A household fire-retardant early warning cloth wire integrated with lightweight material and intelligent AI

Abstract

The application relates to a household flame-retardant early warning cloth electric wire integrating lightweight material and applicable intelligent AI, which comprises a conductor layer and an insulation structure wrapped outside the conductor layer; the insulation structure comprises, from inside to outside, a sensing adhesion layer, an inner insulation layer and an outer flame-retardant layer; the sensing adhesion layer is arranged on the outer surface of the conductor layer; a distributed thermal micro-sensing system is arranged in the sensing adhesion layer; the distributed thermal micro-sensing system comprises a plurality of thermal micro-sensing units distributed at equal intervals along the axial direction of the conductor layer and two parallel micro-connection wires; each thermal micro-sensing unit is connected in series through the micro-connection wires to form a sensing loop. The inner-embedded sensing of the insulation layer realizes direct temperature measurement of the conductor surface, the response speed is improved compared with an external sensor, and the distributed arrangement can realize full-line coverage and reduce the monitoring blind area.

Description

Technical Field

[0001] This invention relates to the field of wire and cable technology, specifically to a household flame-retardant early warning electrical wire that integrates lightweight materials and is suitable for intelligent AI. Background Technology

[0002] According to fire statistics released by the Fire and Rescue Bureau of the Ministry of Emergency Management, residential electrical fires have consistently accounted for over 60% of all fires in my country. Of these, fires caused by three main factors—precursors to short circuits, prolonged overload, and poor contact leading to localized overheating—account for over 85% of all electrical fires. Traditional residential electrical protection systems consist of air switches, fuses, and residual current devices (RCDs). Their core protection logic is current threshold-triggered reactive protection: the circuit breaker only trips when the circuit current exceeds the rated tripping threshold or a leakage occurs. By this time, the insulation layer has often already broken down, and an open flame has already formed, making it impossible to provide early warning and intervention for fire precursors.

[0003] In home renovations with concealed wiring, where wires are buried in walls, ceilings, or floors, early signs of overheating and arcing cannot be detected manually, and locating the fault is difficult. By the time a smoke alarm is triggered, the fire has already progressed, missing the optimal response time. Therefore, the industry urgently needs an active early warning technology that can be integrated into the wiring itself, covering the entire circuit without requiring external accessories, to identify early signs of fire and implement protective measures before the insulation layer ages and breaks down. Summary of the Invention

[0004] (a) Technical problems to be solved

[0005] This invention provides a household flame-retardant early warning electrical wiring that integrates lightweight materials and is compatible with intelligent AI, solving the problems mentioned in the background art.

[0006] (II) Technical Solution

[0007] To achieve the above objectives, the present invention provides the following technical solution: a household flame-retardant early warning electrical wiring device integrating lightweight materials and suitable for intelligent AI, comprising a conductor layer and an insulation structure covering the conductor layer. The insulation structure, from the inside out, comprises a sensing bonding layer, an inner insulation layer, and an outer flame-retardant layer. The sensing bonding layer is bonded to the outer surface of the conductor layer, and a distributed thermal micro-sensing system is arranged inside the sensing bonding layer. The distributed thermal micro-sensing system includes several thermal micro-sensing units evenly distributed along the axial direction of the conductor layer and two parallel micro-connecting wires. Each thermal micro-sensing unit is connected in series through the micro-connecting wires to form a sensing loop. It also includes an end-integrated control unit composed of a signal conditioning unit, a low-power control unit, a passive inductive power supply unit, and a local audible and visual early warning unit. The module assembly includes a lightweight neural decision tree AI model and a hierarchical fault discrimination module embedded within the low-power control unit. The lightweight neural decision tree AI model takes the line's highest temperature, temperature change rate, temperature non-uniformity, and circuit operating current as input features, and outputs eight categories of operating conditions, including normal operating conditions, ambient temperature rise, mild overload, severe overload, mild local overheating, severe local overheating, short circuit precursors, and sensor faults. The hierarchical fault discrimination module, based on the temperature and current sampling data of the entire line collected by the sensor circuit, performs hierarchical classification and identification of fire precursors, and drives the local audible and visual warning unit to execute a graded response based on the classification results. The passive inductive power extraction unit is sleeved on the outside of the conductor layer, and relies on the circuit current to extract and supply power.

[0008] Preferably, the thermistor microsensor unit uses a thin-film encapsulated NTC thermistor chip, and the temperature sensing surface of the thermistor microsensor unit is in close contact with the outer surface of the conductor layer; each thermistor microsensor unit is connected in parallel with a monolithic ceramic capacitor, and the capacitance value of each monolithic ceramic capacitor increases in a fixed step to form a unique resonant frequency identifier; the sensing bonding layer is also filled with low-temperature curing thermally conductive adhesive, which fills the gap between the thermistor microsensor unit and the conductor layer and forms a continuous heat conduction path.

[0009] Preferably, the inner insulation layer and the outer flame-retardant layer are integrally formed using a double-layer co-extrusion process. The inner insulation layer is made of high thermal conductivity, lightweight, and flame-retardant PVC material, and its formula by weight is: 100 parts PVC resin, 28-32 parts surface-modified nano aluminum hydroxide, 7-9 parts surface-silane-modified nano hexagonal boron nitride, 23-27 parts dioctyl phthalate, 4.5-5.5 parts calcium-zinc composite stabilizer, 1.0-1.4 parts calcium stearate, and 2.5-3.5 parts acrylate impact modifier. The outer flame-retardant layer is a low thermal conductivity, lightweight flame-retardant PVC material, and its formula by weight is as follows: 100 parts PVC resin, 33-37 parts surface-modified nano magnesium hydroxide, 2.5-3.5 parts fumed silica, 20-24 parts dioctyl phthalate, 5.0-6.0 parts calcium-zinc composite stabilizer, 0.8-1.2 parts calcium stearate, 3.5-4.5 parts acrylate impact modifier, and 14-16 parts surface-modified hollow glass microspheres.

[0010] In a further preferred embodiment, the signal conditioning unit comprises a frequency-domain reflective multi-point temperature measurement structure consisting of a frequency sweep signal generation module, a signal amplification module, an RMS detection module, and an AD acquisition module. When the signal conditioning unit is working, it outputs a 10kHz-1MHz frequency sweep sinusoidal signal and loads it onto the sensing circuit. By detecting the circuit impedance amplitude at different frequencies, it identifies each resonant frequency point and calculates the temperature at each point by combining the correspondence between the resonant frequency and the NTC resistance value.

[0011] In a further preferred embodiment, the hierarchical fault discrimination module uses a dual-channel analog comparator to perform real-time threshold comparisons of the highest temperature and the loop current, respectively. When the threshold is exceeded, the low-power control unit is woken up. After being woken up, the low-power control unit completes full-point temperature scanning and current sampling, calculates the input features, and completes data filtering. The input features are then processed by the model to obtain the operating condition classification result.

[0012] In a further preferred embodiment, the specific rules for the graded response are as follows: under normal operating conditions and under conditions of increased ambient temperature, the low-power control unit remains in a sleep state; under conditions of slight overload and slight local overheating, a yellow audible and visual warning is triggered and continuous monitoring is performed; under conditions of severe overload, a red audible and visual warning is triggered; under conditions of severe local overheating, a red audible and visual warning is triggered.

[0013] In a further preferred embodiment, the passive inductive power supply unit adopts a through-hole miniature current transformer with a transformation ratio of 200:1, which can output induced electrical energy when the circuit current is ≥0.1A; the induced electrical energy is output as 3.3V DC voltage after full-wave rectification, filtering and low-dropout linear regulation; the passive inductive power supply unit also has a built-in energy storage tantalum capacitor, which can maintain the system operation for a certain period of time for fault data storage after the circuit is de-energized.

[0014] In a further preferred embodiment, the low-power control unit incorporates a dynamic self-calibration module. This module includes dynamic environmental temperature calibration, sensor drift calibration, and adaptive threshold adjustment. The dynamic environmental temperature calibration calculates the lowest temperature across the entire line every hour and updates the environmental temperature benchmark, automatically correcting fault thresholds. The sensor drift calibration corrects the zero-point resistance of the thermistor micro-sensor unit every three months during the line's idle period, offsetting aging drift. The adaptive threshold adjustment fine-tunes the threshold every month based on long-term operating temperature data of the line, adapting to the heat dissipation differences in different wiring environments.

[0015] (III) Beneficial Effects

[0016] Compared with existing technologies, this invention provides a household flame-retardant early warning electrical wiring device that integrates lightweight materials and is compatible with intelligent AI, and has the following beneficial effects:

[0017] In this invention, the temperature of the conductor surface is directly measured by an embedded sensor in the insulation layer, which improves the response speed compared to an external sensor. Moreover, the distributed arrangement can achieve full line coverage and reduce monitoring blind spots.

[0018] In this invention, by setting up a passive inductive power supply unit, it can obtain power completely passively and without communication dependence. It can operate automatically when powered on, without the need for manual configuration and maintenance, and is suitable for all concealed wiring scenarios in home decoration.

[0019] In this invention, by setting an inner insulation layer and an outer flame-retardant layer, different functional focuses can be achieved under a unified extrusion process. The overall density is lower than that of traditional PVC insulation materials, while simultaneously meeting the Class B flame-retardant requirements. Attached Figure Description

[0020] Figure 1 A schematic diagram of the cross-sectional structure of a household flame-retardant early warning electrical wiring according to the implementation plan;

[0021] Figure 2 This is a block diagram of the distributed thermal microsensor system according to the implementation plan;

[0022] Figure 3 This is a flowchart illustrating the three-layer fault diagnosis architecture based on the implementation plan;

[0023] Figure 4 This is a block diagram of the internal circuit of the end-unit integrated control module according to the implementation plan;

[0024] Figure 5 This is a flowchart illustrating the manufacturing process of household flame-retardant early warning electrical wiring according to the implementation plan. Detailed Implementation

[0025] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0026] This embodiment of the integrated lightweight material is suitable for intelligent AI-powered household flame-retardant early warning wiring. It uses the most commonly used 2.5mm² wiring wire in home decoration as the carrier. The specific structural design of the present invention is described in detail. The outer diameter of the finished product is the same as that of the traditional BV-2.5 wire, and it can be directly used with all wiring fittings and terminals.

[0027] In this embodiment, the household flame-retardant early warning wiring cable integrating lightweight materials and suitable for intelligent AI includes a conductor layer and an insulation structure covering the conductor layer. The insulation structure, from the inside out, includes a sensing bonding layer, an inner insulation layer, and an outer flame-retardant layer. The sensing bonding layer is bonded to the outer surface of the conductor layer, and a distributed thermal micro-sensing system is arranged inside the sensing bonding layer. The distributed thermal micro-sensing system includes several thermal micro-sensing units evenly spaced along the axial direction of the conductor layer and two parallel micro-connecting wires. Each thermal micro-sensing unit is connected in series through the micro-connecting wires to form a sensing loop. The household flame-retardant early warning wiring cable integrating lightweight materials and suitable for intelligent AI also includes an end-unit integrated control module composed of a signal conditioning unit, a low-power control unit, a passive inductive power supply unit, and a local sound and light early warning unit.

[0028] In this embodiment, the conductor layer uses Class 1 solid copper conductor as specified in GB / T3956-2008, made of oxygen-free copper, with a nominal cross-sectional area of ​​2.5 mm², corresponding to a diameter of 1.78 mm, and a DC resistance ≤ 7.41 Ω / km (20℃). The conductor surface pretreatment first involves ultrasonic cleaning to remove surface brushing oil and oxide layer, followed by passivation to form a thin passivation film to prevent oxidation during subsequent processing; the surface roughness of the passivated conductor is controlled within a certain range. The following measures are taken to ensure a tight fit with the sensing unit.

[0029] In this embodiment, the sensing bonding layer is located between the outer surface of the conductor and the inner insulating layer, and consists of three parts: a thermistor micro-sensing unit, micro-connecting wires, and a thermally conductive adhesive layer. The overall thickness is ≤0.1mm, without significantly increasing the outer diameter of the wires. The thermistor micro-sensing unit uses a custom-designed thin-film NTC thermistor chip with a bare die size of 0.3mm×0.3mm×0.05mm, a resistance of 10kΩ±1% at 25℃, and a response time constant ≤0.5s (in liquid). The chip is double-sided packaged with a 25μm thick polyimide film, resulting in a packaged size of 0.5mm×0.5mm×0.08mm. The electrode leads are 0.01mm thick gold-plated copper foil with an electrode spacing of 0.3mm. Each sensing unit is connected in parallel with a 0402 packaged monolithic ceramic capacitor, with the capacitance value starting from 1.0nF and increasing in 0.5nF increments, corresponding to a unique resonant frequency. The capacitor and sensing unit are integrally soldered onto the PI film, and the solder joints are coated with UV insulating adhesive and cured. Within the 0-1000mm range of the wire inlet, one sensing unit is placed every 500mm, for a total of three. In the middle section after 1000mm, one sensing unit is placed every 1000mm. Similarly, three sensing units are densely arranged within the last 1000mm of the outlet. A total of 52 sensing units are arranged for a single 50-meter roll of wire. The micro-connecting conductors are two 0.05mm diameter polyurethane enameled copper round wires with a enamel film thickness of 0.008mm and a breakdown voltage ≥50V. The two conductors are arranged parallel to each other at both ends of the horizontal diameter of the conductor cross-section, closely adhering to the conductor surface. All sensing units are connected in series between the two micro-conductors. One electrode of each sensing unit is connected to the first micro-conductor, and the other electrode is connected to the second micro-conductor, forming a series circuit. Each sensing unit and its parallel capacitor form an independent resonant branch, and the overall circuit exhibits multi-resonance peak impedance characteristics. The thermally conductive adhesive layer uses a single-component low-temperature curing thermally conductive epoxy adhesive with a thermal conductivity of [missing information]. The curing temperature is 80℃ and the curing time is 30 minutes. The adhesive is applied between the temperature measuring surface of the sensing unit and the conductor, and the coating thickness is controlled within 0.02mm to fill the interface gap and ensure that the heat of the conductor is efficiently transferred to the sensing chip. After the adhesive is cured, the Shore hardness D≥60, which has sufficient mechanical strength to withstand the pressure of the subsequent extrusion process.

[0030] In this embodiment, the insulation layer is formed in one step using a double-layer co-extrusion process, with a total thickness of 0.8mm, consistent with the nominal insulation thickness of traditional BV-2.5 wire. The inner insulation layer is 0.4mm thick and made of high thermal conductivity, lightweight, flame-retardant PVC material. During extrusion, it adheres tightly to the sensing bonding layer, completely encapsulating the sensing unit and microwires. The inner surface of the inner insulation layer is tightly bonded to the conductor and sensing unit without air gaps, ensuring continuous heat conduction. The outer flame-retardant layer is 0.4mm thick and made of low thermal conductivity, lightweight, flame-retardant PVC material. It is fused with the inner insulation layer at the extrusion die, with no delamination at the interface. The outer surface of the outer flame-retardant layer is smooth, and the outer diameter tolerance is controlled within ±0.1mm. The eccentricity of the insulation layer is ≤5%, ensuring uniform insulation thickness in all directions. The finished product passes a 2kV / 1min power frequency withstand voltage test without breakdown, and the insulation resistance is ≥100MΩ (20℃), meeting all electrical performance requirements of GB / T5023.2.

[0031] In this embodiment, the end-integrated control module is located at the wire inlet, with an overall length of 50mm and a maximum diameter of 8mm, allowing direct insertion into the wiring terminals of a household distribution box. The module housing is injection molded from flame-retardant V0-grade ABS engineering plastic, with internal PCB mounting slots and wire fixing slots. After removing the insulation layer from the wire end, the conductor and micro-wires are soldered to corresponding pads on the PCB, and then the entire module is encapsulated with thermally conductive epoxy resin to achieve an IP65 protection rating. The PCB integrates five major functional circuits: signal conditioning circuit, microcontroller minimum system, passive power supply circuit, micro tripping execution circuit, and audible and visual warning circuit.

[0032] In this embodiment, the inner insulation layer and the outer flame retardant layer are integrally formed by a double-layer co-extrusion process. The inner insulation layer is made of high thermal conductivity, lightweight flame retardant PVC material, and the formula by weight is as follows: 100 parts of PVC resin, 28-32 parts of surface-modified nano aluminum hydroxide, 7-9 parts of surface silane-modified nano hexagonal boron nitride, 23-27 parts of dioctyl phthalate (plasticizer), 4.5-5.5 parts of calcium zinc composite stabilizer, 1.0-1.4 parts of calcium stearate (lubricant), 2.5-3.5 parts of acrylate impact modifier, and 9-11 parts of surface-modified hollow glass microspheres. The outer flame-retardant layer is made of low thermal conductivity, lightweight flame-retardant PVC material, and its formula by weight is as follows: 100 parts PVC resin, 33-37 parts surface-modified nano-magnesium hydroxide, 2.5-3.5 parts fumed silica, 20-24 parts dioctyl phthalate, 5.0-6.0 parts calcium-zinc composite stabilizer, 0.8-1.2 parts calcium stearate, 3.5-4.5 parts acrylate impact modifier, and 14-16 parts surface-modified hollow glass microspheres. The surface-modified nano-aluminum hydroxide in the inner insulation layer is nano-aluminum hydroxide with a surface modified by silane coupling agent KH550. The surface-modified nano-magnesium hydroxide in the outer flame-retardant layer is nano-magnesium hydroxide with a surface modified by stearic acid.

[0033] For performance testing of the inner insulation layer, 100 parts of PVC resin, 30 parts of nano aluminum hydroxide, 8 parts of nano boron nitride, 25 parts of dioctyl phthalate, 5 parts of calcium-zinc stabilizer, 1.2 parts of calcium stearate, 3 parts of acrylate impact modifier, and 10 parts of hollow glass microspheres were used. The PVC resin, calcium-zinc stabilizer, and calcium stearate were added to a high-speed mixer and mixed at 120°C for 10 minutes to ensure full adsorption of the stabilizer. Then, the nano aluminum hydroxide, nano boron nitride, and hollow glass microspheres were pre-mixed in a three-dimensional mixer for 15 minutes to ensure uniform dispersion of the filler. Next, the mixed resin, filler, dioctyl phthalate, and acrylate impact modifier were fed into a twin-screw extruder. The extrusion temperature was set in stages: feeding section 140°C, plasticizing section 160°C, homogenizing section 165°C, and die head 170°C, with a screw speed of 200 rpm. Finally, the extruded strip was water-cooled, air-dried, and granulated to obtain the inner insulation material granules. The performance test results are as follows:

[0034]

[0035] Table 1. Test results of inner insulation layer performance

[0036] For performance testing of the outer flame-retardant layer, the following materials were used: 100 parts PVC resin, 35 parts nano magnesium hydroxide, 3 parts fumed silica, 22 parts dioctyl phthalate, 5.5 parts calcium-zinc stabilizer, 1.0 part calcium stearate, 4 parts acrylate impact modifier, and 15 parts hollow glass microspheres. The preparation process of the outer flame-retardant layer was the same as that of the inner layer, and the extrusion temperature was set in stages: feeding stage 135℃, plasticizing stage 155℃, homogenizing stage 160℃, and die head 165℃. The performance test results are as follows:

[0037]

[0038] Table 2 Performance test results of the outer flame retardant layer

[0039] Test results show that the outer layer material density is reduced by 18.8%, resulting in a more significant lightweight effect; the thermal conductivity is reduced by 29.4% compared to traditional PVC, effectively blocking ambient temperature fluctuations; the oxygen index is increased by 6.1 percentage points, demonstrating excellent flame retardant performance; and the low-temperature impact resistance meets standards, making it suitable for home decoration environments in different regions. The inner and outer layers were extruded in one step using a double-layer co-extrusion mold, and standard samples were prepared for interfacial peel testing. The results show that the interfacial peel strength of the double-layer material is ≥8 N / cm, and the fracture mode is matrix fracture rather than interfacial separation, indicating that the two layers are well-bonded and will not delaminate under conditions such as thermal cycling and bending, ensuring the structural stability of the wire for long-term use.

[0040] In this embodiment, the distributed thermistor microsensor system utilizes frequency-domain reflective distributed temperature measurement based on the series resonant frequency division multiplexing principle. All NTC sensing units are connected in series on two microwires, and each sensing unit is connected in parallel with a capacitor of a unique capacitance value, forming an independent RLC series resonant branch. The resonant frequency of different branches is determined by both the capacitance value and the NTC resistance value. When the NTC resistance value changes with temperature, the resonant frequency of the corresponding branch also shifts. The system outputs a frequency sweep signal to detect the change curve of the loop impedance with frequency, identifies the frequency and amplitude of each resonant peak, calculates the corresponding NTC resistance value, and then converts it to obtain the temperature at that point. This scheme requires only two wires to achieve independent temperature measurement at dozens of points.

[0041] In this embodiment, the signal conditioning circuit is integrated on the PCB of the end-control module and consists of three parts: a sweep frequency generation module, an impedance detection module, and an AD acquisition module. The sweep frequency generation module uses an 8-bit high-speed DAC chip to generate a sinusoidal signal with an output frequency range of 10kHz-1MHz and a frequency resolution of 1kHz. The DAC output is filtered by a second-order low-pass filter to remove harmonics, and then impedance-matched by a voltage follower composed of an operational amplifier, resulting in a sinusoidal sweep frequency signal with a stable amplitude of 1V. The sweep frequency signal from the impedance detection module is applied to the sensing loop via a precision sampling resistor with a resistance of 100Ω and an accuracy of 0.1%. A differential operational amplifier is used to acquire the voltage across the sampling resistor, and the effective value of the loop current is obtained through an RMS conversion chip. Simultaneously, the effective value of the voltage across the sensing loop is acquired, and the total loop impedance is calculated using Ohm's law. The AD acquisition module uses a 12-bit successive approximation ADC with a sampling rate of 100kSPS to acquire the effective values ​​of voltage and current signals, which are then transmitted to the microcontroller for calculation. After the end-unit integrated control module acquires the impedance-frequency curve, it calculates the temperature at each point according to the following steps: First, it performs peak detection on the impedance curve to identify the frequency points corresponding to all local maxima, which are the resonant frequencies of each sensing branch; then, based on the resonant frequency and the known capacitance value, it calculates the real-time resistance value of the corresponding NTC using the series resonance formula. The calculation formula is as follows: ,in The amplitude of the resonant peak impedance. The resonant frequency is L, and the distributed inductance of the circuit (factory calibration value) is L; then, through... The equation converts the NTC resistance value into a temperature value, and the calculation formula is as follows: A, B, and C are the nominal coefficients of the NTC, which are corrected by three-point calibration before leaving the factory. Finally, the offset of each sensing unit obtained by the factory calibration is used for single-point compensation to further improve the temperature detection accuracy.

[0042] In this embodiment, a lightweight neural decision tree AI model and a hierarchical fault discrimination module are embedded within the low-power control unit. The lightweight neural decision tree AI model takes the line's highest temperature, temperature change rate, temperature non-uniformity, and circuit operating current as input features, and outputs eight categories of operating conditions, including normal operation, ambient temperature rise, mild overload, severe overload, mild local overheating, severe local overheating, short-circuit precursors, and sensor faults. The hierarchical fault discrimination module, based on the full-line temperature data and current sampling data collected from the sensor circuit, classifies and identifies fire precursors hierarchically, and drives the local audible and visual warning unit to execute a graded response based on the classification results.

[0043] The lightweight neural decision tree AI model takes four-dimensional physical features as input, all of which can be directly calculated from local sensor data without any external information. The highest line temperature is the maximum real-time temperature of all sensor points, reflecting the degree of heating at the most severe point on the line and serving as a core indicator for assessing overheating risk; the sampling frequency is 1Hz, and the moving average filtering window is 5 seconds. The temperature change rate is the slope of the linear fit of the highest temperature within a 10-second moving window, measured in °C / s, reflecting the speed of temperature rise; short-circuit precursors are characterized by a rapid temperature rise, with a change rate much greater than that of overload and localized overheating; the calculation uses least squares linear fitting to eliminate the influence of instantaneous noise. Temperature non-uniformity is the difference between the highest temperature and the average temperature of all sensor points, reflecting the uniformity of temperature distribution; localized overheating faults are characterized by high temperatures at individual points and low overall temperatures, with a non-uniformity significantly greater than that of overload faults; this feature is the basis for distinguishing between overload and localized overheating. The effective value of the current collected by the current transformer at the input terminal of the circuit operating current; overload faults are inevitably accompanied by current exceeding the standard, while local overheating fault currents are usually within the rated range; this feature can further improve the accuracy of fault classification. The lightweight neural decision tree AI model adopts a four-depth binary neural decision tree, with the following structure: The first layer (root node) has one internal node, completing the initial major category division; the second layer has two internal nodes, performing secondary divisions for "normal" and "fault" classes respectively; the third layer has four internal nodes, completing the fine classification; the fourth layer has eight leaf nodes, corresponding to the final eight operating condition classifications. There are a total of seven internal nodes, each corresponding to a linear perceptron, containing four weight parameters (corresponding to four-dimensional input features) and one bias parameter, thus the number of parameters per node is five. The splitting function of each internal node is calculated using the formula for a linear weighted sum. ;in,

[0044] As weight, As a bias term, when If the condition is met, proceed to the right child node; otherwise, proceed to the left child node. Compared to traditional hard-threshold decision trees, the linear combination splitting method of neural decision trees can achieve a slanted classification hyperplane, achieving higher classification accuracy with fewer nodes. During training, the lightweight neural decision tree AI model collected full-condition data through a power line fault simulation test bench, constructing a dataset containing 13,000 samples. This dataset includes 5,000 samples of normal operating conditions, covering ambient temperatures from -10℃ to 40℃ and load currents from 0 to 1.2 times the rated value, encompassing scenarios such as day-night temperature variations and normal load fluctuations; 3,000 samples of overload conditions, covering 1.2, 1.5, and 2 times the rated current, with continuous temperature increases, including overload scenarios under different ambient temperatures; 3,000 samples of localized overheating, simulating poor contact with contact resistances of 10W-100W at different locations, covering varying degrees of localized overheating; and 2,000 samples of short-circuit precursors, simulating intermittent insulation breakdown arcing using a high-voltage arcing device, with arcing currents of 5A-50A and arcing frequencies of 0.1Hz-1Hz, simulating early short-circuit precursors. The dataset is divided into training and testing sets in a 7:3 ratio to ensure consistent proportions for each type of sample. During training, batch gradient descent was used with cross-entropy loss as the loss function. The initial learning rate was 0.01, which decayed exponentially with each training epoch. The training consisted of 500 epochs, and L2 regularization was applied to prevent overfitting. After training, the model was quantized and compressed, converting all weights and biases to 16-bit fixed-point numbers to further reduce computation and storage requirements. The classification confusion matrix on the test set is as follows (accuracy, %):

[0045]

[0046] Table 3. Test Results of the Lightweight Neural Decision Tree AI Model

[0047] The lightweight neural decision tree AI model is implemented on an 8-bit STM8S003 microcontroller, which has 8KB of Flash memory, 1KB of RAM, and a 16MHz clock speed. The model stores 35 trained parameters (28 weights + 7 biases) as a constant array in the Flash memory. The inference function uses fixed-point arithmetic, implemented entirely with integer addition and shift operations, eliminating the need for floating-point units. During runtime, the model starts from the root node, calculates the linear weighted sum of the current node, determines the sign, and then proceeds to the corresponding child node, looping until a leaf node is reached, returning the classification result.

[0048] In this embodiment, the hierarchical fault discrimination module classifies and identifies fire precursors based on the temperature and current sampling data collected from the entire circuit by the sensing loop, and drives the local audible and visual warning unit to execute a graded response according to the classification results. The hierarchical fault discrimination module uses a dual-channel analog comparator to perform real-time threshold comparisons of the highest temperature and the circuit current. When the threshold is exceeded, the low-power control unit is awakened. The first channel is connected to the rectified voltage of the temperature sensing loop, corresponding to the highest temperature threshold of 60°C; the second channel is connected to the current sampling voltage, corresponding to 1.5 times the rated current threshold. When the highest temperature of the circuit exceeds 60°C, or the circuit current exceeds 1.5 times the rated value, the comparator outputs a high level, triggering an external interrupt and waking up the microcontroller. After the low-power control unit is woken up, it completes full-point temperature scanning and current sampling, calculates input features, and performs data filtering. Upon wake-up, the microcontroller first enters the feature extraction stage, activating the signal conditioning circuit to perform full-point temperature scanning and obtain the temperature values ​​of all sensing points. Then, it initiates current sampling to obtain the effective value of the current in the current loop. Next, it calculates the highest line temperature and temperature non-uniformity features, while simultaneously reading historical temperature data to calculate the temperature change rate. Then, it removes outliers and performs amplitude limiting and normalization on the features to match the model input range. During this stage, rapid fault prediction is also performed. If the temperature exceeds 120℃ or the current exceeds three times the rated value, it is directly judged as a serious fault, skipping model inference and immediately executing a disconnection, further compressing the response time under extreme conditions. The preprocessed four-dimensional features are then processed by the model to obtain the operating condition classification results, and the corresponding response actions are executed based on the classification results. The specific rules for graded response are as follows: Under normal operating conditions and ambient temperature rise conditions, the low-power control unit remains in sleep mode; under mild overload and mild local overheating conditions, a yellow audible and visual warning is triggered and continuous monitoring is performed; under severe overload conditions, a red audible and visual warning is triggered; under severe local overheating conditions, a red audible and visual warning is triggered.

[0049] In this embodiment, the passive inductive power extraction unit is sleeved on the outside of the conductor layer. The passive inductive power extraction unit relies on the loop current to extract and supply power. The passive inductive power extraction unit includes a miniature current transformer, a rectification, filtering, and voltage regulation circuit, an overvoltage protection and discharge circuit, and an energy storage unit. The miniature current transformer is a through-type current transformer with a permalloy core, a transformation ratio of 200:1, and an inner diameter of 3mm, which can be directly sleeved on a 2.5mm² conductor; its rated input current is 5A, and its maximum input current is 50A; sufficient induced energy can be output when the loop current is ≥0.1A. The rectification, filtering, and voltage regulation circuit uses a bridge full-wave rectification, with Schottky diodes to reduce voltage drop; after rectification, it is filtered by a 100μF electrolytic capacitor, and then outputs a stable 3.3V DC voltage through a low-dropout linear regulator (LDO); the LDO has a quiescent current of 2μA and a low dropout voltage of 0.2V, ensuring power supply efficiency under low current conditions. The overvoltage protection and bleedering circuit can clamp the voltage below 5V and protect downstream circuits when a large current short circuit occurs in the circuit, causing a sharp rise in the transformer output voltage. At this point, the Zener diode breaks down and conducts, working in conjunction with the bleedering resistor to release excess energy. It can withstand a maximum short-circuit current surge of 50A. The energy storage unit uses a 100μF tantalum capacitor as the energy storage element, connected in parallel to the regulated output terminal. When the circuit suddenly loses power, the energy storage capacitor can maintain system operation for more than 10 seconds, ensuring that fault data is completely written to the Flash memory and simultaneously providing fault warning.

[0050] After the entire power line is energized, the current transformer gradually builds up the induced voltage. When the voltage rises to 3.0V and stabilizes, the system enters a five-minute reference calibration period, collecting the temperature of all points every minute, for a total of five sets of data. The average of the five sets of data is taken as the initial ambient temperature reference, and the fault temperature thresholds are automatically corrected based on the initial ambient temperature reference: for example, when the ambient temperature is 30℃, the overheating threshold is automatically increased by 5℃ to avoid false alarms caused by high ambient temperatures. After calibration is completed, the microcontroller enters deep sleep mode, the hardware comparator starts working, and the system officially enters normal monitoring mode. To adapt to environmental changes and sensor drift during long-term use, the low-power control unit incorporates a dynamic self-calibration module. This module includes dynamic ambient temperature calibration, sensor drift calibration, and adaptive threshold adjustment. Dynamic ambient temperature calibration calculates the lowest temperature across the entire line every hour and updates the ambient temperature baseline, automatically correcting fault thresholds. This mechanism automatically adapts to diurnal temperature differences and seasonal variations; for example, in summer when the ambient temperature is 35℃, the system will not misinterpret normal load temperature increases as overheating faults. Sensor drift calibration corrects the resistance zero point of the thermistor micro-sensor unit every three months using the line's idle period, offsetting aging drift. Adaptive threshold adjustment fine-tunes the judgment threshold monthly based on long-term operating temperature data to adapt to different wiring environments and heat dissipation differences. For example, in conduit wiring, where heat dissipation is poor and normal temperatures are higher, the system automatically raises the threshold to avoid false alarms; in exposed wiring, where heat dissipation is better, the threshold automatically lowers to improve sensitivity.

[0051] The production process of household flame-retardant early warning electrical wiring integrating lightweight materials and suitable for intelligent AI involves the following steps: conductor pretreatment, sensor unit mounting, micro-wire laying and welding, first curing, double-layer co-extrusion coating, end module integration, and finished product inspection and packaging. During conductor pretreatment, the conductor is released from the pay-off rack and its tension is stabilized by a tension controller. Then, an ultrasonic cleaning tank removes oil and oxide layers from the conductor surface, followed by hot air drying. Next, a thin passivation film is formed in a passivation tank to prevent oxidation and improve adhesion. Finally, it is dried again by hot air before proceeding to the next process. During sensor unit mounting, thermally conductive adhesive is first applied to designated positions on the conductor surface, with the amount controlled at 0.01 μL. Then, a high-precision vision-based placement machine places the thermal micro-sensor unit (with parallel capacitor) into the corresponding position, with pressure controlled at 0.5 N to avoid damaging the chip. After mounting, visual inspection confirms accurate positioning and no misalignment. During the microwire laying and welding process, two microwires are released from the feeder and guided parallel to both sides of the conductor surface via guide wheels, with tension controlled at 0.1N to prevent breakage. Then, a laser welding machine is used to weld the sensor unit electrode to the microwires, with a laser power of 1mW, a pulse width of 1ms, and a welding point diameter ≤0.1mm. After welding, UV insulating adhesive is applied and cured under UV lamps to ensure insulation at the welding points. During curing, the assembled conductor enters an 80℃ oven for 30 minutes to fully cure the thermally conductive adhesive. During the double-layer co-extrusion process, the inner extruder temperature is: body 140~165℃, die 170℃; the outer extruder temperature is: body 135~160℃, die 165℃; the extrusion speed... After extrusion, the wires are cooled in a three-stage water-cooling tank, with the water temperature gradually decreasing from 60℃ to 25℃ to avoid internal stress caused by rapid cooling. After cooling, the outer diameter is monitored online by an outer diameter gauge, and the extrusion parameters are adjusted in real time to ensure the outer diameter tolerance. During the end module integration process, the wires are first cut to a fixed length, and the insulation layer is stripped from the ends to expose the conductors and microwires. Then, the ends are soldered to the control module PCB board, and functional testing is performed after soldering. Finally, the module shell is injection molded and sealed with epoxy resin to ensure the protection level.

[0052] In the description of this invention, it should be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing the invention and for simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the invention. The terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance. Furthermore, unless otherwise explicitly specified and limited, the terms "installed," "connected," and "linked" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances. Moreover, the terms “comprising,” “including,” or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0053] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A household flame-retardant early warning electrical wire integrating lightweight materials and suitable for intelligent AI, comprising a conductor layer and an insulation structure covering the conductor layer, characterized in that: The insulation structure comprises, from the inside out, a sensing bonding layer, an inner insulation layer, and an outer flame-retardant layer; the sensing bonding layer is bonded to the outer surface of the conductor layer, and a distributed thermal micro-sensing system is arranged inside the sensing bonding layer. The distributed thermal micro-sensing system includes several thermal micro-sensing units distributed at equal intervals along the axial direction of the conductor layer and two parallel micro-connecting wires. Each thermal micro-sensing unit is connected in series through the micro-connecting wires to form a sensing circuit. It also includes an end-unit integrated control module consisting of a signal conditioning unit, a low-power control unit, a passive inductive power supply unit, and a local audible and visual warning unit. The low-power control unit is equipped with a lightweight neural decision tree AI model and a hierarchical fault discrimination module. The lightweight neural decision tree AI model takes the line's highest temperature, temperature change rate, temperature non-uniformity, and circuit operating current as input features, and outputs eight categories of operating conditions, including normal operating conditions, ambient temperature rise, mild overload, severe overload, mild local overheating, severe local overheating, short circuit precursors, and sensor faults. The hierarchical fault discrimination module classifies and identifies fire precursors hierarchically based on the temperature data and current sampling data of the entire line collected by the sensor circuit, and drives the local audible and visual warning unit to perform hierarchical responses according to the classification results. The passive inductive power supply unit is sleeved on the outside of the conductor layer and relies on the circuit current to sense and supply power.

2. The household flame-retardant early warning electrical wiring device integrating lightweight materials and suitable for intelligent AI as described in claim 1, characterized in that: The thermistor microsensor unit uses a thin-film packaged NTC thermistor chip, and the temperature sensing surface of the thermistor microsensor unit is in close contact with the outer surface of the conductor layer; each thermistor microsensor unit is connected in parallel with a monolithic ceramic capacitor, and the capacitance value of each monolithic ceramic capacitor increases in a fixed step to form a unique resonant frequency identifier.

3. The household flame-retardant early warning electrical wiring device integrating lightweight materials and suitable for intelligent AI as described in claim 2, characterized in that: The sensing bonding layer is also filled with low-temperature curing thermally conductive adhesive, which fills the gap between the thermally sensitive microsensing unit and the conductor layer and forms a continuous heat conduction path.

4. The household flame-retardant early warning electrical wiring device integrating lightweight materials and suitable for intelligent AI as described in claim 1, characterized in that: The inner insulation layer and the outer flame retardant layer are integrally formed by a double-layer co-extrusion process. The inner insulation layer is made of high thermal conductivity, lightweight, and flame retardant PVC material, and its formula by weight is as follows: 100 parts PVC resin, 28-32 parts surface-modified nano aluminum hydroxide, 7-9 parts surface silane-modified nano hexagonal boron nitride, 23-27 parts dioctyl phthalate, 4.5-5.5 parts calcium-zinc composite stabilizer, 1.0-1.4 parts calcium stearate, 2.5-3.5 parts acrylate impact modifier, and 9-11 parts surface-modified hollow glass microspheres.

5. A household flame-retardant early warning electrical wiring device integrating lightweight materials and suitable for intelligent AI, as described in claim 4, characterized in that: The outer flame-retardant layer is a low thermal conductivity, lightweight flame-retardant PVC material, and its formula by weight is as follows: 100 parts PVC resin, 33-37 parts surface-modified nano magnesium hydroxide, 2.5-3.5 parts fumed silica, 20-24 parts dioctyl phthalate, 5.0-6.0 parts calcium-zinc composite stabilizer, 0.8-1.2 parts calcium stearate, 3.5-4.5 parts acrylate impact modifier, and 14-16 parts surface-modified hollow glass microspheres.

6. The household flame-retardant early warning electrical wiring device integrating lightweight materials and suitable for intelligent AI as described in claim 1, characterized in that: The signal conditioning unit consists of a frequency sweep signal generation module, a signal amplification module, an RMS detection module, and an AD acquisition module, forming a frequency domain reflective multi-point temperature measurement structure. When the signal conditioning unit is working, it outputs a 10kHz-1MHz frequency sweep sinusoidal signal and loads it onto the sensing circuit. By detecting the circuit impedance amplitude at different frequencies, it identifies each resonant frequency point and calculates the temperature at each point by combining the correspondence between the resonant frequency and the NTC resistance value.

7. A household flame-retardant early warning electrical wiring device integrating lightweight materials and suitable for intelligent AI, as described in claim 1, characterized in that: The hierarchical fault discrimination module uses a dual-channel analog comparator to perform real-time threshold comparisons of the highest temperature and the loop current. When the threshold is exceeded, the low-power control unit is woken up. After being woken up, the low-power control unit completes full-point temperature scanning and current sampling, calculates the input features, and performs data filtering. The input features are processed by the model to obtain the working condition classification results.

8. A household flame-retardant early warning electrical wire integrating lightweight materials and suitable for intelligent AI, as described in claim 1, is characterized in that... The specific rules for the graded response are as follows: under normal operating conditions and when the ambient temperature rises, the low-power control unit remains in a sleep state; under mild overload and mild local overheating conditions, a yellow audible and visual warning is triggered and continuous monitoring is performed. Under severe overload conditions, a red audible and visual warning will be triggered; under severe local overheating conditions, a red audible and visual warning will also be triggered.

9. A household flame-retardant early warning electrical wiring device integrating lightweight materials and suitable for intelligent AI, as described in claim 1, characterized in that: The passive inductive power supply unit adopts a through-hole miniature current transformer with a transformation ratio of 200:

1. It can output induced energy when the circuit current is ≥0.1A. The induced energy is rectified by full wave, filtered and regulated by low dropout voltage to output a 3.3V DC voltage. The passive inductive power supply unit also has a built-in tantalum energy storage capacitor, which can maintain the system operation for a certain period of time after the circuit is de-energized for fault data storage.

10. A household flame-retardant early warning electrical wire integrating lightweight materials and suitable for intelligent AI, as described in claim 1, characterized in that: The low-power control unit has a built-in dynamic self-calibration module, which is equipped with dynamic calibration of ambient temperature, sensor drift calibration and threshold adaptive adjustment. The ambient temperature dynamic calibration system calculates the lowest temperature of the entire line every hour and updates the ambient temperature benchmark, automatically correcting fault thresholds. Sensor drift calibration is performed every three months during the line's no-load period to correct the zero resistance of the thermistor microsensor unit and counteract aging drift. The threshold is adaptively adjusted every month based on the long-term operating temperature data of the line to fine-tune the threshold and adapt to the heat dissipation differences of different cabling environments.

Images