentation substrate consumption level has been significantly adjusted. When both the consistency of change direction and the significant change in fermentation substrate consumption rate are met within the same local perception window, it is determined that the distribution of microbial metabolic flux during protocatechuic acid fermentation has changed within the corresponding time period. The time ranges of multiple consecutive local perception windows that meet the judgment conditions are merged to obtain the corresponding time interval, and the time interval is marked as the stage switching interval.

5. The method for identifying and sensing anomalies in time-series monitoring data of a bio-fermentation process according to claim 1, characterized in that: In step S2, the complete time series of the protocatechuic acid fermentation process is segmented using the stage switching interval as the time boundary, and the protocatechuic acid fermentation process is divided into three metabolic stages. Multiple correlation analysis windows are set within each metabolic stage. The correlation analysis window is a time interval that slides along the time axis within the corresponding metabolic stage according to a preset window length. Within each correlation analysis window, the Pearson correlation coefficient calculation method is used to calculate the synchronous correlation coefficient of the fermentation data during the protocatechuic acid fermentation process; Statistical analysis of fermentation data time lag information is performed within the same correlation analysis window; Within a preset time lag range, the Pearson coefficients of discrete time series of different fermentation data under different lag conditions are calculated, and the time lag at which the Pearson coefficient reaches its maximum value is taken as the time lag information.

6. The method for identifying and sensing anomalies in time-series monitoring data of a bio-fermentation process according to claim 5, characterized in that: In step S3, within the correlation analysis window, the average value of each time lag quantity is taken as the time lag strength. The product of the time lag strength and the synchronization correlation coefficient after standardization is used as the time-series correlation strength factor. The product of the temporal correlation strength factor and the preset structure adjustment coefficient is used as the structure aggregation index of the fermentation data.

7. The method for anomaly identification and sensing of time-series monitoring data in a bio-fermentation process according to claim 1, characterized in that: In step S3, within each metabolic stage, cluster analysis is performed on the structural aggregation indices formed by each correlation analysis window; Each association analysis window is used as a clustering sample, and the structural aggregation index corresponding to the association analysis window is used as the sample feature of the sample. The aggregation index distance is calculated based on the structural aggregation index of any two clustered samples. If the aggregation index distance is less than the preset aggregation distance threshold, the two clustered samples are determined to be structurally similar candidate windows. Clusters are obtained by aggregating candidate windows based on their temporal dimension; Count the number of cluster samples contained in each cluster, and select the structural benchmark cluster based on the number of cluster samples; The combination of structural aggregation indices contained in the structural benchmark cluster is used as the benchmark tensor of fermentation data in the current metabolic stage.

8. The method for identifying and sensing anomalies in time-series monitoring data of a bio-fermentation process according to claim 7, characterized in that: In step S4, a baseline fluctuation range for fermentation data is preset. The baseline fluctuation range is used to define a reasonable fluctuation boundary for the structural deviation rate without metabolic imbalance. The average value of the baseline tensor is taken as the baseline mean, and the absolute value of the difference between the aggregation index of each structure in the metabolic stage and the baseline mean is taken to obtain the structural deviation. The result of standardization after taking the maximum value of each structural deviation is used as the structural deviation rate of the fermentation data in the metabolic stage.

9. The method for identifying and sensing anomalies in time-series monitoring data of a bio-fermentation process according to claim 1, characterized in that: In step S4, the structural deviation rate is compared with the baseline fluctuation range to analyze the metabolic state during the metabolic phase. If the structural deviation rate falls within the baseline fluctuation range, the metabolic stage is determined to be in a normal metabolic transition state. Otherwise, the metabolic phase is determined to be in a state of metabolic imbalance, and a sensory abnormality alarm signal is generated.

Images