Bio-based fabric production temperature and humidity intelligent control system

Abstract

The present application relates to the technical field of textile biofabrication and industrial process automation control, in particular to a bio-based fabric production temperature and humidity intelligent regulation and control system, which comprises a data acquisition unit for acquiring temperature and humidity sensor values in a fermentation and solidification workshop in real time, reading current heating power and fan speed instruction of an actuator; a reference state reconstruction unit for receiving an input state vector and generating an interference-free reference vector; a disturbance simulation unit for generating a theoretical damaged state vector representing a specific fault mode; a double-track differential unit for generating a real difference feature vector and a theoretical difference feature vector; a coupling regulation and control unit for calculating the similarity value of the real difference feature vector and the theoretical difference feature vector in the number domain space; the present application effectively eliminates non-ideal factor interference, ensures the physical interpretability and numerical traceability of the reference data.

Description

Technical Field

[0001] This invention relates to the field of textile biomanufacturing and industrial process automation control technology, specifically to an intelligent temperature and humidity control system for bio-based fabric production. Background Technology

[0002] With the continuous development of bio-based fabric production technology, the requirements for precise control of temperature and humidity environment in the fermentation and solidification process have significantly increased. This production environment involves complex biochemical reaction dynamics and multi-physics field coupling, which brings many challenges, especially in the production of high-value fabrics with large batch differences.

[0003] Currently, environmental sensor arrays and industrial communication interfaces are typically used to collect temperature and humidity data within workshops, and traditional threshold alarms or black-box AI models are used for monitoring and control. Systems usually make logical judgments based on real-time sensor readings and actuator states, comparing them to preset standard ranges, or make predictions based on models trained with large amounts of historical data, to maintain environmental parameters within the workshop. However, traditional monitoring and control methods rely on the absolute accuracy of sensor readings and the support of massive amounts of historical fault data, which has significant limitations in practical applications. In high-humidity fermentation environments, humidity-sensitive elements are prone to aging and drift, and the weak bio-heat generated in the early stages of fabric mold growth is easily masked by random environmental noise. Existing technologies struggle to accurately distinguish sensor drift, bio-heating anomalies, and real environmental fluctuations from noisy signals in the absence of prior fault data, easily leading to missed detections due to the system's inability to identify weak bio-anomalies, or incorrect compensation actions due to misjudgment of sensor drift. Therefore, how to effectively isolate environmental noise and accurately distinguish between equipment faults and bio-anomalies without relying on massive amounts of historical data has become a pressing problem in this field. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention provides an intelligent temperature and humidity control system for bio-based fabric production. Specifically, the technical solution of this invention includes:

[0005] A smart temperature and humidity control system for bio-based fabric production includes a data acquisition unit, a reference state reconstruction unit, a disturbance simulation unit, a dual-track differential unit, and a coupling control unit.

[0006] The data acquisition unit is equipped with an industrial communication interface and an environmental sensor group, which is used to collect the temperature and humidity sensor values ​​in the fermentation and curing workshop in real time, read the current heating power and fan speed command of the actuator, and call the pre-stored fabric fermentation dynamic parameters. The above data is combined into an input state vector and sent to the reference state reconstruction unit.

[0007] The reference state reconstruction unit is used to receive the input state vector, perform calculations based on the thermodynamic energy conservation model, obtain the theoretical temperature and humidity after eliminating the influence of environmental noise and equipment failure, generate an interference-free reference vector, and synchronously send the interference-free reference vector to the disturbance simulation unit and the dual-track differential unit.

[0008] The disturbance simulation unit is connected to a database storing fault parameterization operators. It is used to superimpose a preset disturbance factor onto a non-disturbed reference vector to generate a theoretical damaged state vector characterizing a specific fault mode, and send the theoretical damaged state vector to the dual-track differential unit.

[0009] The dual-track differential unit is used to calculate the vector difference between the temperature and humidity sensor values ​​and the interference-free reference vector to generate a real difference feature vector, and at the same time calculate the vector difference between the theoretical damaged state vector and the interference-free reference vector to generate a theoretical difference feature vector.

[0010] The coupling control unit is used to calculate the similarity value between the actual difference feature vector and the theoretical difference feature vector in the number domain space, and to perform the following logical judgment: if the similarity value is greater than a preset judgment threshold, a specific event judgment signal is generated, and a feedforward compensation control command is triggered and sent to the actuator to compensate for the temperature and humidity deviation; if the similarity value is less than or equal to the preset judgment threshold, an environmental noise signal is generated, and the current instruction of the actuator is locked to maintain the current temperature and humidity control strategy unchanged and not respond to the fluctuations in the actual difference feature vector.

[0011] Preferably, the specific steps for the reference state reconstruction unit to perform reference deduction analysis include:

[0012] Extract the actuator's heating power data and the fan's air volume data from the input state vector, and calculate the theoretical bio-self-heating power and theoretical moisture evaporation and dissipation at the current moment based on the fabric fermentation kinetic parameters.

[0013] Perform energy conservation calculations: take the heating power data and theoretical biological self-heating power as heat inputs, subtract the heat loss term calculated from the fan air volume data and the thermal conductivity of the workshop enclosure structure, and obtain the net heat flux;

[0014] Perform mass conservation calculations: Calculate the absolute humidity change rate in the workshop based on the difference between the theoretical moisture evaporation and the fan exhaust rate;

[0015] Based on the net heat flux and the absolute humidity change rate, the theoretical temperature and humidity values ​​at the current moment are calculated by integration, and these theoretical temperature and humidity values ​​are used to construct an interference-free reference vector.

[0016] Preferably, the calculation steps for theoretical biological self-heating power include:

[0017] Read the standard strain growth exothermic curve data and enzyme activity temperature correction coefficient from the fabric fermentation kinetic parameters;

[0018] Obtain the start time of the fermentation process, determine the corresponding time index in the standard strain growth exothermic curve data based on the current system clock, and extract the corresponding basal metabolic calorific value.

[0019] Collect real-time temperature data of the workshop, use enzyme activity temperature correction coefficient to gain or attenuate the basal metabolic calorific value, and output the corrected value as the theoretical biological self-heating power.

[0020] Preferably, the fault simulation injection analysis process is as follows: obtain parameterized operators for various fault types preset in the expert knowledge base, including sensor drift faults and biological contamination anomaly faults; superimpose the parameterized operators onto the values ​​of the interference-free reference vector to simulate the data changes when different fault types occur; generate a theoretical damaged state vector corresponding to each fault type, which represents the theoretical data sequence under the condition that a specific fault occurs and there is no environmental random noise interference.

[0021] The preferred definition and calculation process of the parameterized operator are as follows:

[0022] To address sensor drift faults, a drift interference operator is invoked. This operator defines the drift deviation value as a nonlinear function of the integral value of historical humidity data over time, and calculates the amount of humidity deviation that accumulates over time and superimposes it onto an interference-free reference vector.

[0023] For abnormal faults caused by biological contamination, the heat source interference operator is invoked. This operator defines the abnormal heat source value as an exponential function of time, calculates the additional heat value that grows exponentially, and converts it into the corresponding temperature increment, which is then superimposed on the interference-free reference vector.

[0024] Preferably, the specific steps for the dual-track differential unit to perform the first path difference calculation and the second path difference calculation are as follows:

[0025] First path calculation: The temperature and humidity values ​​collected in real time by the sensor are used to construct a measured state vector. The measured state vector is subtracted from the interference-free reference vector to obtain a real difference feature vector containing real abnormal components and random noise components.

[0026] Second path calculation: Subtract the undisturbed baseline vector from the theoretical damaged state vector to obtain the theoretical difference feature vector containing only the fault mode feature components;

[0027] After aligning the actual difference feature vector with the theoretical difference feature vector in time series, the data is sent to the coupling control unit.

[0028] Preferably, the similarity comparison decision process performed by the coupling control unit is as follows:

[0029] Map the received real-world difference feature vectors and theoretical difference feature vectors to the same feature vector space;

[0030] The optimal matching path distance between two vector sequences on the time axis is calculated using the dynamic time warping algorithm, or the cosine similarity algorithm is used to calculate the cosine value of the angle between two vectors in multidimensional space, thus obtaining a normalized similarity value.

[0031] The similarity value is compared with a preset judgment threshold in the memory to output the specific event judgment signal or environmental noise signal.

[0032] Preferably, the similarity value is calculated using a dynamic time warping algorithm or a cosine similarity algorithm to distinguish the random noise components contained in the real-world difference features and extract fault features that are consistent with the theoretical difference features in terms of change trend.

[0033] Compared with the prior art, the present invention has the following beneficial effects:

[0034] 1. This system generates an interference-free reference vector through a reference state reconstruction unit and uses physical equations to deduce ideal temperature and humidity. This method eliminates the excessive reliance on sensor measurements in traditional strategies, effectively removes interference from non-ideal factors, ensures the physical interpretability and numerical traceability of the reference data, and provides standard reference coordinates for accurately extracting weak biological anomaly signals from noisy environments.

[0035] 2. This system adopts knowledge-driven reverse fault injection technology, which actively simulates fault modes such as sensor drift and biological contamination through preset mathematical operators. This design eliminates the dependence on massive historical labeled data and can realize cold start monitoring of the system without training. It is particularly suitable for high-value production scenarios with large batch differences that lack prior fault data and can sensitively identify potential risks that conform to specific mathematical laws.

[0036] 3. This system constructs two parallel computational paths, one for reality and one for theory, and performs differential operations using an interference-free reference vector as the medium. This dual-track architecture can effectively separate real fault characteristics from environmental random noise and unmodeled residuals, significantly reducing the signal-to-noise ratio requirements for signal processing. This enables accurate differentiation between reading drift caused by sensor aging and real heat changes generated by biological fermentation, avoiding misjudgments.

[0037] 4. This system uses a similarity algorithm to compare waveform trends in the numerical space, rather than simply relying on numerical magnitude for over-limit alarms. This mechanism achieves mathematical immunity to random noise that follows a Gaussian distribution. Even if the sensor data fluctuates slightly, it can be identified as long as its trend conforms to the preset fault model. Conversely, if the fluctuation does not conform to the fault pattern, it is regarded as noise filtering, thereby ensuring the accuracy and robustness of the control decision. Attached Figure Description

[0038] The present invention will be further explained below with reference to the accompanying drawings and embodiments:

[0039] Figure 1 This is a structural diagram of the system of the present invention. Detailed Implementation

[0040] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments. Example 1:

[0041] Please see Figure 1 A smart temperature and humidity control system for bio-based fabric production includes a data acquisition unit, a reference state reconstruction unit, a disturbance simulation unit, a dual-track differential unit, and a coupling control unit.

[0042] The data acquisition unit is equipped with an industrial communication interface and an environmental sensor group. It is used to collect the temperature and humidity sensor values ​​in the fermentation and curing workshop in real time, read the current heating power and fan speed commands of the actuator, and call the pre-stored fabric fermentation dynamic parameters. The above data is combined into an input state vector and sent to the reference state reconstruction unit.

[0043] The reference state reconstruction unit is used to receive the input state vector, perform calculations based on the thermodynamic energy conservation model, obtain the theoretical temperature and humidity after eliminating the influence of environmental noise and equipment failure, generate an interference-free reference vector, and synchronously send the interference-free reference vector to the disturbance simulation unit and the dual-track differential unit.

[0044] The disturbance simulation unit is connected to a database that stores fault parameterization operators. It is used to superimpose preset disturbance factors onto the non-disturbance reference vector to generate a theoretical damaged state vector that characterizes a specific fault mode, and then send the theoretical damaged state vector to the dual-track differential unit.

[0045] The dual-track differential unit is used to calculate the vector difference between the temperature and humidity sensor values ​​and the interference-free reference vector to generate the actual difference feature vector. At the same time, it calculates the vector difference between the theoretical damaged state vector and the interference-free reference vector to generate the theoretical difference feature vector.

[0046] The coupling control unit is used to calculate the similarity value between the actual difference feature vector and the theoretical difference feature vector in the number domain space, and to perform the following logical judgment:

[0047] If the similarity value is greater than the preset judgment threshold, a specific event judgment signal is generated, and a feedforward compensation control command is sent to the actuator to compensate for the temperature and humidity deviation.

[0048] If the similarity value is less than or equal to the preset judgment threshold, an environmental noise signal is generated, and the current instruction of the actuator is locked to maintain the current temperature and humidity control strategy unchanged, without responding to the fluctuations in the real difference feature vector.

[0049] This embodiment provides an intelligent temperature and humidity control system for the production of bio-based fabrics. To address the pain point of existing technologies that cannot distinguish between biological self-heating, sensor drift, and real-world environmental fluctuations, this system adopts a virtual-real combined architecture with dual-track verification.

[0050] The data acquisition unit is the sensing end of the system, and its purpose is to construct a raw dataset reflecting the current physical state of the workshop. In this embodiment, the unit is equipped with an industrial communication interface supporting the Modbus / TCP protocol and a high-precision environmental sensor array. The data acquisition unit interacts with data via the Modbus / TCP protocol, for example, real-time sensor data and actuator status, such as heating power. The speed command is obtained by reading the holding register, and the data type is floating point.

[0051] Meanwhile, the pre-stored fabric fermentation kinetics parameter data structure includes: batch ID, standard strain name, and optimal growth temperature. Enzyme activity broadband coefficient Evaporation coefficient And a time index and basal metabolic rate. A two-dimensional lookup table; specifically, this unit collects the following data in real time: input state vector This includes multiple temperature points within the fermentation and solidification workshop. and humidity Sensor readings, and real-time ambient temperature outside the workshop. and ambient humidity Actuator status: Read the current heating power of the heating resistor. and the speed command of the circulating fan, corresponding to the air volume. Fabric fermentation kinetic parameters: retrieve the pre-stored biological characteristic data of the current production batch of fabric from the database; after aligning the above heterogeneous data with timestamps, this unit packages it into an input state vector and sends it to the reference state reconstruction unit;

[0052] The reference state reconstruction unit is the system's zero-point calibration center, its purpose being to generate an idealized reference that completely eliminates non-ideal factors such as noise and faults. In this embodiment, this unit receives the input state vector and performs calculations based on a thermodynamic energy conservation model. It extrapolates the theoretical temperature and humidity that should be present at the current moment, assuming no external interference, perfect equipment operation, and fabric growth conforming to an ideal biological curve, thus generating an interference-free reference vector. The vector was sent synchronously to the next two units.

[0053] The disturbance simulation unit is the system's backpropagation engine, connected to a database storing fault parameterized operators. Its working logic is proactive trial and error: instead of passively waiting for a fault to occur, it artificially superimposes preset disturbance factors, such as mathematical terms simulating sensor moisture drift, onto a pure, disturbance-free reference vector; thereby generating a theoretical damaged state vector. This vector represents the condition if the first... The standard damaged form that the data should exhibit when a specific fault, such as biological contamination, actually occurs and there is no random noise interference;

[0054] The dual-track differential unit is the system's feature extractor, used to separate signal from noise. It executes two parallel differential calculation paths: the real path, which calculates the difference between the sensor's measured value and the reference value, including the true signal and random noise; and the theoretical path, which calculates the difference between the theoretical damage value and the reference value, including only the true signal features. It generates real difference feature vectors and theoretical difference feature vectors respectively.

[0055] The coupling control unit is the system's decision-making brain, responsible for executing the final logical judgment. It calculates the similarity between the actual difference and the theoretical difference in the numerical space. Logical judgment one: for a specific event, if the similarity value is greater than a preset judgment threshold... This indicates that the actual fluctuation pattern closely matches the theoretical pattern of a certain known fault. At this point, the system generates a specific event judgment signal and triggers a feedforward compensation control command, such as significantly increasing the cooling air volume to compensate for temperature and humidity deviations in the case of biological overheating. Logical judgment two: environmental noise. If the similarity value is less than or equal to the judgment threshold, it indicates that the current temperature and humidity fluctuations are chaotic and do not conform to any physical / biological fault patterns, belonging to random noise. The system generates an environmental noise signal, locks the actuator command, maintains the current strategy, and does not respond to fluctuations in sensor data.

[0056] Through the above settings, this system constructs a closed loop of positive and negative dual verification; it can fundamentally solve the problem of mis-control caused by sensor drift, and can accurately extract weak biological abnormal signals from noisy environmental noise, such as abnormal heat generation in the early stage of fabric mold; this first-principles-based control strategy, compared with the traditional black-box AI model, can achieve cold start monitoring without training with massive historical fault data, and is especially suitable for the production of high-value biological fabrics with large batch differences. Example 2:

[0057] The specific steps for the baseline reconstructing unit to perform baseline derivation analysis include:

[0058] Extract the actuator's heating power data and the fan's air volume data from the input state vector, and calculate the theoretical bio-self-heating power and theoretical moisture evaporation and dissipation at the current moment based on the fabric fermentation kinetic parameters.

[0059] Perform energy conservation calculations: take the heating power data and theoretical biological self-heating power as heat inputs, subtract the heat loss term calculated from the fan air volume data and the thermal conductivity of the workshop enclosure structure, and obtain the net heat flux;

[0060] Perform mass conservation calculations: Calculate the absolute humidity change rate in the workshop based on the difference between the theoretical moisture evaporation and the fan exhaust rate;

[0061] Based on the net heat flux and the absolute humidity change rate, the theoretical temperature and humidity values ​​at the current moment are calculated by integration, and these theoretical temperature and humidity values ​​are constructed as an interference-free reference vector.

[0062] The calculation steps for theoretical biological self-heating power include:

[0063] Read the standard strain growth exothermic curve data and enzyme activity temperature correction coefficient from the fabric fermentation kinetic parameters;

[0064] Obtain the start time of the fermentation process, determine the corresponding time index in the standard strain growth exothermic curve data based on the current system clock, and extract the corresponding basal metabolic calorific value.

[0065] The theoretical temperature value at the current moment is called, and the basal metabolic calorific value is adjusted by gain or attenuation using the enzyme activity temperature correction coefficient. The corrected value is then output as the theoretical biological self-heating power.

[0066] This embodiment describes in detail the specific steps of the benchmark state reconstruction unit performing benchmark extrapolation analysis, and focuses on the construction process of the biothermodynamic model;

[0067] In order to accurately calculate the theoretical temperature and humidity, this embodiment constructs a fabric fermentation energy-mass coupling model;

[0068] Theoretical biological self-heating power ( This refers to the metabolic heat released by microorganisms in the fabric under ideal growth conditions. The calculation steps are as follows:

[0069] ;

[0070] Indicates time The theoretical biological self-heating power, in W;

[0071] Indicates time Basal metabolic calorific value; Source: Obtained by looking up a table in the pre-stored standard strain growth exothermic curve data in the database based on the start time of the fermentation process;

[0072] This represents the enzyme activity temperature correction coefficient; in this embodiment, a Gaussian correction model is used for calculation, and it is based on theoretical temperature values ​​rather than measured temperatures to avoid errors introduced by sensor malfunctions.

[0073] ;

[0074] in, The optimal growth temperature for the strain, such as 37℃. This is the preset bandwidth coefficient for enzyme activity;

[0075] Calculate the net heat flux within the workshop based on the first law of thermodynamics. :

[0076] ;

[0077] This indicates the heating power data of the actuator, in watts (W).

[0078] The data acquisition unit stores a fan characteristic curve table, which is used to look up the corresponding volumetric flow rate, i.e., air volume, based on the read speed command. ;

[0079] This represents the heat loss carried by the fan airflow data; the calculation formula is:

[0080] ;

[0081] in For the specific heat of air, air density, For air volume, The external ambient temperature of the workshop was collected;

[0082] The heat loss of the building envelope is represented by the following formula:

[0083] ;

[0084] in The overall heat transfer coefficient of the workshop walls. This represents the effective heat dissipation surface area of ​​the workshop.

[0085] Based on the mass balance equation, calculate the rate of change of absolute humidity in the workshop:

[0086] ;

[0087] in, This refers to the effective air volume in the fermentation and solidification workshop. Theoretical moisture evaporation and dissipation, in kg / s; this embodiment uses a modified Dalton evaporation formula for calculation:

[0088] ;

[0089] in, The evaporation coefficient is... This refers to the surface area of ​​the fabric. The saturated absolute humidity at the current theoretical temperature. This is the current theoretical absolute humidity; Calculated using the Magnus empirical formula, units are... The calculation formula is as follows:

[0090] ;

[0091] in, The saturated vapor pressure is calculated using the Magnus formula: ;constant hPa, , This formula achieves a precise nonlinear mapping from theoretical temperature to saturated absolute humidity; The fan exhaust rate needs to be converted to kg / s for calculation in this formula; the calculation formula is as follows: ,in The absolute humidity of the external environment of the workshop was collected;

[0092] Based on the above differential equation, the theoretical temperature value at the current moment is calculated by performing discrete-time step integration using the Euler method. and theoretical humidity value :

[0093] ;

[0094] ;

[0095] in The total heat capacity of the air in the workshop ( ), The sampling period is calculated as follows: and Construct an interference-free reference vector In this embodiment, the sampling period is... The preset time is 1 second, which is much smaller than the system's dominant time constant, typically ranging from tens of seconds to minutes. Therefore, the theoretical temperature and humidity extrapolation accuracy obtained using the first-order Euler method meets engineering requirements. It should be noted that the initial conditions for the iterative calculation are as follows: at the system startup initialization time... After the operator confirms that the workshop environment is in a standard, fault-free condition, the current temperature measured by the sensor is directly recorded. and humidity Assign initial values ​​to the undisturbed reference vector and In the subsequent moments, Then, iterative derivation is performed based on the above differential equation;

[0096] By introducing well-defined physical equations, such as the Gaussian model of enzyme activity and the Dalton evaporation model, and by strictly using theoretical state variables ( Instead of using measured variables in the iterative calculation, this embodiment eliminates the risk of black-box parameters and sensor fault coupling in traditional control strategies, ensuring that the generation of the reference vector has strict physical interpretability and numerical traceability, and providing accurate reference coordinates for subsequent extraction of minor fault features. Example 3:

[0097] The fault simulation injection analysis process is as follows:

[0098] The parameterized operators for various fault types are obtained from the expert knowledge base. The fault types include sensor drift faults and biological contamination anomaly faults.

[0099] The parameterized operators are superimposed and calculated onto the values ​​of the interference-free reference vector to simulate the data changes when different fault types occur.

[0100] Generate a theoretical damaged state vector corresponding to each fault type. The theoretical damaged state vector represents the theoretical data sequence under the condition that a specific fault occurs and there is no random environmental noise interference.

[0101] To address sensor drift faults, a drift interference operator is invoked. This operator defines the drift deviation value as a nonlinear function of the integral value of historical humidity data over time, and calculates the amount of humidity deviation that accumulates over time and superimposes it onto an interference-free reference vector.

[0102] For abnormal faults caused by biological contamination, a heat source disturbance operator is invoked. This operator defines the abnormal heat source value as an exponential function of time, calculates the cumulative heat energy value that grows exponentially, and converts it into the corresponding temperature increment, which is then superimposed on the undisturbed reference vector.

[0103] This embodiment describes in detail the fault simulation injection analysis process in the disturbance simulation unit, especially the definition and calculation of parameterized operators;

[0104] In order to identify risks that have never occurred but theoretically exist, this invention employs a knowledge-driven reverse fault injection technique;

[0105] The system retrieves preset fault mechanism models from an expert knowledge base; in this embodiment, it focuses on two typical types of faults:

[0106] Sensor drift fault: refers to the monotonic deviation in readings caused by the aging of capacitive humidity-sensitive elements under high humidity conditions;

[0107] Biological contamination abnormality: refers to the invasion of miscellaneous bacteria, such as mold, which causes abnormal local temperature rise in the fermentation bed;

[0108] The system superimposes the corresponding mathematical operators to calculate the interference-free reference vector. In the numerical values; for example, generating a vector simulating drift faults. and vectors of simulated biocontamination These vectors are collectively referred to as theoretical damaged state vectors.

[0109] For sensor drift faults - drift interference operator: This operator defines the drift deviation value as a nonlinear function of the integral value of historical humidity data over time, and its calculation formula is as follows:

[0110] ;

[0111] in, This is the preset characteristic lifetime constant of the sensor. This represents the upper limit of the humidity measurement range; the formula within parentheses uses division by... Dimensionless processing of the integral value was implemented, thereby ensuring The physical rationale behind exponential operations;

[0112] :time The superimposed humidity deviation;

[0113] Baseline humidity data for historical periods;

[0114] and The preset drift coefficient and nonlinearity factor are derived from sensor aging test data;

[0115] The system calculates the cumulative increase in humidity deviation over time and adds it to... Above, simulate sensor poisoning phenomena;

[0116] For the anomaly fault caused by biological contamination - heat source interference operator: This operator defines the value of the abnormal heat source as an exponential function of time, and its calculation formula is as follows:

[0117] ;

[0118] in, :time Superimposed temperature increment; The sampling period; Inject the start time into the preset simulated fault;

[0119] Initial pollution thermal power coefficient, in W;

[0120] The growth rate constant of miscellaneous bacteria is usually much larger than that of the target bacteria.

[0121] : The heat capacity constant of this region;

[0122] The exponentially increasing additional heat value is calculated and converted into the corresponding temperature increment, which is then added to... superior;

[0123] By defining well-defined mathematical operators, the integral function simulates cumulative damage, and the exponential function simulates biological outbreaks. This invention transforms fuzzy empirical faults into precise mathematical waveforms. The effect of this is that the system no longer relies on massive amounts of labeled data for training, but directly uses physical and biological mechanisms to identify faults. Even if the production line is running for the first time, as long as a temperature rise that meets the characteristics of exponential growth is detected, the system can immediately determine that it is biological contamination, rather than simply overheating. Example 4:

[0124] The specific steps for the dual-track differential unit to perform the first path difference calculation and the second path difference calculation are as follows:

[0125] First path calculation: The temperature and humidity values ​​collected in real time by the sensor are used to construct a measured state vector. The measured state vector is subtracted from the interference-free reference vector to obtain a real difference feature vector containing real abnormal components and random noise components.

[0126] Second path calculation: Subtract the undisturbed baseline vector from the theoretical damaged state vector to obtain the theoretical difference feature vector containing only the fault mode feature components;

[0127] After aligning the actual difference feature vector with the theoretical difference feature vector in time series, the data is sent to the coupling control unit.

[0128] This embodiment describes in detail the specific steps of the dual-track differential unit to perform path difference calculation;

[0129] To extract the true fault characteristics from noisy industrial field data, this embodiment designs the following two independent calculation paths:

[0130] The first path calculation, real-world difference extraction, constructs a measured state vector from the real-time temperature and humidity values ​​collected by the sensor. Perform subtraction:

[0131] ;

[0132] The real-world difference feature vector contains three components: real system anomalies, environmental random noise, and unmodeled residuals.

[0133] The second path calculation, theoretical difference extraction, targets the first... For each failure mode, the theoretical damaged state vector is subtracted from the undisturbed baseline vector:

[0134] ;

[0135] The theoretical difference feature vector only contains the first... The pure characteristic component of a failure mode, for example, is simply a pure exponentially rising curve without any spikes;

[0136] The system performs time window truncation and sequence alignment on the two vectors mentioned above, and sends them to the coupling control unit in parallel;

[0137] The core innovation of this embodiment lies in using a pure reference as a medium to achieve differential countermeasures; traditional control systems directly process... It is difficult to distinguish between signal and noise; this invention reduces the noise level by subtracting the signal. It cleverly eliminates normal temperature and humidity changes, such as those occurring during normal heating processes; this makes... Any non-zero value in the equation represents an unexpected event; while This provides standard failure modes for unexpected events; this dual-track design significantly reduces the signal-to-noise ratio requirements for signal processing. Example 5:

[0138] The similarity comparison decision process performed by the coupling control unit is as follows:

[0139] Map the received real-world difference feature vectors and theoretical difference feature vectors to the same feature vector space;

[0140] The optimal matching path distance between two vector sequences on the time axis is calculated using the dynamic time warping algorithm, or the cosine similarity algorithm is used to calculate the cosine value of the angle between two vectors in multidimensional space, thus obtaining a normalized similarity value.

[0141] The similarity value is compared with a preset judgment threshold in the memory to output the specific event judgment signal or environmental noise signal.

[0142] The similarity score is calculated using either the dynamic time warping algorithm or the cosine similarity algorithm. This algorithm is used to distinguish the random noise components contained in the real-world differences and to extract fault features that are consistent with the theoretical differences in terms of their changing trends.

[0143] This embodiment describes in detail the similarity comparison and decision process performed by the coupling control unit;

[0144] Preprocessing steps: The system is set with a length of ,For example A first-in-first-out (FIFO) buffer queue for each sampling period; the actual difference feature vector and the theoretical difference feature vector are not values ​​at a single moment, but are determined by the current moment. and the past A sliding time window sequence matrix consisting of data from each moment;

[0145] ;

[0146] ;

[0147] In order to determine Is the fluctuation in the middle actually...? The system calculates the similarity value between the predicted fault and the two faults in the number domain space. This embodiment provides two specific algorithm implementations:

[0148] Algorithm A: The cosine similarity algorithm is suitable for scenarios that are sensitive to waveform direction but not to amplitude.

[0149] ;

[0150] The algorithm calculates the cosine of the angle between two vectors in a multidimensional space; if the result is close to 1, it indicates that the two vectors have a highly consistent trend.

[0151] Algorithm B: Dynamic Time Warping (DTW) is suitable for scenarios where there is a slight offset in the time of the fault occurrence. This algorithm uses Euclidean distance as the base distance metric to calculate the distance of the best matching path between two vector sequences on the time axis.

[0152] ;

[0153] The DTW algorithm allows sequences to be non-linearly twisted along the time axis, thus enabling the identification of fault features that are identical in shape but occur at slightly different times.

[0154] Calculated Compared with the preset judgment threshold in the memory For example, 0.85; Comparison: If System confirmation The fluctuations in are caused by the first This is caused by a fault; in this case, a specific event determination signal is output; if System judgment The fluctuations in the signal do not match any known fault model and are classified as environmental noise; at this point, the environmental noise signal is output; effect description: this step achieves mathematical immunity to random noise; because random noise usually follows a Gaussian distribution, it is almost orthogonal to the theoretical fault vector with specific physical laws, such as exponential functions and integral functions, in the feature space;

[0155] By employing cosine similarity or DTW algorithms, this invention no longer relies on simple over-limit alarms, i.e., absolute numerical magnitude, to determine faults, but instead uses morphological recognition. Even if the sensor data fluctuations are small, as long as the trend of change conforms to the exponential characteristics of biological contamination, the system can still accurately identify them. Even if the sensor data fluctuations are large, such as instantaneous changes caused by someone opening or closing a door in the workshop, as long as the trend of change does not conform to any fault model, the system will identify it as noise and ignore it, without taking erroneous control actions. This sensitive yet stable control characteristic perfectly meets the requirements of bio-based fabric production, which is extremely sensitive to the environment.

[0156] Engineering design parameters: Overall heat transfer coefficient of workshop walls The optimal growth temperature for the microbial strain was obtained by consulting the workshop building structure design manual. and enzyme activity broadband coefficient Obtained by reviewing experimental data on the biological characteristics of the target bacterial species;

[0157] Experimental calibration parameters: Evaporation coefficient The actual evaporation and moisture dissipation rate of the fabric was obtained by least squares regression of the measured actual evaporation and moisture dissipation rate with the theoretical calculated value in a controlled experimental environment with no load and constant temperature and humidity in the workshop.

[0158] Sensor drift parameters and The data was obtained by placing a large number of sensors in a high-humidity aging chamber for long-term experiments, collecting their reading drift data, and then using a nonlinear regression algorithm to fit the drift deviation function.

[0159] Biological pollution parameters and The initial exothermic power and growth rate of the mixed bacteria were measured at different temperatures through in vitro culture experiments, and obtained by fitting exponential curves.

[0160] Decision threshold: the threshold for making a judgment The value, such as 0.85, is determined by conducting backtesting tests on historical fault data in an offline simulation environment, such as ROC curve analysis, to ensure the highest fault detection rate while controlling the false alarm rate within an acceptable range, such as within 5%.

[0161] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A smart temperature and humidity control system for bio-based fabric production, characterized in that, It includes a data acquisition unit, a reference state reconstruction unit, a disturbance simulation unit, a dual-track differential unit, and a coupling control unit; The data acquisition unit is equipped with an industrial communication interface and an environmental sensor group, which is used to collect the temperature and humidity sensor values ​​in the fermentation and curing workshop in real time, read the current heating power and fan speed command of the actuator, and call the pre-stored fabric fermentation dynamic parameters. The above data is combined into an input state vector and sent to the reference state reconstruction unit. The reference state reconstruction unit is used to receive the input state vector, perform calculations based on the thermodynamic energy conservation model, obtain the theoretical temperature and humidity after eliminating the influence of environmental noise and equipment failure, generate an interference-free reference vector, and synchronously send the interference-free reference vector to the disturbance simulation unit and the dual-track differential unit. The disturbance simulation unit is connected to a database storing fault parameterization operators. It is used to superimpose a preset disturbance factor onto a non-disturbed reference vector to generate a theoretical damaged state vector characterizing a specific fault mode, and send the theoretical damaged state vector to the dual-track differential unit. The dual-track differential unit is used to calculate the vector difference between the temperature and humidity sensor values ​​and the interference-free reference vector to generate the actual difference feature vector, and at the same time calculate the vector difference between the theoretical damaged state vector and the interference-free reference vector to generate the theoretical difference feature vector. The coupling control unit is used to calculate the similarity value between the actual difference feature vector and the theoretical difference feature vector in the number domain space, and to perform the following logical judgment: if the similarity value is greater than a preset judgment threshold, a specific event judgment signal is generated, and a feedforward compensation control command is sent to the actuator to compensate for the temperature and humidity deviation; if the similarity value is less than or equal to the preset judgment threshold, an environmental noise signal is generated, and the current instruction of the actuator is locked to maintain the current temperature and humidity control strategy unchanged and not respond to the fluctuations in the actual difference feature vector. The specific steps for the benchmark state reconstruction unit to perform benchmark deduction analysis include: Extract the actuator's heating power data and fan air volume data from the input state vector, and calculate the theoretical bio-self-heating power and theoretical moisture evaporation and dissipation at the current moment based on the fabric fermentation kinetic parameters; Perform energy conservation calculations: take the heating power data and theoretical biological self-heating power as heat inputs, subtract the heat loss term calculated from the fan air volume data and the thermal conductivity of the workshop envelope, and obtain the net heat flux; Perform mass conservation calculations: Calculate the absolute humidity change rate in the workshop based on the difference between the theoretical moisture evaporation and the fan exhaust rate; Based on the net heat flux and the absolute humidity change rate, the theoretical temperature and humidity values ​​at the current moment are calculated by integration, and these theoretical temperature and humidity values ​​are used to construct an interference-free reference vector.

2. The intelligent temperature and humidity control system for bio-based fabric production according to claim 1, characterized in that, The calculation steps for the theoretical biological self-heating power include: Read the standard strain growth exothermic curve data and enzyme activity temperature correction coefficient from the fabric fermentation kinetic parameters; Obtain the start time of the fermentation process, determine the corresponding time index in the standard strain growth exothermic curve data based on the current system clock, and extract the corresponding basal metabolic calorific value. Collect real-time temperature data of the workshop, use enzyme activity temperature correction coefficient to gain or attenuate the basal metabolic calorific value, and output the corrected value as the theoretical biological self-heating power.

3. The intelligent temperature and humidity control system for bio-based fabric production according to claim 1, characterized in that, The fault simulation injection analysis process is as follows: obtain parameterized operators for various fault types preset in the expert knowledge base, including sensor drift faults and biological contamination anomaly faults; superimpose the parameterized operators onto the values ​​of the interference-free reference vector to simulate the data changes when different fault types occur; generate a theoretical damaged state vector corresponding to each fault type, which represents the theoretical data sequence under the condition that a specific fault occurs and there is no environmental random noise interference.

4. The intelligent temperature and humidity control system for bio-based fabric production according to claim 3, characterized in that, The definition and calculation process of the parameterized operator are as follows: To address sensor drift faults, a drift interference operator is invoked. This operator defines the drift deviation value as a nonlinear function of the integral value of historical humidity data over time, and calculates the amount of humidity deviation that accumulates over time and superimposes it onto an interference-free reference vector. For abnormal faults caused by biological contamination, the heat source interference operator is invoked. This operator defines the abnormal heat source value as an exponential function of time, calculates the additional heat value that grows exponentially, and converts it into the corresponding temperature increment, which is then superimposed on the interference-free reference vector.

5. The intelligent temperature and humidity control system for bio-based fabric production according to claim 1, characterized in that, The specific steps for the dual-track differential unit to perform the first path difference calculation and the second path difference calculation are as follows: First path calculation: The temperature and humidity values ​​collected in real time by the sensor are used to construct a measured state vector. The measured state vector is subtracted from the interference-free reference vector to obtain a real difference feature vector containing real abnormal components and random noise components. Second path calculation: Subtract the undisturbed baseline vector from the theoretical damaged state vector to obtain the theoretical difference feature vector containing only the fault mode feature components; After aligning the actual difference feature vector with the theoretical difference feature vector in time series, the data is sent to the coupling control unit.

6. The intelligent temperature and humidity control system for bio-based fabric production according to claim 1, characterized in that, The similarity comparison decision process performed by the coupling control unit is as follows: Map the received real-world difference feature vectors and theoretical difference feature vectors to the same feature vector space; The optimal matching path distance between two vector sequences on the time axis is calculated using the dynamic time warping algorithm, or the cosine similarity algorithm is used to calculate the cosine value of the angle between two vectors in multidimensional space, thus obtaining a normalized similarity value. The similarity value is compared with a preset judgment threshold in the memory to output the specific event judgment signal or environmental noise signal.

7. The intelligent temperature and humidity control system for bio-based fabric production according to claim 6, characterized in that, The similarity values ​​are calculated using a dynamic time warping algorithm or a cosine similarity algorithm to distinguish random noise components contained in real-world differences and extract fault features that are consistent with theoretical differences in their trends of change.

Images