A U→A recursive tensor uncertainty transformation system and method based on energy conservation

Abstract

This invention provides a U→A recursive tensor uncertainty transformation system and method based on energy conservation. The system includes a field sensing module, a potential energy accumulation module, a recursive tensor module, an A-state triggering module, an energy conservation verification and self-calibration module, and a state output module. Uncertainty (U-state) is treated as accumulable potential energy. A recursive tensor machine is used to model the historical potential energy sequence. When the accumulated potential energy exceeds a critical threshold, a state transition is triggered, outputting a deterministic action (A-state). The energy conservation verification and self-calibration module ensures that the decision-making process conforms to physical laws. This invention introduces the principle of energy conservation into uncertainty decision-making systems, offering advantages such as strong interpretability, recursive memory, cross-domain applicability, and self-calibration. It can be widely applied in fields such as financial quantitative trading, traditional Chinese medicine auxiliary diagnosis, and industrial sensor networks.

Description

Technical Field

[0001] This invention relates to the fields of artificial intelligence and automatic control technology, and in particular to a recursive tensor system and method for realizing the transformation from the U state (uncertain state) to the A state (abnormal evolution state) in a five-state dynamic logic system of -1, 0, 1, U, and A. It can be widely applied in the fields of industrial automation control, robot path planning, intelligent sensor networks, multi-source information fusion, financial decision support, and medical auxiliary diagnosis. Background Technology

[0002] In complex dynamic systems, the core challenge for decision-makers is to identify the true state of the system and respond effectively when information is incomplete and the state is uncertain. Existing technologies have the following problems: 1. Treating uncertainty as noise and filtering it directly. Traditional control systems use filtering algorithms (such as Kalman filtering and particle filtering) to filter out uncertain signals as noise, thus losing valuable information contained in the uncertainty. Experiments show that there is a significant correlation (p<0.01) between the intensity, frequency, and duration of uncertain signals and the probability of system state transitions. Direct filtering will lead to the loss of the system's ability to predict state transitions.

[0003] 2. Lack of energy conservation constraints. Existing decision-making systems (including expert systems, fuzzy logic systems, and neural network systems) do not adhere to the principle of energy conservation, resulting in state transitions that do not conform to physical laws. For example, a system cannot spontaneously transition from a stable state to an unstable state without any external input; existing systems cannot constrain such non-physical state transitions.

[0004] 3. The decision-making process is unexplainable. Deep learning models are typical "black box" systems, whose decision-making processes are untraceable and unexplainable. In fields such as industrial control and medical diagnosis, where interpretability is highly critical, the deployment of black box models is severely limited.

[0005] 4. Lack of recursive memory. Most existing systems make decisions based on the current state and cannot "remember" the accumulation process of historical uncertainty. However, a large amount of experimental data shows that the probability of system state transitions is strongly correlated with the historical uncertainty sequence (correlation coefficient r>0.7), and the lack of recursive memory will lead to a serious lag in judging the timing of transitions.

[0006] 5. Patent application No. 2025115881700, entitled "A Self-Evolving Computation System Based on Dynamic Logic and Recursive Tensors", proposed a five-state dynamic logic framework of -1, 0, 1, U, and A, and defined the five-state transformation rules and self-evolution mechanism. However, it did not specifically disclose the recursive tensor calculation method for the transformation from the U state (uncertain state) to the A state (abnormal evolution state), which limited the practical application of the five-state logic system in engineering.

[0007] Therefore, there is an urgent need for a U→A recursive tensor system and method that can explicitly model the "uncertainty → determinism" transformation process, obey the principle of energy conservation, and has interpretability and recursive memory capabilities. Summary of the Invention

[0008] This invention overcomes the shortcomings of the prior art and provides a U→A recursive tensor uncertainty transformation system and method based on energy conservation.

[0009] The object of this invention is achieved in the following manner: The uncertainty transformation system in this invention is a dynamic logic system with a set of logic values ​​of {-1, 0, 1, U, A}, which consists of a first deterministic state (-1), a second deterministic state (0), a third deterministic state (+1), an uncertainty state (U), and an abnormal evolutionary state (A).

[0010] U-state (uncertainty state): one of the five states, representing the critical state of the system, where potential energy continues to accumulate but has not yet reached the trigger threshold.

[0011] A-state (abnormal evolutionary state): one of the five states, representing a deterministic action state after potential energy accumulation exceeds a threshold.

[0012] U→A Recursive Tensor Machine: The core engine for U→A transformation.

[0013] Energy conservation constraint: α+β+γ=1, ensuring the balance between the system's input energy and output energy.

[0014] A U→A recursive tensor uncertainty transformation system based on energy conservation includes: The field sensing module is configured to receive external uncertain U-state input signals and calculate the field vibration intensity F(t); The potential energy accumulation module is configured to convert the input energy into accumulable potential energy P(t) according to the field vibration intensity F(t) and the law of conservation of energy. The recursive tensor module is configured to perform deep recursive learning on historical potential energy sequences to extract deep features of uncertainty evolution. The A-state triggering module is configured to monitor the accumulated potential energy P(t). When P(t) reaches or exceeds the preset triggering threshold, a deterministic A-state confirmation is triggered. The energy conservation verification and self-calibration module is configured to monitor the balance of system input energy, output energy and dissipated energy, and automatically adjust the energy distribution coefficient when the energy deviation exceeds the preset tolerance. The priority of this calibration module is higher than that of the A-state triggering module. The status output module is configured to output system status transition instructions based on the A-state trigger signal.

[0015] Further: 1. Field sensing module (U-state input unit) Receive an externally uncertain U-state input signal and calculate the field vibration intensity F(t): F(t) = Σ(W_i × S_i(t)) / Σ(W_i) Where S_i(t) is the signal strength of the i-th channel at time t, including physical signals such as sensor data and measured values; event signals such as state changes and triggering events; and environmental signals such as temperature, humidity, and vibration. W_i is the weighting coefficient of the i-th channel.

[0016] The U-state input signal includes at least one or more of the following fields: financial time series data, physiological time series data, sensor network data, and robot kinematics data; the system is different from a hardware reconfigurable computing system, its reconfiguration is performed at the decision logic layer rather than the hardware circuit layer, the reconfiguration trigger condition is the system-level contradiction caused by U-state exploration, and the reconfiguration goal is the dimensional leap of the system's cognitive ability.

[0017] 2. Potential Energy Accumulation Module (Energy Conversion Prediction Unit) The potential energy accumulation module models historical potential energy sequences using a recursive tensor machine to predict potential energy accumulation trends. Its core energy conversion formula is: P(t) = α•F(t) + β•Σ_{i=1}^{n} (λ^{i} × F(ti)) - γ•D(t) in: P(t) represents the amount of potential energy accumulated at the current moment; F(t) represents the field vibration intensity at the current moment; Σ_{i=1}^{n} (λ^{i} × F(ti)) is the recursive weighted sum of the historical potential energy sequence, λ is the decay factor, 0<λ<1, and n is the recursive memory depth; D(t) is the energy dissipation term; α, β, and γ are energy distribution coefficients that satisfy α+β+γ=1 (energy conservation constraint).

[0018] 3. Recursive Tensor Module Using an LSTM or GRU network structure, the input of this module is the potential energy accumulation sequence P(tn) to P(t), and the output is the potential energy breakthrough time window and the prediction of the conversion direction. It is also configured with an online learning mode: when the prediction error of N consecutive U→A conversions exceeds the preset threshold, the system automatically expands the hidden layer dimension H or increases the recursion depth L until the prediction error drops below the threshold.

[0019] 4. A-state triggering module (dynamic release unit) It is equipped with a conversion direction selection unit and a hysteresis comparator: when the potential energy accumulation P(t) ≥ P_threshold_up, the forward conversion of state A is triggered U→A+ (recommended value 0.70~0.85); when P(t) ≤ P_threshold_down, the reverse conversion of state A is triggered U→A-; a hysteresis interval ΔP_hyst is set between the upper and lower thresholds to prevent frequent conversions at the threshold boundaries.

[0020] 5. Energy Conservation Verification and Self-Calibration Module Monitor the system's input energy E_in, output energy E_out, and dissipated energy E_dissipation to verify the energy conservation constraint: AE = E_in - E_out - E_dissipation in: E_in = ΣF(t), the total input energy; E_out = ΣΔE(release), outputting the total energy; E_dissipation = ΣD(t), the total energy dissipated; When |AE| exceeds the preset tolerance ε (recommended value 0.05~0.10), the system automatically triggers self-calibration, adjusting the energy distribution coefficients α, β, and γ to meet the constraint α+β+γ=1, ensuring that the energy consumption of the system is minimized during long-term operation. ε is the threshold of the system's "self-calibration" mechanism, ensuring that the system does not deviate from the law of conservation of energy during long-term operation, and is a key engineering indicator regarding "self-evolution" capability.

[0021] 6. Status Output Module The system state transition instructions of the status output module include: The amount of potential energy accumulated at the time of triggering, P(t); Change direction, either forward or reverse; Suggested actions are specific control commands output based on the application scenario.

[0022] The system is configured for continuous self-operation: after a U→A conversion is completed, the field sensing module automatically resets and begins the next cycle of U-state monitoring without requiring an external reset signal; when the system detects an abnormality in the calculation of the potential energy accumulation P(t) or an energy deviation exceeding the preset tolerance 2ε, it automatically resets the recursive tensor module state and reinitializes the memory window n. When the absolute value of the energy deviation further deteriorates beyond 2ε, the system determines it to be a serious anomaly, thereby automatically resetting the recursive tensor module state and reinitializing the memory window.

[0023] The system has self-evolution capabilities, including a meta-learning module configured to monitor the commonalities and differences in historical U-state transition tasks. After the system completes K different types of U→A transition tasks within a preset time window, the meta-learning module updates the initial parameters of the recursive tensor module, thereby reducing the cold start prediction error of the system on new tasks.

[0024] The implementation method of the U→A recursive tensor uncertainty transformation system based on energy conservation includes the following steps: Step S1: Sensing external U-state input and calculating field vibration intensity F(t); Step S2: Convert the input energy into accumulable potential energy P(t) according to the law of conservation of energy; Step S3: Perform deep recursive learning on the historical potential energy sequence using a recursive tensor machine to predict the timing of potential energy breakout; Step S4: When the potential energy accumulation P(t) reaches or exceeds the preset trigger threshold, trigger state A confirmation and determine the conversion direction; Step S5: Monitor the balance of input energy, output energy and dissipated energy of the system, and automatically calibrate the energy distribution coefficient when the energy deviation exceeds the preset tolerance; Step S6: Output system state transition instructions; The computer has a readable storage medium on which a computer program is stored, which, when executed by a processor, implements all or part of the steps of the method.

[0025] The beneficial effects of this invention are: 1. Introducing the principle of energy conservation into uncertainty decision-making systems: using physical laws to constrain logical transformations, making the decision-making process interpretable; 2. Possesses recursive memory capability: Accurately captures state transition opportunities by modeling historical potential energy sequences using recursive tensor machines; 3. High interpretability: The energy conservation verification module records the energy flow path, and every decision has a physical basis; 4. Cross-domain applicability: It has been proven effective in three completely non-overlapping fields: financial quantitative trading, traditional Chinese medicine assisted diagnosis and treatment, and industrial sensor networks; 5. Self-calibration capability: Automatically calibrates the energy distribution coefficient during long-term operation to ensure minimal energy consumption. Attached Figure Description

[0026] Figure 1 This is a schematic diagram of the overall system architecture of the present invention; Figure 2 This is a schematic diagram of the U-state input and self-looping layout of the present invention; Figure 3 This is a schematic diagram of the recursive tensor network structure of the present invention (LSTM unit unfolded diagram); Figure 4 This is a flowchart illustrating the potential energy accumulation and A-state triggering process of this invention. Figure 5 This is a schematic diagram of the five-state transition of the present invention; Figure 6 This is a flowchart illustrating the energy conservation verification and self-calibration process of this invention. Detailed Implementation

[0027] A U→A recursive tensor uncertainty transformation system based on energy conservation includes: The field sensing module is configured to receive external uncertain U-state input signals and calculate the field vibration intensity F(t); The potential energy accumulation module is configured to convert the input energy into accumulable potential energy P(t) according to the field vibration intensity F(t) and the law of conservation of energy. The recursive tensor module is configured to perform deep recursive learning on historical potential energy sequences to extract deep features of uncertainty evolution. The A-state triggering module is configured to monitor the accumulated potential energy P(t). When P(t) reaches or exceeds the preset triggering threshold, a deterministic A-state confirmation is triggered. The energy conservation verification and self-calibration module is configured to monitor the balance of system input energy, output energy and dissipated energy, and automatically adjust the energy distribution coefficient when the energy deviation exceeds the preset tolerance. The priority of this calibration module is higher than that of the A-state triggering module. The status output module is configured to output system status transition instructions based on the A-state trigger signal.

[0028] Field sensing module: Receive an externally uncertain U-state input signal and calculate the field vibration intensity F(t): F(t) = Σ(W_i × S_i(t)) / Σ(W_i) Where S_i(t) is the signal strength of the i-th channel at time t, including physical signals such as sensor data and measured values; event signals such as state changes and triggering events; and environmental signals such as temperature, humidity, and vibration. W_i is the weighting coefficient of the i-th channel.

[0029] Potential energy accumulation module: By modeling historical potential energy sequences using a recursive tensor machine, the trend of potential energy accumulation is predicted. The core energy conversion formula is as follows: P(t) = α•F(t) + β•Σ_{i=1}^{n} (λ^{i} × F(ti)) - γ•D(t) in: P(t) represents the amount of potential energy accumulated at the current moment; F(t) represents the field vibration intensity at the current moment; Σ_{i=1}^{n} (λ^{i} × F(ti)) is the recursive weighted sum of the historical potential energy sequence, λ is the decay factor, 0<λ<1, and n is the recursive memory depth; D(t) is the energy dissipation term; α, β, and γ are energy distribution coefficients that satisfy α + β + γ = 1.

[0030] Recursive Tensor Module: Using an LSTM or GRU network structure, the input of this module is the potential energy accumulation sequence P(tn) to P(t), and the output is the potential energy breakthrough time window and the prediction of the conversion direction. It is also configured with an online learning mode: when the prediction error of N consecutive U→A conversions exceeds the preset threshold, the system automatically expands the hidden layer dimension H or increases the recursion depth L until the prediction error drops below the threshold.

[0031] A-state triggering module: It is equipped with a conversion direction selection unit and a hysteresis comparator: when the potential energy accumulation P(t) ≥ P_threshold_up, the forward conversion of state A is triggered U→A+; when P(t) ≤ P_threshold_down, the reverse conversion of state A is triggered U→A-; a hysteresis interval ΔP_hyst is set between the upper and lower thresholds to prevent frequent conversions at the threshold boundaries.

[0032] Energy conservation verification and self-calibration module: Configured to monitor the system's input energy E_in, output energy E_out, and dissipated energy E_dissipation, to verify the energy conservation constraint: AE = E_in - E_out - E_dissipation in: E_in = ΣF(t), the total input energy; E_out = ΣΔE(release), outputting the total energy; E_dissipation = ΣD(t), the total energy dissipated; When |AE| exceeds the preset tolerance ε, the system automatically triggers self-calibration, adjusting the energy distribution coefficients α, β, and γ to meet the constraint α+β+γ=1, ensuring that the energy consumption of the system is minimized during long-term operation.

[0033] Status output module: System state transition instructions include: The amount of potential energy accumulated at the time of triggering, P(t); Change direction, either forward or reverse; Suggested actions are specific control commands output based on the application scenario.

[0034] The system is configured for continuous self-running mode: after a U→A conversion is completed, the field sensing module automatically resets and starts the next cycle of U-state monitoring without the need for an external reset signal; when the system detects that the potential energy accumulation P(t) calculation is abnormal or the energy deviation exceeds the preset tolerance 2ε, it automatically resets the state of the recursive tensor module and reinitializes the memory window n.

[0035] The system has self-evolution capabilities, including a meta-learning module configured to monitor the commonalities and differences in historical U-state transition tasks. After the system completes K different types of U→A transition tasks within a preset time window, the meta-learning module updates the initial parameters of the recursive tensor module, thereby reducing the cold start prediction error of the system on new tasks.

[0036] The implementation method of the U→A recursive tensor uncertainty transformation system based on energy conservation includes the following steps: Step S1: Sensing external U-state input and calculating field vibration intensity F(t); Step S2: Convert the input energy into accumulable potential energy P(t) according to the law of conservation of energy; Step S3: Perform deep recursive learning on the historical potential energy sequence using a recursive tensor machine to predict the timing of potential energy breakout; Step S4: When the potential energy accumulation P(t) reaches or exceeds the preset trigger threshold, trigger state A confirmation and determine the conversion direction; Step S5: Monitor the balance of input energy, output energy and dissipated energy of the system, and automatically calibrate the energy distribution coefficient when the energy deviation exceeds the preset tolerance; Step S6: Output system state transition instructions; The computer has a readable storage medium on which a computer program is stored, which, when executed by a processor, implements all or part of the steps of the method.

[0037] Figure 1 The diagram shows the six modules and their data flow.

[0038] It contains six core modules, arranged from top to bottom: Module 1: Field Sensing Module (U-State Input Unit) Module 2: Potential Energy Accumulation Module (Energy Storage Unit) Module 3: Recursive Tensor Module (LSTM / GRU Network) Module 4: A-state triggering module (kinetic energy release unit) Module 5: Energy Conservation Verification and Self-Calibration Module Module 6: Status Output Module (Decision Instruction Output) Data flow direction: Solid arrow (main data flow): External U-state signal → Module 1 → Module 2 → Module 3 → Module 4 → Module 6 → Execution layer (adding to position / reducing position / holding); Dashed arrow (feedback data flow): Module 5 is bidirectionally connected to Module 2 and Module 4 respectively, forming a closed-loop self-calibration circuit.

[0039] Figure 2 The diagram shows a nine-square grid U-state matrix, weight allocation, and self-circulating feedback.

[0040] Figure 2 The left side of the middle section is a 3×3 grid matrix, where each cell represents a U-shaped input channel: Location, label, U-state type, weight range; First row, first column, S1: Geopolitical factors 0.10-0.20 The first row, second column (S2) shows the funding factor: 0.15-0.25. The first row, third column (S3) shows a technical factor of 0.15-0.25. Second row, first column, S4: Emotional factors 0.10-0.20 Second row, second column, S5: Policy factors 0.05-0.15 The external factors in the second row, third column (S6) are 0.05-0.10. The third row, first column (S7) shows the performance factor: 0.05-0.15. The third row, second column, S8 represents psychological factors ranging from 0.05 to 0.10. Third row, third column, S9: Environmental Factor Reference Weights Data flow direction: ① U-state data acquisition: The nine-grid matrix is ​​output to the field sensing module. ② Self-loop feedback: The energy conservation verification module feeds back the calibration parameters (energy distribution coefficients α, β, γ) to the U-state input matrix, updates the weights W_i, and forms a closed-loop self-loop.

[0041] Figure 3The diagram illustrates the LSTM unit, recursion depth, and memory depth, demonstrating the unfolded structure of the LSTM (Long Short-Term Memory) unit along the time axis. Horizontal direction: Expand by time steps t-2, t-1, t, t+1, t+2 to show the recursion depth L. Each time step contains one LSTM unit, which includes a forget gate, an input gate, and an output gate. Cell state update lines run throughout the entire unit sequence, preserving long-term memory information. Input-output relationship: x_t: Current input (potential energy accumulation P(t)) h_t: Current output (hidden state, i.e., the probability prediction result of state A). C_t: Current cell state (memory unit) Core recurrence relation: C_t = f_t * C_{t-1} + i_t * g_t h_t = o_t * tanh(C_t) The recursion depth L ranges from 2 to 16 levels, and the memory depth n ranges from 50 to 5000. Figure 4 The process of calculating F(t) and P(t) and determining the threshold is shown.

[0042] Figure 4 It includes the following steps: 1. Beginning (ellipse) 2. Acquire U-state input signal (rectangular) 3. Calculate the field vibration intensity F(t) (rectangular) 4. Calculate the accumulated potential energy P(t) (rectangle) according to the law of conservation of energy. 5. Determine if P(t) is greater than or equal to the trigger threshold P_threshold (diamond shape). If not, return to step 2. If so, proceed to step 6. 6. A-state trigger (rectangle) 7. Output A-state command (rectangle) 8. End (ellipse) Relevant formulas: F(t) = Σ(W_i × S_i(t)) / Σ(W_i) P(t) = α·P(t-1) + β·F(t) - γ·D(t), where α+β+γ=1 Triggering condition: P(t) ≥ P_threshold and ΔP / Δt ≥ 0 Parameter range: P_threshold = 0.7~0.85 Figure 5 The five states -1, 0, 1, U, and A are defined and their transition conditions are shown. Transition paths between the five states: -1 state (first deterministic state): located in the lower left, representing a downward trend; 0 state (second deterministic state): located at the center, representing a sideways trend; State 1 (Third Deterministic State): Located in the lower right corner, representing an upward trend; U-state (uncertainty state): Located directly above, and directly connected to the 0-state; A state (abnormal evolutionary state): located at the boundary between the U state and the -1 / 0 / 1 state.

[0043] Conversion path: Solid double-headed arrows: -1↔0, 0↔1, 0↔U, indicate normal state transitions; The bold red one-way arrow: U→A, indicates the core innovation of this invention (U→A boundary logic); Dashed diverging arrow: A→-1 / 0 / 1, select the target state according to the direction of the output command.

[0044] Triggering conditions: U→A: Potential energy accumulation P(t) ≥ trigger threshold P_threshold, and ΔP / Δt ≥ 0; A→-1 / 0 / 1: Selected according to the direction of the A-state output command (forward → 1 state, hold → 0 state, reverse → -1 state).

[0045] Figure 6 The calculation of ΔE, self-calibration mode, and α / β / γ adjustment are shown.

[0046] Figure 6 It includes the following steps: 1. System in operation (elliptical shape) 2. Calculate the input energy E_in = ΣF(t) (rectangular) 3. Calculate the output energy E_out = ΣΔE_release (rectangle) 4. Calculate the dissipated energy E_dissipation = ΣD(t) (rectangle) 5. Calculate the energy deviation ΔE = E_in - E_out - E_dissipation (rectangle) 6. Determine if |ΔE| ≤ ε (rhombus) If so, return to step 1 (system runs normally). If not, proceed to step 7. 7. Enter self-calibration mode (rectangle) 8. Adjust the energy distribution coefficients α, β, and γ to maintain α + β + γ = 1 (rectangular shape). 9. Recalculate ΔE (rectangle) 10. Determine if |ΔE| ≤ ε (rhombus) If not, return to step 8. If so, exit self-calibration mode and return to step 1. Parameter range: ε = 0.05~0.10 Example 1: Financial Quantitative Trading System System configuration: The field perception module integrates 8-dimensional market features (price momentum, trading volume, volatility, fund flow, sector strength, market breadth, VIX, basis), with a communication frequency of 1Hz. The recursive memory depth is n=200, and the trigger threshold P_threshold=0.75. The recursive tensor module has a hidden layer dimension of 64 and a recursion depth of 3 layers. Energy allocation coefficients are α=0.50, β=0.40, and γ=0.10.

[0047] Workflow: Step 1: The field perception module receives multi-dimensional market data and calculates the market uncertainty intensity F(t). Step 2: When multiple features simultaneously indicate an unclear direction, F(t) increases, potential energy P(t) begins to accumulate, and the system enters state U. Step 3: Use the recursive tensor module to mine historical potential energy sequences and predict when trends will form. Step 4: When P(t)≥0.75, trigger state A and output the "open position / add position" instruction. Step 5: Monitor energy balance and automatically calibrate parameters. Example 2: Traditional Chinese Medicine Auxiliary Diagnosis and Treatment System (Dynamic Medication Adjustment for Chronic Diseases in the Elderly) System configuration: The field perception module integrates 7 dimensions of patient status (emotional score, sleep quality, appetite, daytime activity, cognitive function, side effect load, and vital signs), with a communication frequency of 0.1Hz. The recursive memory depth is n=90, and the trigger threshold P_threshold=0.70. The recursive tensor module has a hidden layer dimension of 32 and a recursion depth of 2 layers. Energy allocation coefficients are α=0.55, β=0.35, and γ=0.10.

[0048] Workflow: Step 1: The field sensing module receives multidimensional state data of the patient and calculates the uncertainty intensity F(t) of the health status. Step 2: When abnormal fluctuations occur in multiple state dimensions, F(t) increases, potential energy P(t) begins to accumulate, and the system enters state U (corresponding to the chaotic state of "struggle between righteous and evil" and "syndrome transformation" in traditional Chinese medicine). Step 3: The recursive tensor module identifies abnormal patterns and determines the direction of disease progression (development towards mania / depression). Step 4: When P(t)≥0.70, trigger state A and output dynamic medication adjustment suggestions (such as "increase olanzapine dose to prevent positivity" or "increase venlafaxine dose to raise positivity"), completing the U→A transition. This system is the first to quantify the traditional Chinese medicine concept of "when the body's vital energy is abundant, pathogens cannot invade" as a process of potential energy accumulation and release. The U-state corresponds to the critical point in the human body's evolution from health to disease, and the A-state corresponds to the decisive point that triggers therapeutic intervention. The amount of potential energy accumulated, P(t), is directly mapped to the intensity of the "disease."

[0049] Example 3: Industrial Sensor Network System (Multi-Source Information Fusion Anomaly Detection) System configuration: The field sensing module connects to 32 wireless sensor nodes to collect data such as temperature, humidity, vibration, and gas concentration, with a communication frequency of 10Hz. The recursive memory depth is n=200, and the trigger threshold P_threshold=0.70. The recursive tensor module has a hidden layer dimension of 128 and a recursion depth of 4 layers. The energy distribution coefficients are α=0.55, β=0.35, and γ=0.10.

[0050] Workflow: Step 1: The field sensing module receives data from each node and calculates the overall network uncertainty F(t). Step 2: When multiple nodes simultaneously exhibit abnormal readings, F(t) increases, potential energy P(t) begins to accumulate rapidly, and the system is in state U. Step 3: The recursive tensor module identifies abnormal patterns and determines whether it is a systemic failure. Step 4: When P(t)≥0.70, trigger state A, output "system abnormality" alarm and fault location information, and complete the U→A conversion.

Claims

1. A U→A recursive tensor uncertainty transformation system based on energy conservation, characterized in that, include: The field sensing module is configured to receive external uncertain U-state input signals and calculate the field vibration intensity F(t); The potential energy accumulation module is configured to convert the input energy into accumulable potential energy P(t) according to the field vibration intensity F(t) and the law of conservation of energy. The recursive tensor module is configured to perform deep recursive learning on historical potential energy sequences to extract deep features of uncertainty evolution. The A-state triggering module is configured to monitor the accumulated potential energy P(t). When P(t) reaches or exceeds the preset triggering threshold, a deterministic A-state confirmation is triggered. The energy conservation verification and self-calibration module is configured to monitor the balance of system input energy, output energy and dissipated energy, and automatically adjust the energy distribution coefficient when the energy deviation exceeds the preset tolerance. The priority of this calibration module is higher than that of the A-state triggering module. The status output module is configured to output system status transition instructions based on the A-state trigger signal.

2. The U→A recursive tensor uncertainty transformation system based on energy conservation according to claim 1, characterized in that, The field sensing module receives an external uncertain U-state input signal and calculates the field vibration intensity F(t): F(t) = Σ(W_i × S_i(t)) / Σ(W_i) Where S_i(t) is the signal strength of the i-th channel at time t, including physical signals such as sensor data and measured values; event signals such as state changes and triggering events; and environmental signals such as temperature, humidity, and vibration. W_i is the weighting coefficient of the i-th channel.

3. The U→A recursive tensor uncertainty transformation system based on energy conservation according to claim 1, characterized in that, The potential energy accumulation module models historical potential energy sequences using a recursive tensor machine to predict potential energy accumulation trends. Its core energy conversion formula is: P(t) = α•F(t) + β•Σ_{i=1}^{n} (λ^{i} × F(ti)) - γ•D(t) in: P(t) represents the amount of potential energy accumulated at the current moment; F(t) represents the field vibration intensity at the current moment; Σ_{i=1}^{n} (λ^{i} × F(ti)) is the recursive weighted sum of the historical potential energy sequence, λ is the decay factor, 0<λ<1, and n is the recursive memory depth; D(t) is the energy dissipation term; α, β, and γ are energy distribution coefficients that satisfy α + β + γ = 1.

4. The U→A recursive tensor uncertainty transformation system based on energy conservation according to claim 1, characterized in that, The recursive tensor module adopts an LSTM or GRU network structure. The input of this module is the potential energy accumulation sequence P(tn) to P(t), and the output is the potential energy breakthrough time window and the prediction of the conversion direction. It is also equipped with an online learning mode: when the prediction error of N consecutive U→A conversions exceeds the preset threshold, the system automatically expands the hidden layer dimension H or increases the recursion depth L until the prediction error drops below the threshold.

5. The U→A recursive tensor uncertainty transformation system based on energy conservation according to claim 1, characterized in that, The A-state triggering module is configured with a conversion direction selection unit and a hysteresis comparator: when the potential energy accumulation P(t) ≥ P_threshold_up, the forward conversion U→A+ of the A-state is triggered; when P(t) ≤ P_threshold_down, the reverse conversion U→A- of the A-state is triggered; a hysteresis interval ΔP_hyst is set between the upper and lower thresholds to prevent frequent conversions at the threshold boundaries.

6. The U→A recursive tensor uncertainty transformation system based on energy conservation according to claim 1, characterized in that, The energy conservation verification and self-calibration module is configured to monitor the system's input energy E_in, output energy E_out, and dissipated energy E_dissipation to verify the energy conservation constraints. AE = E_in - E_out - E_dissipation in: E_in = ΣF(t), the total input energy; E_out = ΣΔE(release), outputting the total energy; E_dissipation = ΣD(t), the total energy dissipated; When |AE| exceeds the preset tolerance ε, the system automatically triggers self-calibration, adjusting the energy distribution coefficients α, β, and γ to meet the constraint α+β+γ=1, ensuring that the energy consumption of the system is minimized during long-term operation.

7. The U→A recursive tensor uncertainty transformation system based on energy conservation according to claim 1, characterized in that, The system state transition instructions of the status output module include: The amount of potential energy accumulated at the time of triggering, P(t); Change direction, either forward or reverse; Suggested actions are specific control commands output based on the application scenario.

8. The U→A recursive tensor uncertainty transformation system based on energy conservation according to claim 1, characterized in that, The system is configured for continuous self-running mode: after a U→A conversion is completed, the field sensing module automatically resets and starts the next cycle of U-state monitoring without the need for an external reset signal; when the system detects that the potential energy accumulation P(t) calculation is abnormal or the energy deviation exceeds the preset tolerance 2ε, it automatically resets the state of the recursive tensor module and reinitializes the memory window n.

9. The U→A recursive tensor uncertainty transformation system based on energy conservation according to claim 1, characterized in that, The system has self-evolution capabilities, including a meta-learning module configured to monitor the commonalities and differences in historical U-state transition tasks. After the system completes K different types of U→A transition tasks within a preset time window, the meta-learning module updates the initial parameters of the recursive tensor module, thereby reducing the cold start prediction error of the system on new tasks.

10. A method for implementing a U→A recursive tensor uncertainty transformation system based on energy conservation as described in any of claims 1-9, characterized in that, Includes the following steps: Step S1: Sensing external U-state input and calculating field vibration intensity F(t); Step S2: Convert the input energy into accumulable potential energy P(t) according to the law of conservation of energy; Step S3: Perform deep recursive learning on the historical potential energy sequence using a recursive tensor machine to predict the timing of potential energy breakout; Step S4: When the potential energy accumulation P(t) reaches or exceeds the preset trigger threshold, trigger state A confirmation and determine the conversion direction; Step S5: Monitor the balance of input energy, output energy and dissipated energy of the system, and automatically calibrate the energy distribution coefficient when the energy deviation exceeds the preset tolerance; Step S6: Output system state transition instructions; The computer has a readable storage medium on which a computer program is stored, which, when executed by a processor, implements all or part of the steps of the method.

Images