A production behavior water footprint tracking method based on hydraulic entropy value characteristics and blockchain consensus verification

Abstract

This invention relates to a method for tracking the water footprint of production activities based on hydraulic entropy characteristics and blockchain consensus verification. The method includes the following steps: Step 1: Install high-precision sensors at key locations in the water-using unit, and collect real-time data on pipeline water pressure, flow rate, and velocity at different time scales through a triple sampling sequence processing layer of high frequency, medium frequency, and low frequency. Step 2: Based on an on-chain-off-chain collaborative verification architecture, design an entropy feature extraction function for off-chain edge computing nodes, and construct a multi-dimensional entropy calculation model. Step 3: Based on the on-chain-off-chain collaborative verification architecture, on-chain smart contracts perform cross-node feature consistency verification; design a trusted feature library self-updating mechanism, automatically updating the template library when a new feature is verified as a trusted feature. Step 4: Call the working condition identification module to analyze real-time entropy features, calculate the water entropy value of each stage of the production chain, and return the complete water footprint chain by querying through the product ID.

Description

Technical Field

[0001] This invention belongs to the field of water use behavior analysis and production behavior identification technology, specifically involving a method and system for tracking the water footprint of production behavior based on hydraulic entropy characteristics and blockchain consensus verification. It is applicable to the accurate identification of water use behavior in high water-consuming industries and the management of water resource consumption in the production chain. Background Technology

[0002] The global water shortage problem is becoming increasingly severe, and water footprint tracking technology has become an important means to optimize water resource management and promote sustainable development. Traditional water use behavior identification methods are not accurate enough when dealing with complex operating conditions, especially in the production environments of water-intensive industries. Some companies have begun to adopt water footprint tracking technology to track water consumption at each stage of the production and supply chain, but the following problems still exist:

[0003] (1) Traditional methods often fail to effectively capture and analyze nonlinear hydraulic characteristics under scenarios such as equipment start-up and shutdown, and sudden load changes, resulting in a high misjudgment rate of water use behavior;

[0004] (2) In the existing technology, the data acquisition and authentication process is separated. Sensor data needs to be uploaded to the central server and then manually reviewed. There is a risk of tampering in the intermediate process, which makes it difficult to guarantee the credibility of the identification results.

[0005] (3) Existing technologies mostly focus on the water consumption statistics of single processes, lacking a mechanism for the flow of production water consumption across processes, making it difficult to achieve supply chain-level collaborative management of water resources.

[0006] Therefore, this invention proposes a method and system for tracking the water footprint of production behavior based on hydraulic entropy characteristics and blockchain consensus verification. By combining hydraulic entropy characteristic analysis with blockchain consensus verification, the method achieves integrated tracking, analysis and storage of the water footprint of production behavior, enabling accurate identification of water use behavior under complex working conditions and highly reliable management of water resource consumption in the production chain. Summary of the Invention

[0007] To address the shortcomings of the aforementioned water use behavior collection and identification technologies, this invention provides a method and system for tracking the water footprint of production behavior based on hydraulic entropy characteristics and blockchain consensus verification. By combining multi-scale hydraulic entropy characteristic analysis with blockchain consensus verification, the method achieves accurate tracking of the water footprint of production behavior, enabling real-time, reliable, and traceable management of water resource consumption in the production chain.

[0008] To achieve the above objectives, the technical solution adopted by this invention is: a method for tracking the water footprint of production behavior based on hydraulic entropy characteristics and blockchain consensus verification, the specific steps of which include:

[0009] Step 1: Install high-precision sensors at key locations in the water-using unit. Through a triple sampling sequence processing layer of high frequency, medium frequency, and low frequency, collect real-time data on pipeline water pressure, flow rate, and velocity at different time scales.

[0010] S11: Install high-precision sensors at key locations such as industrial water pipes and equipment inlets to collect data such as water pressure, flow rate, and flow velocity in the pipeline.

[0011] Optionally, the power pulse signal of the device can be collected synchronously.

[0012] S12: Design a triple sampling sequence processing layer to capture the characteristics of water pressure, flow rate and other data at different time scales in real time through high-frequency, medium-frequency and low-frequency triple sampling.

[0013] Furthermore, the triple sampling sequence processing layer includes a high-frequency sampling layer, a mid-frequency sampling layer, and a low-frequency sampling layer, which respectively capture data features at different time scales.

[0014] Specifically, the high-frequency sampling layer captures transient fluctuations in data such as water pressure and flow rate, with a short sampling interval, capturing sudden changes in water flow caused by valve opening and closing and equipment start-up and shutdown; the medium-frequency sampling layer captures operational cycle characteristics, with a moderate sampling interval, used to identify periodic operations such as equipment cleaning and cooling cycles; and the low-frequency sampling layer captures long-term trends in data such as water pressure and flow rate, with a longer sampling interval, enabling monitoring of water consumption trends during equipment operation and product production throughout the day.

[0015] Step 2: Based on the on-chain-off-chain collaborative verification architecture, design the entropy feature extraction function for off-chain edge computing nodes. By constructing a multi-dimensional entropy calculation model, calculate the primary entropy value and generate a multi-dimensional entropy feature vector; by dynamically adjusting the entropy weights through a sliding window variance detector, calculate the comprehensive entropy feature vector.

[0016] S21: Deploy edge computing devices near the data source to input the collected data such as pipe water pressure and flow rate into the edge processing module.

[0017] Specifically, the edge processing module includes a multidimensional entropy calculation unit and a feature entropy processing unit.

[0018] S22: Construct a multidimensional entropy calculation model within the multidimensional entropy calculation unit, calculate the entropy features and fuse them with the time series to form a multidimensional entropy feature vector under the time series.

[0019] Furthermore, the multidimensional entropy calculation model includes three entropy calculation methods: multi-scale permutation entropy, time-frequency domain wavelet entropy, and adaptive fuzzy entropy, which are used to calculate entropy features at different scales and frequency bands.

[0020] (1) Multiscale permutation entropy is a nonlinear time series analysis method that decomposes a time series into multiple time-domain scales, calculates the permutation entropy value at each scale, and combines them to obtain the complexity characteristics of the entire time series. Specifically, the calculation steps of multiscale permutation entropy include:

[0021] First, the time series is processed by normalization and noise reduction. Then, wavelet decomposition is performed to obtain approximate coefficient sequences at different scales. These approximate sequences are then arranged to obtain permutation sequences at different scales. The permutation entropy value at each scale is calculated, and the permutation entropy values ​​at different scales are combined to obtain the multi-scale permutation entropy value of the entire sequence. The relevant formulas and principles are as follows:

[0022]

[0023] Where m is the dimension, H d Let be the multiscale permutation entropy in dimension m, π be the permutation pattern, and p(π) be the probability of the permutation entropy occurring.

[0024] (2) Time-frequency domain wavelet entropy combines the time-frequency localization characteristic of wavelet analysis with the concept of entropy, enabling it to better describe the complexity and randomness of signals at different scales. Specifically, the calculation steps for time-frequency domain wavelet entropy include:

[0025] First, the signal is decomposed into wavelet coefficients to obtain a series of wavelet coefficients. The probability distribution of each wavelet coefficient is calculated, and finally the entropy value is calculated based on the probability distribution.

[0026]

[0027] Among them, H s Let be the wavelet entropy in the time-frequency domain, p(k) be the energy probability of the k-th subband, and k be the number of subband sequences. j be the number of sample sequences.

[0028] (3) Adaptive fuzzy entropy, combining fuzzy membership functions and adaptive parameter adjustment, measures the probability of a time series generating new patterns as its dimension changes. The higher the probability of a sequence generating new patterns, the higher the complexity of the sequence. Specifically, the calculation steps for adaptive fuzzy entropy include:

[0029]

[0030] Where i and j are the number of sequences from different samples, m is the dimension, r is the adaptive similarity tolerance threshold, and n is the similarity tolerance boundary gradient. N is the total number of time series, and D is the fuzzy similarity metric. H z It is an adaptive fuzzy entropy.

[0031] S23: A sliding window variance detector is used to monitor entropy changes in real time. The fusion weights are dynamically adjusted based on the variance, and a comprehensive entropy feature vector is output for the time series. Specific steps include:

[0032] S231: Extracting the entropy sequence of the M most recent time points at time t based on the sliding window method;

[0033] Specifically, select a starting position t, initialize the window size, and calculate the initial entropy value. Move the window forward one position and update the entropy value within the window. Repeat this operation M times to obtain M entropy value sequences.

[0034] S232: Calculate the variance of each type of entropy value within the window, reflecting the degree of dispersion of each entropy value; calculate the mean of each type of entropy value separately, and calculate the variance separately. The larger the variance, the greater the dispersion of the entropy value. The relevant formulas are as follows:

[0035]

[0036] Where B is the number of windows, H is the entropy value, and V is the variance of the corresponding entropy value.

[0037] S233: Calculate the weights of each entropy feature dynamically based on variance;

[0038] Specifically, Softmax weighting is used to amplify variance differences, and the weight distribution of each entropy value is dynamically controlled by adjusting the temperature coefficient. The relevant formulas are as follows:

[0039]

[0040] Where δ a Let Γ be the variance of the a-th feature entropy value, Γ be the temperature coefficient controlling the weight distribution, and δ be the variance of the a-th feature entropy value. l Let ω be the variance of the l-th feature entropy value, A be the total variance of the feature entropy values, and ω be the variance of the l-th feature entropy value. a is the weight of the entropy value of the a-th feature.

[0041] S234: After normalizing the three types of entropy values ​​respectively, a weighted fusion is performed based on their respective weights to output a comprehensive entropy feature vector for the time series. The underlying formulas are as follows:

[0042]

[0043] in, The normalized multi-scale permutation entropy takes values ​​in the range [0,1]. ln(m!) represents the theoretical maximum value of the permutation entropy.

[0044]

[0045] in, Let L be the normalized time-frequency wavelet entropy, and L be the total number of subbands.

[0046]

[0047] in, For normalized adaptive fuzzy entropy, ln(m!) represents the theoretical maximum value of permutation entropy.

[0048]

[0049] Where H d,s,z The comprehensive feature entropy value, ω d ω s ω z These are the weights of the multi-scale permutation entropy, the time-frequency domain wavelet entropy, and the adaptive fuzzy entropy, respectively.

[0050] S24: The eigenvector of the comprehensive entropy value is reduced in dimension and compressed by principal component analysis and output to memory to reduce the subsequent on-chain storage pressure.

[0051] Step 3: Based on the on-chain-off-chain collaborative verification architecture, on-chain smart contracts perform cross-node feature consistency verification; design a trusted feature library self-updating mechanism, automatically updating the template library when a new feature is verified as a trusted feature; abnormal features trigger alarms, and multi-party signatures are used for evidence storage for audit traceability; based on the random forest model, different working conditions are associated with entropy features to build a working condition identification module.

[0052] S31: Off-chain edge nodes upload the output feature entropy vector to the blockchain network.

[0053] S32: On-chain smart contracts execute blockchain consensus verification, comparing entropy feature vectors submitted by multiple edge nodes to perform cross-node feature consistency verification.

[0054] Specifically, the steps for executing a smart contract include:

[0055] S321: The on-chain smart contract receives entropy feature data from multiple edge nodes, calculates the cosine similarity between each node, and selects highly similar node pairs to form a set C.

[0056] Specifically, for the entropy feature vectors of J edge nodes, the cosine similarity between any two nodes p and q is calculated using the following formula:

[0057]

[0058] Among them, S p,q It is the similarity between nodes p and q; v p and v q Let v be the entropy eigenvectors of nodes p and q; p ‖ and ‖v q || represents vector vp and v q The Euclidean norm,

[0059] C={q≠p|S p,q ≥T}

[0060] Where T is the similarity threshold, typically ranging from 0.8 to 0.9.

[0061] S322: If the number of nodes in set C exceeds K, the smart contract performs weighted voting based on the dynamic weight mechanism of each node. If the voting results reach a consensus, it is determined to be a trustworthy feature and triggers the trustworthy feature library self-update mechanism.

[0062] Furthermore, the dynamic weighted voting mechanism includes the following steps:

[0063] (1) Set weighting factors and calculate the overall weight.

[0064] a. Historical credibility weight w1:

[0065]

[0066] Where P1 represents the percentage of node q that has historically participated in consensus and passed the judgment.

[0067] b. Real-time consistency weight w2:

[0068]

[0069] in, Let q be the average similarity between node q and set C, and T be the similarity threshold of 55.

[0070] c. Online stability weight w3:

[0071]

[0072] Wherein, P2 is the proportion of time that node q remains online and reports data on time during the monitoring period.

[0073] For a set C of highly similar node pairs, calculate the voting weight w of each node. q And after normalization, we get w′ q The specific calculation formula is as follows:

[0074] w q =w1+w2+w3

[0075]

[0076] (2) Weighted voting and consensus determination

[0077] The smart contract calculates the average similarity of each node. Automatically assign "agree" or "disagree" and calculate the weighted agreement rate R, using the following formula:

[0078]

[0079] When R exceeds the consensus rate threshold α, the node's characteristics are determined to be trustworthy, triggering the trustworthy characteristic database self-update mechanism.

[0080] S323: If the similarity between a node feature k and most nodes is less than the threshold T, it is determined to be an abnormal feature and the anomaly detection mechanism is triggered.

[0081] S33: Establish a self-updating mechanism for the trusted feature library: write features that meet the criteria for trusted feature determination into the trusted feature library, generate a new version of the feature template, and record it on the blockchain as a comparison benchmark for subsequent consistency verification.

[0082] S34: Establish an anomaly detection mechanism: Store the anomaly features collected by the anomaly detection mechanism into the blockchain evidence repository, and use multi-party signatures for evidence storage to ensure data traceability.

[0083] S35: Based on the random forest model, different operating conditions are associated with entropy features to establish an operating condition identification module. Specific steps include:

[0084] S351: Extract the historical entropy value feature vector and label the corresponding working conditions;

[0085] Specifically, different operating conditions include normal production, equipment maintenance, and abnormal leaks.

[0086] S352: Entropy feature vectors and their operating conditions are used as training sets, input into a random forest model to train a classification model, and the model parameters are adjusted through cross-validation.

[0087] S353: Convert the multi-condition output of the classification model into a probability distribution matrix, and set corresponding probability thresholds to represent the probability of each condition. When the predicted probability exceeds the threshold, the system is considered to be in that condition.

[0088] Step 4: Call the working condition identification module to analyze the real-time entropy value characteristics, calculate the water entropy value of each stage of the production chain, and realize the unique recording of water footprint certificates through blockchain smart contracts; construct a dynamic path map of water footprint based on the production timeline, entropy intensity, and water resource flow path, and build a reverse traceability mechanism to return the complete water footprint chain by querying through product ID.

[0089] S41: When a product enters a critical process, the system first calls the entropy feature calculation module to extract and process the entropy feature vector, and then calls the working condition identification module to calculate the water entropy value of this stage in combination with the actual water consumption data of this process.

[0090] S42: Analyze the current entropy characteristics through smart contracts and update the trusted feature library in real time; calculate and accumulate the water footprint increment of the product in this stage and update the water footprint; write the water footprint attributes into the blockchain to generate a unique water footprint certificate to ensure that it cannot be tampered with.

[0091] The formula for calculating the cumulative increase in water footprint is as follows:

[0092] WF = WF total +WF physical +γ×WF stage

[0093] Among them, WF total Cumulative water footprint associated with product ID; WF physical For direct water use in the current stage; WF stage For the current stage of water resource consumption, WF stage =‖v stage ||;||v stage ‖ represents the current water resource entropy feature vector v stage The Euclidean norm; γ is a dynamically adjusted parameter for different operating conditions.

[0094] Specifically, the water footprint certificate includes basic information (such as certificate ID, product ID, process information, and production time), water footprint data (such as details of water resource consumption at each production stage, direct water consumption at each stage, and the updated cumulative total water footprint), entropy characteristics and operating condition information, reliable evidence storage information, and associated anomaly records.

[0095] S43: Based on the production timeline (X-axis), entropy intensity (Y-axis), and water resource flow path (Z-axis), a 3D topology map is drawn in real time, dynamically displaying the flow path and intensity changes of the water footprint.

[0096] Specifically, the X-axis represents the process completion time or production progress percentage, the Y-axis reflects the oscillation intensity of the operating state using the entropy norm, and the Z-axis arranges process nodes at equal intervals and indicates the direction and magnitude of water flow with directed lines.

[0097] S44: Construct a reverse traceability mechanism: Query the blockchain through the final product ID, and the smart contract returns complete water footprint chain certificate information according to the on-chain records, including the entropy change curve of each process, water consumption details, and the signatures of the blockchain nodes that participated in the verification.

[0098] Specifically, the query process is as follows: each product is assigned a unique QR code, and consumers scan the code to send the product ID to the blockchain; the smart contract retrieves all water footprint credentials associated with that ID, including text, path maps, and entropy analysis reports.

[0099] An electronic device includes a processor, a memory, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the aforementioned method for tracking the water footprint of production behavior based on hydraulic entropy characteristics and blockchain consensus verification.

[0100] A computer-readable storage medium includes a stored computer program, wherein, when the computer program is executed, it controls the device where the computer-readable storage medium is located to perform the aforementioned method for tracking the water footprint of production behavior based on hydraulic entropy characteristics and blockchain consensus verification.

[0101] This invention achieves precise tracking of the water footprint of production activities by combining multi-scale hydraulic entropy characteristic analysis with blockchain consensus verification.

[0102] Compared with the prior art, the present invention has the following beneficial effects:

[0103] This invention uses high-precision sensors and triple samplers to collect data from key water-using nodes. It uses high-frequency, medium-frequency, and low-frequency triple sampling layers to capture transient fluctuations, periodic operations, and long-term trends in water usage at the nodes, respectively, thus solving the problem of insufficient sensitivity of traditional single sampling methods to nonlinear hydraulic characteristics.

[0104] This invention constructs a multi-dimensional entropy calculation model to calculate multi-scale permutation entropy, time-frequency domain wavelet entropy, and adaptive fuzzy entropy, forming a multi-dimensional entropy feature vector. A sliding window variance detector is introduced to monitor entropy changes in real time and automatically adjust feature weights to obtain a comprehensive entropy feature, ensuring robustness under different production scenarios. Compared to traditional single entropy methods, this invention improves the accuracy of water usage behavior identification under complex operating conditions such as equipment transient fluctuations, process switching, and load abrupt changes. It breaks through the dependence of traditional statistical models on steady-state conditions and adapts to the high noise, nonlinearity, and strong time-varying characteristics of industrial sites.

[0105] This invention adopts an on-chain-off-chain collaborative architecture. On the one hand, it deploys edge computing devices off-chain to calculate the comprehensive entropy characteristics of each node based on the edge processing module, while filtering out some redundant data to reduce the on-chain storage load. On the other hand, it uses on-chain smart contracts to perform cross-node feature consistency verification to ensure the integrity of the data written to the blockchain.

[0106] This invention, based on a blockchain consensus mechanism, uploads key data such as entropy characteristics, water consumption, and timestamps to the chain, forming an immutable water footprint traceability chain, significantly enhancing tamper resistance. Through smart contracts, it automatically triggers water footprint certificate updates and constructs a multi-party signature trusted feature library mechanism. Both trusted and abnormal features must be jointly verified by multiple nodes before being stored, greatly improving data credibility. Compared to traditional unilateral evidence storage models, this invention solves the problems of easily tampered traditional centralized databases and high costs of mutual trust among multiple parties in the supply chain.

[0107] This invention designs a 3D water footprint topology map based on the production timeline (X-axis), entropy intensity (Y-axis), and water resource flow path (Z-axis), intuitively displaying the dynamic flow relationships of the water footprint and improving regulatory transparency. It also constructs a water footprint reverse traceability mechanism, supporting automatic accounting and unique storage of the water footprint throughout the product's entire lifecycle, enabling rapid location of abnormal processes and reverse traceability of the entire water footprint process. Compared to traditional water footprint tracking methods, this invention overcomes the static limitations of traditional paper certificates, achieving dynamic binding of water footprints and transparent cross-industry circulation. Attached Figure Description

[0108] Figure 1 This is a flowchart of a method for tracking the water footprint of production activities based on hydraulic entropy characteristics and blockchain consensus verification.

[0109] Figure 2 This is a three-dimensional topological diagram illustrating the flow path and intensity variations of a water footprint.

[0110] Figure 3 This is a schematic diagram of a water footprint tracking device for production activities based on hydraulic entropy characteristics and blockchain consensus verification.

[0111] Figure 4 This is a schematic diagram of the physical structure of an electronic device. Detailed Implementation

[0112] The technical solution of the present invention will be clearly and completely described below with reference to specific embodiments. 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.

[0113] Example: The technical solution adopted in this invention is: a method for tracking the water footprint of production behavior based on hydraulic entropy characteristics and blockchain consensus verification, the specific steps of which include:

[0114] To achieve the above objectives, the technical solution adopted by this invention is: a method for tracking the water footprint of production behavior based on hydraulic entropy characteristics and blockchain consensus verification, the specific steps of which include:

[0115] Step 1: Install high-precision sensors at key locations in the water-using unit. Through a triple sampling sequence processing layer of high frequency, medium frequency, and low frequency, collect real-time data on pipeline water pressure, flow rate, and velocity at different time scales to obtain a long-term time series dataset X0 (water pressure, flow rate, velocity) of pipeline water pressure and flow rate.

[0116] S11: Install high-precision sensors at key locations such as industrial water pipes and equipment inlets to collect data such as water pressure and flow rate in the pipeline.

[0117] Optionally, the power pulse signal of the device can be collected synchronously.

[0118] S12: Design a triple sampling sequence processing layer to capture the characteristics of water pressure, flow rate, and flow velocity at different time scales in real time through high-frequency, medium-frequency, and low-frequency triple sampling.

[0119] Furthermore, the triple sampling sequence processing layer includes a high-frequency sampling layer, a mid-frequency sampling layer, and a low-frequency sampling layer, which respectively capture data features at different time scales.

[0120] Specifically, the high-frequency sampling layer captures transient fluctuations in data such as water pressure and flow rate, with a sampling interval of 60 seconds, to capture sudden changes in water flow caused by valve opening and closing and equipment start-up and shutdown; the medium-frequency sampling layer captures operational cycle characteristics, with a sampling interval of 15 minutes, to identify periodic operations such as equipment cleaning and cooling circulation; and the low-frequency sampling layer captures long-term trends in data such as water pressure and flow rate, with a sampling interval of 1 hour, to monitor the water consumption trends of equipment operation throughout the day and product production.

[0121] Step 2: Based on the on-chain-off-chain collaborative verification architecture, design the entropy feature extraction function for off-chain edge computing nodes. By constructing a multi-dimensional entropy calculation model, calculate the primary entropy value and generate a multi-dimensional entropy feature vector; by dynamically adjusting the entropy weights through a sliding window variance detector, calculate the weighted comprehensive entropy feature vector.

[0122] S21: Deploy edge computing devices near the data source and input the collected pipeline data into the edge processing module.

[0123] Furthermore, the edge processing module includes a multidimensional entropy calculation unit and a feature entropy processing unit.

[0124] Specifically, edge computing devices are deployed at key nodes (such as water-intensive plants, chemical plants, and food processing plants) to aggregate and perform preliminary processing of sensor data. Each edge node receives raw dataset X0 containing water pressure, flow rate, and other parameters from nearby processes (such as power plant A, chemical plant B, and food processing plant C).

[0125] S22: Construct a multidimensional entropy calculation model within the multidimensional entropy calculation unit, calculate the entropy features and fuse them with the time series to form a multidimensional entropy feature vector X under the time series. Q,t (H d,t H s,t H z,t ).

[0126] Furthermore, the multidimensional entropy calculation model includes three entropy calculation methods: multi-scale permutation entropy, time-frequency domain wavelet entropy, and adaptive fuzzy entropy, which are used to calculate entropy features at different scales and frequency bands.

[0127] (1) Multiscale permutation entropy is a nonlinear time series analysis method that decomposes a time series into multiple time-domain scales, calculates the permutation entropy value at each scale, and combines them to obtain the complexity characteristics of the entire time series. Specifically, the calculation steps of multiscale permutation entropy include:

[0128] First, the time series is processed by normalization and noise reduction. Then, wavelet decomposition is performed to obtain approximate coefficient sequences at different scales. These approximate sequences are then arranged to obtain permutation sequences at different scales. The permutation entropy value at each scale is calculated, and the permutation entropy values ​​at different scales are combined to obtain the multi-scale permutation entropy value of the entire sequence. The relevant formulas and principles are as follows:

[0129]

[0130] Where m is the dimension, H d Let be the multiscale permutation entropy in dimension m. Let π be the permutation pattern. Let p(π) be the probability of the permutation entropy occurring.

[0131] (2) Time-frequency domain wavelet entropy combines the time-frequency localization characteristic of wavelet analysis with the concept of entropy, enabling it to better describe the complexity and randomness of signals at different scales. Specifically, the calculation steps for time-frequency domain wavelet entropy include:

[0132] First, the signal is decomposed into wavelet coefficients to obtain a series of wavelet coefficients. The probability distribution of each wavelet coefficient is calculated, and finally the entropy value is calculated based on the probability distribution.

[0133]

[0134] Among them, H s Let be the wavelet entropy in the time-frequency domain, p(k) be the energy probability of the k-th subband, and k be the number of subband sequences. j be the number of sample sequences.

[0135] (3) Adaptive fuzzy entropy, combining fuzzy membership functions and adaptive parameter adjustment, measures the probability of a time series generating new patterns as its dimension changes. The higher the probability of a sequence generating new patterns, the higher the complexity of the sequence. Specifically, the calculation steps for adaptive fuzzy entropy include:

[0136]

[0137] Where i and j are the number of sequences from different samples, m is the dimension, r is the adaptive similarity tolerance threshold, and n is the similarity tolerance boundary gradient. N is the total number of time series, and D is the fuzzy similarity metric. H z It is an adaptive fuzzy entropy.

[0138] S23: A sliding window variance detector is used to monitor entropy changes in real time, and the fusion weights are dynamically adjusted based on variance to output a comprehensive entropy feature vector H in the time series. c,t The specific steps include:

[0139] S231: Set the window length M = 50, and extract the entropy value sequence of the 50 most recent time points at time t based on the sliding window method;

[0140] Specifically, select a starting position 't', initialize the window size, and calculate the initial entropy value. Move the window forward one position and update the entropy value within the window. Repeat this process 50 times to obtain a sequence of 50 entropy values.

[0141] S232: Calculate the variance of each type of entropy value within the window to reflect the degree of dispersion of each entropy value;

[0142] Calculate the mean of each type of entropy value, and then calculate the variance for each. The larger the variance, the greater the dispersion of the entropy values. The relevant formulas are as follows:

[0143]

[0144] Where B is the number of windows, H is the entropy value, and V is the variance of the corresponding entropy value.

[0145] When the operating mode changes (such as when cleaning is started), the variance of isoentropy values ​​such as flow rate and pressure will increase significantly.

[0146] S233: Calculate the weights of each entropy feature dynamically based on variance;

[0147] Specifically, Softmax weighting is used to amplify variance differences. By adjusting the temperature coefficient, the weight distribution of each entropy value is dynamically controlled, thus amplifying the impact of abnormal changes. The relevant formulas are as follows:

[0148]

[0149] Where δ aLet Γ be the variance of the a-th feature entropy value, Γ be the temperature coefficient controlling the weight distribution, and δ be the variance of the a-th feature entropy value. l Let ω be the variance of the l-th feature entropy value, A be the total variance of the feature entropy values, and ω be the variance of the l-th feature entropy value. a is the weight of the entropy value of the a-th feature.

[0150] S234: After normalizing the three types of entropy values ​​respectively, perform weighted fusion based on their respective weights to output the entropy feature vector H of the time series. c,t This is used to represent the comprehensive characteristics of the system's current operating state. The underlying formula is as follows:

[0151]

[0152] in, The normalized multi-scale permutation entropy takes values ​​in the range [0,1]. ln(m!) represents the theoretical maximum value of the permutation entropy.

[0153]

[0154] in, Let L be the normalized time-frequency wavelet entropy, and L be the total number of subbands.

[0155]

[0156] in, For normalized adaptive fuzzy entropy, ln(m!) represents the theoretical maximum value of permutation entropy.

[0157]

[0158] Where H d,s,z The comprehensive feature entropy value, ω d ω s ω z These are the weights for multi-scale permutation entropy, time-frequency domain wavelet entropy, and adaptive fuzzy entropy, respectively. Q,t (,H s,t H z,t )

[0159] Specifically, taking power plant A as an example, the normalized multi-scale permutation entropy, normalized time-frequency domain wavelet entropy, and normalized adaptive fuzzy entropy at times t1, t2, and t3 in the time series are calculated respectively, and the comprehensive entropy value features are calculated based on the obtained weights.

[0160] Taking time t1 as an example, the values ​​of the normalized multi-scale permutation entropy, the normalized time-frequency wavelet entropy, and the normalized adaptive fuzzy entropy are, respectively, H. d,t1 =0.78, H s,t1 =0.84, H z,t1 =0.76; the weights of the three are ω in order. d,t1=0.35, ω s,t1 =0.45, ω z,t1 =0.2, then the comprehensive feature entropy value H c,t1 =0.803. The calculation process for times t2 and t3 is the same as that for time t1, and finally the comprehensive entropy characteristics at times t1, t2 and t3 can be obtained.

[0161] <![CDATA[H d ]]> <![CDATA[H s ]]> <![CDATA[H z ]]> <![CDATA[ω d ]]> <![CDATA[ω s ]]> <![CDATA[ω z ]]> <![CDATA[H d,s,z ]]> <![CDATA[Time t1]]> 0.78 0.84 0.76 0.35 0.45 0.20 0.803 <![CDATA[Time t2]]> 0.76 0.85 0.72 0.32 0.50 0.18 0.798 <![CDATA[Time t3]]> 0.73 0.82 0.78 0.30 0.48 0.22 0.784

[0162] S24: Principal component analysis is used to reduce the dimensionality and compress the entropy eigenvectors, and the output is sent to memory to reduce the storage pressure on subsequent chains.

[0163] Step 3: Based on the on-chain-off-chain collaborative verification architecture, on-chain smart contracts perform cross-node feature consistency verification; design a trusted feature library self-updating mechanism, automatically updating the template library when a new feature is verified as a trusted feature; abnormal features trigger alarms, and multi-party signatures are used for evidence storage for audit traceability; based on the random forest model, different working conditions are associated with entropy features to build a working condition identification module.

[0164] S31: Off-chain edge nodes upload the output feature entropy vector to the blockchain network.

[0165] S32: On-chain smart contracts execute blockchain consensus verification, comparing entropy feature vectors submitted by multiple edge nodes to perform cross-node feature consistency verification.

[0166] Specifically, the steps for executing a smart contract include:

[0167] S321: The on-chain smart contract receives entropy feature data from multiple edge nodes, calculates the cosine similarity between each node, and selects node pairs with similarity higher than the threshold T = 0.85 to form a set C;

[0168] Specifically, for the entropy feature vectors of J edge nodes, the cosine similarity between any two nodes p and q is calculated using the following formula:

[0169]

[0170] Among them, S p,q It is the similarity between nodes p and q; v p and v q Let v be the entropy eigenvectors of nodes p and q; p ‖ and ‖v q || represents vector v p and v q The Euclidean norm,

[0171] C={q≠p|S p,q ≥T}

[0172] Where T is the similarity threshold, which is set to 0.85 here.

[0173] S322: If the number of nodes in set C exceeds K=3, the smart contract performs weighted voting based on the dynamic weight mechanism of each node. The voting weight comprehensively considers factors such as the historical credibility of the node, real-time consistency and online stability. The smart contract calculates the weighted agreement rate and compares it with the consensus threshold α=0.85. When the agreement rate exceeds the threshold, the entropy value feature is determined to be a trustworthy feature and the trustworthy feature library self-update mechanism is triggered.

[0174] Furthermore, the dynamic weighted voting mechanism includes the following steps:

[0175] (1) Set weighting factors and calculate the overall weight.

[0176] a. Historical credibility weight w1:

[0177]

[0178] Where P1 represents the percentage of node q that has historically participated in consensus and passed the judgment.

[0179] b. Real-time consistency weight w2:

[0180]

[0181] in, Let q be the average similarity between node q and set C.

[0182] c. Online stability weight w3:

[0183]

[0184] Wherein, P2 is the proportion of time that node q remains online and reports data on time during the monitoring period.

[0185] For a set C of highly similar node pairs, calculate the voting weight w of each node. q And after normalization, we get w′ q The specific calculation formula is as follows:

[0186] w q =w1+w2+w3

[0187]

[0188] (2) Weighted voting and consensus determination

[0189] The smart contract calculates the average similarity of each node. Automatically assign "agree" or "disagree" and calculate the weighted agreement rate R, using the following formula:

[0190]

[0191] When R exceeds the consensus rate threshold α = 0.85, the node's characteristics are determined to be trustworthy, triggering the trustworthy characteristic database self-update mechanism.

[0192] S323: If the similarity between a node feature and any of the three nodes (K=3 or less) is less than the threshold of 0.85, it is determined to be an abnormal feature and the anomaly detection mechanism is triggered.

[0193] Specifically, taking the example of an on-chain smart contract receiving entropy feature vectors from four edge nodes A, B, C, D, and E, V A = [1.0, 0.00], V B = [0.90, 0.20], V C = [0.87, 0.30], V D =[0.86,0.50]、V D = [0.88, 0.40]; Using node A as the reference, calculate the cosine similarity of each pair of nodes, S A,B =0.976, S A,C =0.946, S A,D =0.864, S A,E =0.91; Select all nodes with a similarity higher than 0.85 to node A to form a set C, C = {B, C, D, E}, |C| = 4 > K (= 3); For each node in set C, calculate its similarity with the other three nodes and take the average.

[0194] For each node, its three weights are calculated, and the historical pass rate P1 and online rate P2 of each node are set as follows:

[0195] node P1 P2 B 95％ 98％ C 80％ 96％ D 65％ 89％ E 92％ 99％

[0196] Substituting the values, we obtain the three weights for each node as shown in the table below:

[0197] node <![CDATA[w1]]> <![CDATA[w2]]> <![CDATA[w3]]> <![CDATA[w′ q ]]> B 1.2 1.831 1.2 4.231 C 1.0 1.938 1.0 3.938 D 0.8 1.842 0.8 3.442 E 1.2 1.927 1.2 4.327

[0198] The three weights of each node are summed and normalized, ∑w q =15.938, ∑w′ B =0.265, ∑w′ C =0.247, ∑w′ D =0.216, ∑w′ E =0.271.

[0199] If and only if the average similarity of nodes vote in favor q=1. All here Since all values ​​are ≥0.85, all votes are "Agree". Therefore, R = ∑ q∈C w′ q If ×1=1.0≥α, the contract determines that the feature set is “trustworthy” and triggers the trustworthy feature library self-update mechanism.

[0200] S33: Establish a self-updating mechanism for the trusted feature library: write features that meet the criteria for trusted feature determination into the trusted feature library, generate a new version of the feature template, and record it on the blockchain as a benchmark for subsequent comparisons.

[0201] S34: Establish an anomaly detection mechanism: Store the anomaly features collected by the anomaly detection mechanism into the blockchain evidence repository, and use multi-party signatures for evidence storage to ensure data traceability.

[0202] S35: Based on the random forest model, different operating conditions are associated with entropy features to establish an operating condition identification module. Specific steps include:

[0203] S351: Extract entropy feature vectors that have passed consistency verification from the "Trusted Feature Library" and label them as "Normal Production" or "Equipment Maintenance" and other operating conditions; extract marked abnormal entropy feature vectors from the "Abnormal Evidence Library" and label them as "Abnormal Leakage" or "Sensor Failure" and other abnormal operating conditions.

[0204] S352: The above entropy feature vector and its working conditions are used as the training set, input into the random forest model to train the classification model, and the model parameters are adjusted through cross-validation.

[0205] S353: Convert the multi-condition output of the classification model into a probability distribution matrix, and set corresponding probability thresholds to represent the probability of each condition. When the predicted probability exceeds the threshold, the system is considered to be in that condition.

[0206] Step 4: Call the working condition identification module to analyze the real-time entropy value characteristics, calculate the water entropy value of each stage of the production chain, and realize the unique recording of water footprint certificates through blockchain smart contracts; construct a dynamic path map of water footprint based on the production timeline, entropy intensity, and water resource flow path, and build a reverse traceability mechanism to return the complete water footprint chain by querying through product ID.

[0207] S41: When a product enters a critical process (such as cleaning, cooling, or other processes with high water intensity or sensitive changes), the system first calls the entropy feature analysis module deployed in the off-chain edge device to extract and process the entropy feature vector of the current process in real time. This feature vector is then input into the deployed random forest classification model, outputting the current operating condition label. This operating condition label serves as one of the conditions in the water resource consumption calculation formula, used to dynamically adjust the calculation parameters, thereby accurately estimating the water resource consumption increment of this process.

[0208] S42: Analyze the current entropy characteristics through smart contracts and update the trusted feature library in real time; calculate and accumulate the water footprint increment of the product in this stage and update the water footprint; write the water footprint attributes into the blockchain to generate a unique water footprint certificate to ensure that it cannot be tampered with.

[0209] The formula for calculating the cumulative increase in water footprint is as follows:

[0210] WF = WF total +WF physical +γ×WF stage

[0211] Among them, WF total Cumulative water footprint associated with product ID; WF physical For direct water use in the current stage; WF stage For the current stage of water resource consumption, WF stage =‖v stage ||;||v stage ‖ represents the current water resource entropy feature vector v stage The Euclidean norm; γ is a dynamically adjusted parameter for different operating conditions.

[0212] Specifically, the water footprint certificate includes basic information (such as certificate ID, product ID, process information, and production time), water footprint data (such as details of water resource consumption at each production stage, direct water consumption at each stage, and the updated cumulative total water footprint), entropy characteristics and operating condition information, reliable evidence storage information, and associated anomaly records.

[0213] S43: Based on the production timeline (X-axis), entropy intensity (Y-axis), and water resource flow path (Z-axis), a 3D topology map is drawn in real time, dynamically displaying the flow path and intensity changes of the water footprint.

[0214] Specifically, the X-axis (production time axis) can use the absolute timestamp of each process completion or normalize the overall production progress to 0–100% percentage; the Y-axis (entropy intensity) takes the entropy norm corresponding to each process, with a larger value indicating a higher degree of system fluctuation or anomaly; the Z-axis (water resource flow path) projects each process node at equal distances and connects them with directed lines. The width of the lines is scaled according to the actual water flow rate. The direction of the lines along the +Z-axis represents forward production flow, and along the -Z-axis represents back circulation or water recycling flow. Users can rotate, scale, or pan the graph to comprehensively view the water footprint flow direction and fluctuation intensity between different processes.

[0215] Figure 2This is a 3D topological diagram illustrating the flow path and intensity changes of water footprint, showing the changes in water footprint at each production node during the production of a certain product. The bottom horizontal axis represents the production time axis, the bottom vertical axis represents the entropy intensity axis, and the axis perpendicular to the bottom vertical axis represents the water resource flow path axis. The arrows indicate the direction of water flow, and the arrow width represents the actual amount of water used.

[0216] S44: Construct a reverse traceability mechanism: Query the blockchain through the final product ID, and the smart contract returns complete water footprint chain certificate information according to the on-chain records, including the entropy change curve of each process, water consumption details, and the signatures of the blockchain nodes that participated in the verification.

[0217] Specifically, the query process is as follows: each product is assigned a unique QR code, and consumers scan the code to send the product ID to the blockchain; the smart contract retrieves all water footprint credentials associated with that ID, including text, path maps, and entropy analysis reports.

[0218] Taking a plastic processing product with ID NA2024-001 as an example:

[0219] Product ID NA2024-001 Cumulative water footprint 3100L Maximum Entropy (Abnormal) Process Midsole injection molding (0.72) Optimal Entropy Process Assembly and packaging (0.29) Number of abnormal events 1 time (short-term abnormality in cooling water circulation) Last Update 2024-07-20 18:35UTC

[0220] Process time Place Water consumption Entropy refers to state Raw material glue extraction 2024-06-01 Rubber Plantation in City A 1500L 0.40 good Midsole injection molding 2024-06-15 Injection Molding Factory in City B 800L 0.72 warn Hot pressing 2024-07-10 C City Hot Press Center 500L 0.35 Stablize Assembly and packaging 2024-07-20 D City Assembly Center 300L 0.29 excellent

[0221] Figure 3 This is a schematic diagram of the structure of a production behavior water footprint tracking device based on hydraulic entropy characteristics and blockchain consensus verification provided in an embodiment of the present invention, as shown below. Figure 3 As shown, the device includes:

[0222] The system comprises a multi-data acquisition layer 201, an edge processing layer 202, a blockchain evidence storage layer 203, and an application interaction layer 204. The multi-data acquisition layer 201 controls a multimodal sensor group and a triple sampling controller to collect real-time data on pipeline water pressure, flow rate, and other characteristics at different time scales, providing raw input for feature entropy calculation. The edge processing layer 202 controls edge computing nodes to extract multi-dimensional entropy feature vectors through a multi-dimensional entropy calculation model and dynamically adjusts entropy weights using a sliding window variance detector to calculate weighted entropy features, reducing on-chain load. The blockchain evidence storage layer 203 implements cross-node data feature consistency verification and unique water footprint evidence storage. It records production water footprints through a water footprint certificate management contract and stores disputed data through an anomaly evidence storage contract. The application interaction layer 204 provides visual interaction and monitoring tools, constructing a dynamic water footprint path map based on the production timeline, entropy intensity, and water resource flow path, and building a reverse traceability mechanism to return the complete water footprint chain through product ID queries.

[0223] The apparatus embodiments provided in this invention are for implementing the above-described method embodiments. For specific processes and details, please refer to the above-described method embodiments, which will not be repeated here.

[0224] Figure 4 This is a schematic diagram of the physical structure of an electronic device provided in an embodiment of the present invention, such as... Figure 4 As shown, the electronic device may include: a processor 301, a communication interface 302, a memory 303, and a computer program stored in the memory 303 and configured to be executed by the processor 301. The communication interface 302 can be used for information transmission of the electronic device. The processor 301 can call the computer program in the memory 303, and when executing the computer program, implement the production behavior water footprint tracking method based on hydraulic entropy characteristics and blockchain consensus verification described in the above embodiments.

[0225] Preferably, the computer program can be divided into one or more modules (such as computer program 1, computer program 2, ...), and the one or more modules are stored in the memory 303 and executed by the processor 301. The one or more modules / units can be a series of computer program instruction segments capable of performing specific functions. These instruction segments are used to describe the execution process of the computer program in the terminal device, specifically including: calculating entropy features based on a multidimensional entropy calculation model and forming a multidimensional entropy feature vector under time series; using a sliding window variance detector to monitor entropy changes in real time, dynamically adjusting the fusion weights based on variance, and outputting a weighted entropy feature vector under time series; performing cross-node feature consistency verification based on on-chain smart contracts, recording production water footprints through a water footprint certificate management contract, and storing disputed data through an anomaly storage contract; constructing a dynamic path graph of water footprints based on the production timeline, entropy intensity, and water resource flow path, and constructing a reverse tracing mechanism to return the complete water footprint chain through product ID query.

[0226] The processor 301 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Alternatively, the processor 301 may be any conventional processor. The processor 301 is the control center of the terminal device and connects various parts of the terminal device using various interfaces and lines.

[0227] The memory 303 mainly includes a program storage area and a data storage area. The program storage area can store the operating system, at least one application program required for a given function, etc., while the data storage area can store related data, etc. Furthermore, the memory 303 can be a high-speed random access memory, or a non-volatile memory, such as a plug-in hard disk, smart memory card, flash memory card, etc., or the memory 303 can also be other volatile solid-state storage devices. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the above-described method embodiments of the present invention.

[0228] This invention also provides a computer-readable storage medium, which includes a stored computer program. When the computer program is executed, it controls the device where the computer-readable storage medium is located to perform the production behavior water footprint tracking method based on hydraulic entropy characteristics and blockchain consensus verification described in the above embodiments.

[0229] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0230] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of various embodiments or some parts of embodiments.

Claims

1. A method for tracking the water footprint of production activities based on hydraulic entropy characteristics and blockchain consensus verification, characterized in that, The method includes the following steps: Step 1: Install high-precision sensors at key locations within the water-using unit. Through a triple sampling sequence processing layer (high-frequency, medium-frequency, and low-frequency), collect real-time data on pipeline water pressure, flow rate, and velocity at different time scales. Step 2: Based on the on-chain-off-chain collaborative verification architecture, design the entropy feature extraction function for off-chain edge computing nodes. This involves constructing a multi-dimensional entropy calculation model to calculate the initial entropy and generate a multi-dimensional entropy feature vector. A sliding window variance detector is then used to dynamically adjust the entropy weights and calculate the comprehensive entropy feature vector. Step 3: Based on the on-chain-off-chain collaborative verification architecture, on-chain smart contracts perform cross-node feature consistency verification; a trusted feature library self-updating mechanism is designed, automatically updating the template library when a new feature is verified as a trusted feature; abnormal features trigger alarms, and multi-party signatures are used for evidence storage for audit traceability; based on the random forest model, different operating conditions are associated with entropy features to construct an operating condition identification module. Step 4: Call the working condition identification module to analyze the real-time entropy value characteristics, calculate the water entropy value of each stage of the production chain, and realize the unique recording of water footprint certificates through blockchain smart contracts; construct a dynamic path map of water footprint based on the production timeline, entropy intensity, and water resource flow path, and build a reverse traceability mechanism to return the complete water footprint chain by querying through product ID.

2. The method for tracking the water footprint of production behavior based on hydraulic entropy characteristics and blockchain consensus verification according to claim 1, characterized in that, Step 1 is as follows: S11: High-precision sensors are installed at key locations in industrial water pipes and equipment inlets to collect data on water pressure, flow rate, and flow velocity in the pipes, and simultaneously collect electrical pulse signals from the equipment. S12: Design a triple sampling sequence processing layer. Through high-frequency, medium-frequency and low-frequency triple sampling, the characteristics of water pressure and flow data at different time scales are captured in real time. The high-frequency sampling layer captures the transient fluctuations of water pressure and flow data. The sampling interval is short, and it captures sudden changes in water flow caused by valve opening and closing and equipment start and stop. The intermediate frequency sampling layer captures the characteristics of the operation cycle, with a moderate sampling interval, which is used to identify the periodic operation of equipment cleaning and cooling cycles; The low-frequency sampling layer captures long-term trends in water pressure and flow data, with a relatively long sampling interval, enabling monitoring of water consumption trends during equipment operation and product production throughout the day.

3. The method for tracking the water footprint of production behavior based on hydraulic entropy characteristics and blockchain consensus verification according to claim 1, characterized in that, Step 2 is as follows: S21: Deploy edge computing devices near the data source to input the collected pipeline water pressure and flow data into the edge processing module. The edge processing module includes a multi-dimensional entropy calculation unit and a feature entropy processing unit. S22: Construct a multidimensional entropy calculation model within the multidimensional entropy calculation unit, calculate entropy features and fuse them with the time series to form a multidimensional entropy feature vector under the time series. The multidimensional entropy calculation model includes three entropy calculation methods: multi-scale permutation entropy, time-frequency domain wavelet entropy, and adaptive fuzzy entropy, used to calculate entropy features at different scales and frequency bands. (1) Multiscale permutation entropy is a nonlinear time series analysis method that decomposes a time series into multiple time-domain scales, calculates the permutation entropy value at each scale, and combines them to obtain the complexity characteristics of the entire time series. Specifically, the calculation steps of multiscale permutation entropy include: First, the time series is normalized and denoised. Then, wavelet decomposition is performed to obtain approximate coefficient sequences at different scales. These approximate sequences are then arranged to obtain permutation sequences at different scales. The permutation entropy value at each scale is calculated, and the permutation entropy values ​​at different scales are combined to obtain the multi-scale permutation entropy value of the entire sequence. The relevant formulas and principles are as follows: in, For dimension, dimension Multiscale permutation entropy, For the arrangement pattern, The probability of the permutation entropy occurring. (2) Time-frequency domain wavelet entropy combines the time-frequency localization characteristics of wavelet analysis with the concept of entropy, enabling it to better describe the complexity and randomness of signals at different scales. Specifically, the calculation steps of time-frequency domain wavelet entropy include: First, the signal is decomposed using wavelet decomposition to obtain a series of wavelet coefficients. The probability distribution of each wavelet coefficient is then calculated, and finally, the entropy value is calculated based on the probability distribution. in, For wavelet entropy in the time-frequency domain, For the first The energy probability of each element The number of sub-band sequences. The number of sample sequences. (3) Adaptive fuzzy entropy, combined with fuzzy membership function and adaptive parameter adjustment, measures the probability of a time series generating a new pattern when the dimension changes. The higher the probability of the sequence generating a new pattern, the higher the complexity of the sequence. The calculation steps of adaptive fuzzy entropy include: in, , For the number of sequences in different samples, For dimension, To adapt the similarity tolerance threshold, For similarity tolerance boundary gradient, The total number of time series, For fuzzy similarity measurement, For adaptive fuzzy entropy, S23: A sliding window variance detector is used to monitor entropy changes in real time. The fusion weights are dynamically adjusted based on the variance, and the comprehensive entropy feature vector of the time series is output. The specific steps include: S231: Extract the entropy value sequence of the M most recent time points at time t based on the sliding window method; specifically, select the starting position t, initialize the window size, calculate the initial entropy value, move the window forward by one position, update the entropy value within the window, repeat the operation M times, and finally obtain the M entropy value sequence. S232: Calculate the variance of each type of entropy value within the window, reflecting the degree of dispersion of each entropy value; calculate the mean of each type of entropy value separately, and calculate the variance separately. The larger the variance, the greater the dispersion of the entropy value. The relevant formulas are as follows: in, For the number of windows, The entropy value. This corresponds to the mean entropy value. S233: Calculate the weights of each entropy feature dynamically based on variance; Specifically, Softmax weighting is used to amplify variance differences. By adjusting the temperature coefficient, the weight distribution of each entropy value is dynamically controlled. The relevant formulas are as follows: in For the first The variance of the entropy values ​​of each feature To control the temperature coefficient of the weight distribution, For the first The variance of the entropy values ​​of each feature The total variance of the feature entropy values. For the first The weights of each feature entropy value S234: After normalizing the three types of entropy values ​​respectively, a weighted fusion is performed based on their respective weights to output the comprehensive entropy feature vector of the time series. The formula and principle involved are as follows: in, The normalized multi-scale permutation entropy takes values ​​in the range [0,1]. This represents the theoretical maximum value of the permutation entropy. in, To normalize the time-frequency domain wavelet entropy, For the total number of sub-bands, in, To normalize adaptive fuzzy entropy, This represents the theoretical maximum value of the permutation entropy. in The entropy value is the comprehensive feature value. , , These are the weights for multi-scale permutation entropy, time-frequency domain wavelet entropy, and adaptive fuzzy entropy, respectively. S24: The dimensionality reduction and compression of the comprehensive entropy value eigenvector are performed by principal component analysis and output to memory to reduce the subsequent on-chain storage pressure.

4. The method for tracking the water footprint of production behavior based on hydraulic entropy characteristics and blockchain consensus verification according to claim 1, characterized in that, Step 3 is as follows: S31: Off-chain edge nodes upload the output feature entropy vector to the blockchain network. S32: On-chain smart contracts execute blockchain consensus verification, comparing the entropy feature vectors submitted by multiple edge nodes to perform cross-node feature consistency verification. The steps involved in running a smart contract include: S321: The on-chain smart contract receives entropy feature data from multiple edge nodes, calculates the cosine similarity between nodes, and selects highly similar node pairs to form a set. ; Specifically, for the entropy feature vectors of J edge nodes, any two nodes and The formula for calculating cosine similarity is as follows: in, It is a node and The similarity between them; and For nodes and The entropy value eigenvector; and For vectors and The Euclidean norm, in, The similarity threshold is set between 0.8 and 0.

9. S322: If set If the number of nodes exceeds K, the smart contract will conduct weighted voting based on the dynamic weight mechanism of each node. If the voting results reach a consensus, it will be determined as a trustworthy feature and trigger the self-update mechanism of the trustworthy feature library. The dynamic weighted voting mechanism includes the following steps: (1) Set weighting factors to calculate the overall weight. a. Historical credibility weight : in, For nodes The proportion of historical participation in the consensus-building process. b. Real-time consistency weight : in, For nodes With sets The average similarity, The similarity threshold is 55. c. Online stability weights : in, For nodes The percentage of time during which data is kept online and reported on time during the monitoring period. For a set of highly similar node pairs Calculate the voting weight of each node. And after normalization, we obtain The specific calculation formula is as follows: (2) Weighted voting and consensus determination The smart contract calculates the average similarity of each node. It automatically assigns "agree" or "disagree" and calculates the weighted agreement rate. The calculation formula is as follows: when Exceeding the consensus rate threshold If the node's feature is deemed trustworthy, the trustworthy feature database self-update mechanism is triggered. S323: If a node has characteristics The similarity with most nodes is less than the threshold. If this is detected, it is considered an abnormal feature, and the anomaly detection mechanism is triggered. S33: Establish a self-updating mechanism for the trusted feature library: Features that meet the criteria for trusted feature determination are written into the trusted feature library, a new version of the feature template is generated, and it is recorded on the blockchain as a comparison benchmark for subsequent consistency verification. S34: Establish an anomaly detection mechanism: Store the anomaly features collected by the anomaly detection mechanism in a blockchain evidence repository, and use multi-party signatures for evidence storage to ensure data traceability. S35: Based on the random forest model, different working conditions are associated with entropy features to establish a working condition identification module. Specific steps include: S351: Extract the historical entropy feature vector and label the corresponding working conditions; Specifically, different operating conditions include normal production, equipment maintenance, and abnormal leakage. S352: Entropy feature vectors and their operating conditions are used as training sets, input into a random forest model to train a classification model, and the model parameters are adjusted through cross-validation. S353: Convert the multi-condition output of the classification model into a probability distribution matrix, and set corresponding probability thresholds to represent the probability of each condition. When the predicted probability exceeds the threshold, the system is considered to be in that condition.

5. The method for tracking the water footprint of production behavior based on hydraulic entropy characteristics and blockchain consensus verification according to claim 1, characterized in that, Step 4 is as follows: S41: When a product enters a critical process, the system first calls the entropy feature calculation module to extract and process the entropy feature vector, and then calls the operating condition identification module to calculate the water entropy value for this stage based on the actual water consumption data of that process. S42: Analyze the current entropy characteristics through smart contracts and update the trusted feature database in real time; calculate and accumulate the water footprint increment of the product in this stage, and update the water footprint; write the water footprint attributes into the blockchain to generate a unique water footprint certificate, ensuring immutability. The formula for calculating the cumulative increase in water footprint is as follows: in, The cumulative water footprint associated with the product ID; For direct water use in the current stage; This represents the current water consumption level. ; The current water resource entropy feature vector The Euclidean norm; To dynamically adjust parameters for different operating conditions, The water footprint certificate includes basic information, water footprint data, entropy characteristics and operating condition information, reliable evidence storage information, and associated anomaly records. S43: Based on the production timeline (X-axis), entropy intensity (Y-axis), and water flow path (Z-axis), a 3D topology map is drawn in real time, dynamically displaying the flow path and intensity changes of the water footprint. In this system, the X-axis represents the process completion time or production progress percentage, the Y-axis reflects the oscillation intensity of the operating state using the entropy norm, and the Z-axis arranges process nodes at equal intervals and uses directed lines to indicate the water flow direction and flow rate. S44: Construct a reverse traceability mechanism: Query the blockchain through the final product ID, and the smart contract returns complete water footprint chain certificate information according to the on-chain records, including the entropy change curve of each process, water consumption details, and the signatures of the blockchain nodes that participated in the verification.

6. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the production behavior water footprint tracking method based on hydraulic entropy characteristics and blockchain consensus verification as described in any one of claims 1 to 5.

7. A computer-readable storage medium storing computer instructions thereon, characterized in that, When the computer instruction is executed by the processor, it implements a production behavior water footprint tracking method based on hydraulic entropy characteristics and blockchain consensus verification as described in any one of claims 1-5.

Images