CN-122024935-A - Online intelligent evaluation modeling method for activity parameters of fermentation strain

CN122024935ACN 122024935 ACN122024935 ACN 122024935ACN-122024935-A

Abstract

The invention discloses an online intelligent evaluation modeling method for activity parameters of fermentation strains, in particular to the field of fermentation process data processing, which is used for solving the problem that the online identification is difficult because the quantity representation and the actual activity state are easily misplaced in the fermentation process; the method comprises the steps of obtaining viable bacteria concentration data or thallus concentration estimated value, acid-base data, feed supplement data, temperature data and ventilation data of a current fermentation batch, constructing a quantity representation sequence and a process response sequence, combining a historical normal batch reference interval to identify a dislocation window, calculating a reference conversion stable quantity time quantity, a maintenance load conversion time quantity and a surface activity mismatch coefficient, generating a state deviation fragment, a stage activity track and a strain activity parameter evaluation result, outputting an activity state label and a production treatment prompt according to the state deviation fragment, and realizing continuous judgment and management expression of internal and external activity deviations in a fermentation stage, so that the on-line evaluation result and a fermentation batch treatment link form a direct corresponding relation.

Inventors

DENG GANG
YANG RU
LAI JIABAO
YAO ZHIKUN
SUN YAN
ZHANG JIAFAN
LIU WEI
Kang Dongtao

Assignees

汉中天然谷生物科技股份有限公司

Dates

Publication Date: 20260512
Application Date: 20260410

Claims (10)

1. A fermentation strain activity parameter online intelligent evaluation modeling method is characterized by comprising the following steps: s1, acquiring various data of a current fermentation batch, rearranging according to a uniform time index to form a quantity representation sequence and a process response sequence, and dividing the quantity representation sequence and the process response sequence into continuous time windows respectively corresponding to the quantity representation sequence and the process response sequence; s2, reading continuous time windows one by one, mapping historical normal batch reference intervals of the fermentation stage to a quantity representation sequence and a process response sequence, identifying dislocation windows of which the process response sequence falls outside the reference interval and the quantity representation sequence falls inside the reference interval, generating reference conversion stable quantity time, maintenance load conversion time and surface activity mismatch coefficients, and connecting to obtain a state deviation segment; S3, sequencing the state deviation fragments according to a unified time index, calling a reference fragment corresponding to a fermentation stage, aligning the quantity representation sequence, the process response sequence and the surface activity mismatch coefficient in the state deviation fragments time by time, determining a deviation starting point, a deviation continuous range and a deviation ending point, forming a stage activity track and determining a strain activity parameter evaluation result; s4, reading the strain activity parameter evaluation result and the stage activity track, extracting track connection positions between adjacent fermentation stages and deviation distribution results inside each fermentation stage, generating an activity state change result, and mapping the strain activity parameter evaluation result and the activity state change result into an activity state label and a production treatment prompt.
2. The method for online intelligent evaluation modeling of activity parameters of a fermentation strain according to claim 1, wherein the step S1 comprises: Firstly, determining a unique quantity data source between the viable bacteria concentration data and the bacterial concentration estimated value, deleting a time stamp missing item and a measured value missing item for the viable bacteria concentration data or the bacterial concentration estimated value, acid-base data, feed supplement data, temperature data and ventilation data, retaining a last written item when the same time stamp is repeated, and then generating a unified time index according to a common effective observation interval and a fixed sampling interval.
3. The method for online intelligent evaluation modeling of activity parameters of a fermentation broth according to claim 2, wherein step S1 further comprises: And performing piecewise linear interpolation on the viable bacteria concentration data or the bacterial concentration estimated value, the acid-base data and the temperature data along the unified time index, performing zero-order maintenance on the feed supplement data and the ventilation data, forming continuous index sections after eliminating the observation fracture sections, and generating continuous time windows from the inside of each continuous index section according to the preset window length and the adjacent window handover rule.
4. The method for online intelligent evaluation modeling of activity parameters of a fermentation broth according to claim 3, wherein step S2 comprises: Comparing each continuous time window with each fermentation stage time range one by one, determining a stage mapping window according to the fermentation stage with the largest overlapping duration, calling the historical normal batch processed by the step S1, respectively generating a number of reference intervals and process reference intervals of acid base, feed supplement, temperature and ventilation on a unified time point covered by the stage mapping window, and mapping the reference intervals to a number characterization sequence and a process response sequence.
5. The method for online intelligent evaluation modeling of activity parameters of a fermentation broth of claim 4, wherein step S2 further comprises: Identifying a number of bounded fragments with duration reaching a number of bounded duration thresholds in a phase mapping window, identifying a process deviation fragment with duration reaching a process deviation duration threshold, determining a part with duration reaching a misplacement overlapping duration threshold in an overlapping time period of the bounded fragments and the process deviation fragment as a misplacement window, and determining the number of the misplacement window in the bounded time period as a steady fragment.
6. The method for online intelligent evaluation modeling of activity parameters of a fermentation broth of claim 5, wherein step S2 further comprises: Obtaining a stable load bearing time product by accumulating the product of the quantity value and the fixed sampling interval point by point along the stable segment, and converting the stable load bearing time product into a reference converted stable time by taking the average result of median values of the historical normal batch quantity at the unified time point corresponding to the stable segment as a reference quantity standard; meanwhile, process channels which fall outside the process reference interval in the stable quantity segment are counted point by point and accumulated to obtain the maintenance load conversion time quantity, and then the table activity mismatch coefficient is output by the two-channel interaction learning model.
7. The method for online intelligent evaluation modeling of activity parameters of a fermentation broth of claim 6, wherein step S3 comprises: the method comprises the steps of firstly sequencing state deviation fragments according to a unified time index, spreading the surface activity mismatch coefficients of effective dislocation windows forming the state deviation fragments into a time spreading surface activity mismatch sequence, then calling reference fragments covering the same unified time point set from a historical normal batch, wherein the reference fragments comprise a quantity reference upper boundary and a quantity reference lower boundary, a process reference upper boundary and a surface activity mismatch reference upper boundary, and aligning the quantity representation sequence, a process response sequence and the time spreading surface activity mismatch sequence time by time on the same unified time point.
8. The method for online intelligent evaluation modeling of activity parameters of a fermentation broth of claim 7, wherein step S3 further comprises: and determining continuous track deviation sections with the duration reaching a deviation confirmation duration threshold as initial candidate deviation sections according to the quantity deviation, the process deviation and the coefficient deviation generation track deviation mark, and determining a deviation starting point, a deviation duration range and a deviation ending point by using the parallel candidate deviation sections with the largest coefficient accumulation quantity after parallel adjacent initial candidate deviation sections with intervals not exceeding a deviation tolerance threshold.
9. The method for online intelligent evaluation modeling of activity parameters of a fermentation broth of claim 8, wherein step S3 further comprises: and simultaneously, respectively extracting the deviation coverage time ratio, the longest continuous deviation time quantity and the phase peak value coefficient in each fermentation stage, and forming a strain activity parameter evaluation result by the deviation coverage time ratio, the longest continuous deviation time quantity and the phase peak value coefficient.
10. The method for online intelligent evaluation modeling of activity parameters of a fermentation broth according to claim 9, wherein step S4 comprises: And determining an activity state change result according to the track connection position and the connection judgment threshold value between adjacent fermentation stages, determining an activity stage code of each fermentation stage according to the coverage judgment threshold value, the continuous judgment threshold value and the peak judgment threshold value, extracting a first stage serial number and a main migration type from the activity state change result, forming a ternary index with the activity stage code of the last stage, and uniquely mapping the ternary index into an activity state label and a production treatment prompt according to a preset rule set.

Description

Online intelligent evaluation modeling method for activity parameters of fermentation strain Technical Field The invention relates to the field of fermentation process data processing, in particular to an online intelligent evaluation modeling method for activity parameters of fermentation strains. Background In the biological fermentation production process, strain activity is a key factor affecting substrate utilization, metabolic conversion and process stability. The prior art generally indirectly judges fermentation status by detecting viable bacteria concentration on line or establishing a soft measurement model based on process parameters. For example, the prior art of the device for real-time online detection of the concentration of viable bacteria in fermentation broth (application number: CN 200420005114) focuses on real-time online detection of the concentration of viable bacteria in fermentation broth, and the prior art of the method for online soft measurement of the concentration of bacteria in glutamic acid fermentation process (application number: CN 201310460718.4) carries out online estimation of the concentration of bacteria by collecting real-time process data in the fermentation process and establishing a soft measurement model. The scheme shows that the prior art can acquire continuous information around the fermentation process and realize online characterization, but mainly reflects the change of the number, growth state or external process of the thalli, and an effective and direct online evaluation system is not formed on the internal parameter of the strain activity which can more embody the real metabolic capability. However, in the actual fermentation process, the strain activity does not necessarily change in synchronization with the viable cell concentration or the cell concentration. The reason is that when the bacterial cells are affected by changes in culture conditions, fluctuations in mass transfer conditions, switching of nutrient supply, and changes in metabolic load, the bacterial cells are often first shown as changes in metabolic capacity, environmental response capacity, and sustained conversion capacity, and such changes are not immediately shown as obvious changes in the number of bacterial cells, and are difficult to accurately reveal only by conventional process variables. As a result, although the prior art can perform online characterization on the growth trend or the local process state of the thalli, the problems of evaluation lag, judgment deviation and control basis distortion are easy to occur when the activity of the strain is subjected to recessive attenuation, stage turning or state mismatch. Therefore, how to establish a method which can distinguish thallus quantity characterization from strain activity characterization and is applicable to online intelligent evaluation modeling of fermentation process is still a key technical problem to be solved in the prior art. In order to solve the above problems, a technical solution is now provided. Disclosure of Invention In order to overcome the defects of the prior art, the embodiment of the invention provides an online intelligent evaluation modeling method for activity parameters of fermentation strains, which comprises the steps of obtaining active bacteria concentration data or thallus concentration estimated value, acid-base data, feed supplement data, temperature data and ventilation data of a current fermentation batch, constructing a quantity representation sequence and a process response sequence, combining a historical normal batch reference interval to identify a dislocation window, calculating a reference conversion stable quantity time, a maintenance load conversion time and a surface activity mismatch coefficient, generating state deviation fragments, a stage activity track and a strain activity parameter evaluation result, outputting an activity state label and a production treatment prompt according to the quantity, realizing continuous judgment and management expression of internal and external activity deviations in a fermentation stage, and enabling the online evaluation result and a fermentation batch treatment link to form a direct corresponding relation so as to solve the problems in the background technology. In order to achieve the above purpose, the present invention provides the following technical solutions: s1, acquiring various data of a current fermentation batch, rearranging according to a uniform time index to form a quantity representation sequence and a process response sequence, and dividing the quantity representation sequence and the process response sequence into continuous time windows respectively corresponding to the quantity representation sequence and the process response sequence; s2, reading continuous time windows one by one, mapping historical normal batch reference intervals of the fermentation stage to a quantity representation sequence and a process response s