Integrated circuit control system based on artificial intelligence

Abstract

The invention discloses an integrated circuit control system based on artificial intelligence. The integrated circuit control system comprises a data acquisition module, a feature extraction module, a control decision mapping module, a loss design module, an integrated circuit control model establishment module and an integrated circuit control module. The invention belongs to the field of circuit control, and particularly relates to an integrated circuit control system based on artificial intelligence, which adopts a double-branch parallel architecture to lock burst, peak and slow variation trends; through the control decision mapping design, a response can be quickly made for a large-amplitude control instruction change; smooth nonlinear feedback is provided for small errors and stable states, and excessive adjustment and unnecessary fluctuation are avoided. By constructing an adjustable smooth loss function, when accidental sensing noise and sudden transient occur, a logarithmic function is adopted to perform an inhibition effect, so that control jitter is avoided; in a small error region, a proper feedback signal can be provided for a control system; and the control effect of the integrated circuit is improved.

Description

Technical Field

[0001] This invention relates to the field of circuit control, specifically to an integrated circuit control system based on artificial intelligence. Background Technology

[0002] An integrated circuit control system is a system based on integrated circuit technology that monitors and analyzes the operating state of a circuit, and adjusts and controls the circuit parameters according to the analysis results to achieve specific functions and performance requirements. However, general integrated circuit control systems suffer from insufficient ability to capture drastic local changes in the integrated circuit, inadequate characterization of long-term trends and slow-changing characteristics, and a lack of stability strategies for large-amplitude control, leading to poor integrated circuit control performance. Furthermore, general integrated circuit control systems are sensitive to sporadic sensor noise and sudden transient anomalies in sensor signals, and have poor feedback in small error regions, resulting in control jitter. Summary of the Invention

[0003] To address the above issues and overcome the shortcomings of existing technologies, this invention provides an integrated circuit control system based on artificial intelligence. Addressing the problems of general integrated circuit control systems, such as insufficient ability to capture drastic local changes in the integrated circuit, inadequate characterization of long-term trends and slow-changing characteristics, and a lack of stability strategies for large-amplitude control, leading to poor integrated circuit control performance, this solution employs a dual-branch parallel architecture. A small core with attention-based approach quickly locks onto sudden spikes and peaks, while a large core with channel gating locks onto slow-changing trends. Based on multi-level global-local feature enhancement, irrelevant redundancy is eliminated, preserving both rapid response capability and capturing the global slow-changing trends and periodic changes of the integrated circuit. Through control decision mapping design, it can quickly respond to large-amplitude changes in control commands, enhancing... The system's ability to respond to equipment malfunctions, fault repairs, or high-temperature alarms is enhanced. For small errors and stable states, it provides smooth nonlinear feedback, avoiding excessive adjustment and unnecessary fluctuations, thus improving the stability of the integrated circuit control system. Addressing the common problem of integrated circuit control systems being sensitive to occasional sensor noise and sudden transient anomalies in sensor signals, and exhibiting poor feedback in small error regions, leading to control jitter, this solution constructs an adjustable smoothing loss function. When occasional sensor noise and sudden transients occur, causing large error fluctuations, a logarithmic function is used to suppress them, making the control system insensitive to these anomalies and avoiding control jitter. Within the small error region, it provides suitable feedback signals to the control system, thereby improving the integrated circuit control effect.

[0004] The technical solution adopted by the present invention is as follows: The present invention provides an integrated circuit control system based on artificial intelligence, including a data acquisition module, a feature extraction module, a control decision mapping module, a loss design module, an integrated circuit control model establishment module, and an integrated circuit control module;

[0005] The data acquisition module collects historical signals from inside the integrated circuit and constructs a historical circuit sample set.

[0006] The feature extraction module captures local mutations and global long-range trends based on a dual-branch parallel architecture, and adaptively fuses the features of both.

[0007] The control decision mapping module flattens the fused features into vectors, inputs them into the lightweight control network, and maps the network output into actual circuit control commands through an activation function.

[0008] The loss design module designs a loss function based on the actual response and the expected response of the circuit.

[0009] The integrated circuit control model establishment module establishes an integrated circuit control model based on a historical circuit sample set, combined with a feature extraction module, a control decision mapping module, and a loss design module.

[0010] The integrated circuit control module controls the real-time integrated circuit based on the established integrated circuit control model.

[0011] Furthermore, the data acquisition module acquires historical signals from inside the integrated circuit and determines the expected response, using the expected response as a data label; it then divides the data into a sliding window of length T to construct a historical circuit sample set.

[0012] Furthermore, the feature extraction module performs feature extraction on the input X of the historical circuit sample set, including two parallel branches: a local fine-tuning branch and a global receptive branch. The local fine-tuning branch uses a small convolutional kernel of size k1×k1, followed by channel attention and spatial reconstruction, as follows: Where F1 is the local branch output; CR(·) is the Channel-attention+Spatial-reconstruction module; The convolution operation is performed using a k1×k1 kernel; the global receptive branch uses a large k2×k2 convolution kernel, combined with learnable channel weights, as follows: Where F2 is the global branch output; W gate It is a learnable channel weight vector; (·) is a convolution operation performed with a kernel of size k2×k2; prior feature fusion is represented as: F comb =F1 + F2; where, Fcomb It is the prior fusion output; the exception response channel is strengthened through global branching, represented as: X G =CIBlock k (GMP(F comb )); where X G It is a feature enhanced by global branching; GMP(·) is global max pooling; CIBlock k (·) represents k cascaded CIBlocks; CIBlock consists of: pointwise convolution → PReLU → Dropout → BatchNorm; local branches enhance channel interaction, represented as: X L =CIBlock k (F comb ); where X L This is the fusion feature after local branch processing; the fusion weights are then used to output the fusion, and the formula used is: X′ G =Repeat(X) G size(X) L )); X concat =Concat(X) G ,X′ G ); X fused =PWConv(X concat W = σ s (X fused ); Here, size(·) returns shape information; X′ G It is a global feature after broadcasting; X out It is a fused feature tensor; Repeat(X) G size(X) L )) is to use global features X G Copy to X L Same shape; X concat It is the concatenation result; PWConv(·) is a 1×1 convolution; X fused This involves mapping the concatenated features back to the original channel dimensions to obtain the fused feature representation; Concat(X G ,X′ G W = σ(X) is the global feature that concatenates the original and broadcast versions along the channel dimension; fused ) is the adaptive fusion coefficient tensor; σ s (·) is the Sigmoid activation function.

[0013] Furthermore, the control decision mapping module will X out Flatten the vector into a vector x, input it into the lightweight control network, and represent it as: x = Flatten(X) out); u = CtrlNet(x); where Flatten(·) is the flattening operation; CtrlNet(·) is the function mapping of the lightweight control network; u is the multiplexing control instruction, and each component u j Corresponding to one physical control quantity; for u j Mapping activation functions to actual control instructions is represented as follows: in, It is the actual control instruction, mapped to the actual control command range; σ(·) is the activation function mapping; λ, a, and ε control the shape and characteristics of the activation function; U max and U min These are the maximum and minimum values ​​of the control command range, respectively.

[0014] Furthermore, for the i-th sample, the loss design module takes the actual sensor signal waveform exhibited by the circuit after applying the actual control command as the actual response, and assumes that the error between the actual response and the expected response is r. i And construct an adjustable smoothing loss L, using the following formula: The overall gradient is expressed as: Where N is the total number of samples; k is the sensitivity parameter; and r is the population error.

[0015] Furthermore, the integrated circuit control model building module divides the historical circuit sample set into a test set and a training set; it constructs an integrated circuit control model based on the feature extraction module, control decision mapping module, and loss design module; it sets an error threshold, and when the integrated circuit control model converges to the training set loss or reaches the maximum number of training iterations, the integrated circuit control model training is complete; it calculates the error index between the actual response and the expected response of the trained integrated circuit control model to the test set, and if the error index is less than the corresponding threshold, the integrated circuit control model building is complete; otherwise, it adjusts the model parameters and retrains.

[0016] Furthermore, the integrated circuit control module, based on the established integrated circuit control model, collects the internal signals of the integrated circuit in real time and inputs them into the integrated circuit control model, and performs integrated circuit control based on the actual control commands obtained from the model.

[0017] The beneficial effects achieved by the present invention using the above solution are as follows:

[0018] (1) To address the problems of poor control performance of general integrated circuit control systems, such as insufficient ability to capture local drastic changes in integrated circuits, insufficient characterization of long-term trends and slow-changing characteristics, and lack of stability strategies for large-amplitude control, this solution adopts a dual-branch parallel architecture. It uses a small core + attention to quickly lock in sudden and sharp changes, and a large core + channel gating to lock in slow-changing trends. Based on multi-level global-local feature enhancement, irrelevant redundancy is eliminated, which not only retains the fast response capability, but also captures the global slow-changing trends and periodic changes of integrated circuits. Through the control decision mapping design, it can quickly respond to large-amplitude changes in control commands, and enhance the system's ability to cope with equipment abnormalities, fault repairs, or high-temperature alarms. For small errors and stable states, it provides smooth nonlinear feedback, avoiding excessive adjustment and unnecessary fluctuations, and improving the stability of the integrated circuit control system.

[0019] (2) To address the problem that general integrated circuit control systems are sensitive to occasional sensor noise and sudden transient anomalies in sensor signals, and have poor feedback in the small error range, which leads to control jitter in the integrated circuit, this solution constructs an adjustable smoothing loss function. When occasional sensor noise and sudden transients occur and the error fluctuates greatly, a logarithmic function is used to suppress them, thereby making the control system insensitive to these anomalies and avoiding control jitter. In the small error range, it can provide a suitable feedback signal to the control system, thereby improving the control effect of the integrated circuit. Attached Figure Description

[0020] Figure 1 This invention provides a flowchart of an integrated circuit control system based on artificial intelligence.

[0021] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof. Detailed Implementation

[0022] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0023] In the description of this invention, it should be understood that the terms "upper", "lower", "front", "rear", "left", "right", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and 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 this invention.

[0024] Example 1, see Figure 1 The present invention provides an integrated circuit control system based on artificial intelligence, including a data acquisition module, a feature extraction module, a control decision mapping module, a loss design module, an integrated circuit control model establishment module, and an integrated circuit control module;

[0025] The data acquisition module collects historical signals from inside the integrated circuit, constructs a historical circuit sample set, and sends the data to the feature extraction module.

[0026] The feature extraction module captures local mutations and global long-range trends based on a dual-branch parallel architecture, and adaptively fuses the features of both; then it sends the data to the control decision mapping module.

[0027] The control decision mapping module flattens the fused features into a vector, inputs it into the lightweight control network, and maps the network output into actual circuit control commands through an activation function; and sends the data to the loss design module.

[0028] The loss design module designs a loss function based on the actual response and expected response of the circuit, and sends the data to the integrated circuit control model building module.

[0029] The integrated circuit control model building module establishes an integrated circuit control model based on a historical circuit sample set, combined with a feature extraction module, a control decision mapping module, and a loss design module; and sends the data to the integrated circuit control module.

[0030] The integrated circuit control module controls the real-time integrated circuit based on the established integrated circuit control model.

[0031] Example 2, see Figure 1 This embodiment is based on the above embodiment. The data acquisition module acquires historical signals inside the integrated circuit and determines the desired response. The desired response is used as a data tag. The historical signals include voltage waveforms, current, temperature and frequency. The historical circuit sample set is constructed by dividing the sample set into segments with a sliding window of length T. Each input matrix of the historical circuit sample set is regarded as a two-dimensional matrix of sensor channel × sampling time, taking into account the interaction between different sensor channels and time modes. The desired response is the target waveform of each channel of the integrated circuit under ideal control.

[0032] Example 3, see Figure 1 This embodiment is based on the above embodiment. The feature extraction module uses a dual-branch parallel architecture to simultaneously capture sudden events and long-term trends. For the input X of the historical circuit sample set, feature extraction is performed, including two parallel branches: a local fine-grained branch and a global sensing branch. The local fine-grained branch uses a small 3×3 convolutional kernel, followed by channel attention and spatial reconstruction, focusing on capturing local abrupt changes in the sensing signal such as sudden current spikes and temperature surges, expressed as: F1 = CR(Conv3(X)); where F1 is the output of the local branch, focusing on representing short-term spikes and surges; CR(·) is the Channel-attention + Spatial-reconstruction module; Conv3(·) performs convolution operations with a 3×3 kernel. The global sensing branch uses a large 7×7 convolutional kernel and combines learnable channel weights to focus on the global periodicity and long-term dependence of slowly varying hot spots and frequency drift trends, expressed as: F2 = W gate ⊙Conv7(X); where F2 is the global branch output; W gate It is a learnable channel weight vector with the same shape as the number of channels; Conv7(·) is a 7×7 convolution operation with a convolution kernel; prior feature fusion is represented as: F comb =F1 + F2; where, F comb It is the prior fusion output; the most severe anomaly response channel is enhanced through global branches, represented as: X G =CIBlock k (GMP(F comb )); where X G It is a feature enhanced by global branching; GMP(·) is global max pooling; CIBlock k (·) represents k cascaded CIBlocks; CIBlock consists of: pointwise convolution → PReLU → Dropout → BatchNorm; local branches enhance channel interaction, represented as: X L =CIBlock k (F comb ); where X L This is the fusion feature after local branch processing; the fusion weights are then used to output the fusion, and the formula used is: X′ G =Repeat(X) G size(X) L )); X concat =Concat(X) G ,X′ G ); X fused =PWConv(X concat W = σ s (X fused); Here, size(·) returns shape information; X′ G It is a global feature after broadcasting; X out It is a fused feature tensor that captures both local and global information of the integrated circuit; Repeat(X) G size(X) L )) is to use global features X G Copy to X L Same shape; X concat It is the concatenation result; PWConv(·) is a 1×1 convolution; X fused This involves mapping the concatenated features back to the original channel dimensions to obtain the fused feature representation; Concat(X G ,X′ G W = σ(X) is the global feature that concatenates the original and broadcast versions along the channel dimension; fused ) is the adaptive fusion coefficient tensor; σ s (·) is the Sigmoid activation function; it eliminates irrelevant redundancy through low-level convolution, saving edge computing power, captures slow-changing trends with a large receptive field at high levels, provides early warning of hot spots, and adaptively fused attention to dynamically allocate attention based on the current dramatic changes vs. stability, taking into account both fast response and global stability.

[0033] Example 4, see Figure 1 This embodiment is based on the above embodiment, and the control decision mapping module will X out Flatten the vector into a vector x, input it into the lightweight control network, and represent it as: x = Flatten(X) out ); u = CtrlNet(x); where Flatten(·) is the flattening operation; CtrlNet(·) is the function mapping of the lightweight control network; u is the multiplexing control instruction, and each component u j This corresponds to one physical control quantity, including voltage regulation, current magnitude, temperature threshold, and switching signal; for u j Mapping activation functions to actual control instructions is represented as follows: in, It is the actual control instruction, mapped to the actual control command range; σ(·) is the activation function mapping; λ, a, and ε control the shape and characteristics of the activation function; U max and U min These are the maximum and minimum values ​​of the control command range, respectively; to ensure differentiability at the connection point of the three segments, take... Many integrated circuit control systems involve large-amplitude control commands, including rapid adjustments to voltage and current. Especially in cases of equipment malfunction, fault repair, or high-temperature alarms, the response to these large input changes is relatively slow. This solution enhances the response capability to such abrupt signals through the linear components at both ends. Furthermore, it provides smooth nonlinear feedback for small errors or steady states, which is crucial for the stability of the control system. In practical applications, integrated circuit control systems need to maintain small errors during smooth operation and converge quickly to the target state to avoid excessive adjustment and unnecessary fluctuations.

[0034] By performing the above operations, this solution addresses the problems of general integrated circuit control systems, such as insufficient ability to capture local drastic changes in integrated circuits, inadequate characterization of long-term trends and slow-changing characteristics, and a lack of stability strategies for large-amplitude control, leading to poor integrated circuit control performance. It adopts a dual-branch parallel architecture, using a small core with attention to quickly lock onto sudden spikes and peaks, and a large core with channel gating to lock onto slow-changing trends. Based on multi-level global-local feature enhancement, irrelevant redundancy is eliminated, preserving both rapid response capability and capturing the global slow-changing trends and periodic changes of the integrated circuit. Through control decision mapping design, it can quickly respond to large-amplitude changes in control commands, enhancing the system's ability to cope with equipment anomalies, fault repairs, or high-temperature alarms. For small errors and stable states, it provides smooth nonlinear feedback, avoiding excessive adjustment and unnecessary fluctuations, thus improving the stability of the integrated circuit control system.

[0035] Example 5, see Figure 1 This embodiment is based on the above embodiment. For the i-th sample, the loss design module takes the actual sensor signal waveform exhibited by the circuit after the actual control command is applied as the actual response. Let the error between the actual response and the expected response be r. i And construct an adjustable smoothing loss L, using the following formula: The overall gradient is expressed as: Where N is the total number of samples; k is the sensitivity parameter; r is the overall error; it is insensitive to occasional sensing noise and sudden transients with gradient saturation within ±k, avoiding control jitter caused by abnormal readings; it provides smooth nonlinear feedback in the small error region, accelerating convergence and ensuring control accuracy.

[0036] By performing the above operations, this solution addresses the problem of general integrated circuit control systems being sensitive to occasional sensor noise and sudden transient anomalies in sensor signals, exhibiting poor feedback in small error regions, and consequently leading to control jitter. This solution constructs an adjustable smoothing loss function. When occasional sensor noise and sudden transients occur, causing significant error fluctuations, a logarithmic function is used to suppress them, making the control system insensitive to these anomalies and avoiding control jitter. Furthermore, within small error regions, it provides suitable feedback signals to the control system, thereby improving the control performance of the integrated circuit.

[0037] Example 6, see Figure 1 This embodiment is based on the above embodiment. The integrated circuit control module collects the internal signals of the integrated circuit in real time based on the established integrated circuit control model and inputs them into the integrated circuit control model. Based on the actual control commands obtained from the model, the integrated circuit is controlled.

[0038] Example 7, see Figure 1 This embodiment is based on the above embodiment, and selects the control command range [U] min U max = [0V, 5V]; Transition band width a = 2.5, smoothing parameter ε = 1.0; Critical slope; First calculate

[0039] Case A: Small amplitude output u j =1.0, falls in [-a,a], σ(u j ) = 0.7311; In the intermediate nonlinear range, the output is not strictly linear, nor is it fully saturated, and it can respond sensitively to subtle control signals such as fine-tuning current or temperature threshold.

[0040] Case B: Positive mutation u j =3.0, greater than a, σ(u j ) = 0.9591; Entering the linear tail, the output is very close to the upper limit but not fully saturated, retaining a margin of 0.2V, which helps the hardware to tolerate extreme commands.

[0041] Case C: Negative mutation u j =―3.0, less than -a, σ(u) j ) = 0.0409; A slight upward adjustment near the lower limit helps avoid a hardware dead zone caused by commands returning to zero.

[0042] 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.

[0043] The present invention and its embodiments have been described above. This description is not restrictive, and the accompanying drawings are only one embodiment of the present invention; the actual structure is not limited thereto. In conclusion, if those skilled in the art are inspired by this description and design similar structures and embodiments without departing from the spirit of the invention, such designs should fall within the protection scope of the present invention.

Claims

1. An intelligent protection and control system for electric water heaters, characterized in that: The system includes a data acquisition module, a feature extraction module, a control decision mapping module, a loss design module, an integrated circuit control model establishment module, and an integrated circuit control module; The data acquisition module collects historical signals from inside the integrated circuit and constructs a historical circuit sample set. The feature extraction module captures local mutations and global long-range trends based on a dual-branch parallel architecture, and adaptively fuses the features of both. The control decision mapping module flattens the fused features into vectors, inputs them into the lightweight control network, and maps the network output into actual circuit control commands through an activation function. The loss design module designs a loss function based on the actual response and the expected response of the circuit. The integrated circuit control model establishment module establishes an integrated circuit control model based on a historical circuit sample set, combined with a feature extraction module, a control decision mapping module, and a loss design module. The integrated circuit control module controls the real-time integrated circuit based on the established integrated circuit control model.

2. The intelligent electric water heater protection control system according to claim 1, characterized in that: The feature extraction module extracts features from the input X of the historical circuit sample set, including two parallel branches: a local fine-tuning branch and a global receptive branch. The local fine-tuning branch uses a small convolutional kernel of size k1×k1, followed by channel attention and spatial reconstruction, as shown below: Where F1 is the local branch output; CR(·) is the Channel-attention+Spatial-reconstruction module; (·) performs a convolution operation with a kernel size of k1×k1; the global receptive branch uses a large convolution kernel of size k2×k2, combined with learnable channel weights, and is expressed as: Where F2 is the global branch output; W gate It is a learnable channel weight vector; Convolution is performed using a k2×k2 kernel; prior feature fusion is represented as: F comb =F1 + F2; where, F comb It is the prior fusion output; the exception response channel is strengthened through global branching, represented as: X G =CIBlock k (GMP(F comb )); where X G It is a feature enhanced by global branching; GMP(·) is global max pooling; CIBlock k (·) represents k cascaded CIBlocks; CIBlock consists of: pointwise convolution → PReLU → Dropout → BatchNorm; local branches enhance channel interaction, represented as: X L =CIBlock k (F comb ); where X L It is the fused feature after local branch processing; the fused weights are then output as fused X. out .

3. The intelligent electric water heater protection control system according to claim 2, characterized in that: The control decision mapping module will X out Flatten the vector into a vector x, input it into the lightweight control network, and represent it as: x = Flatten(X) out ); u = CtrlNet(x); where Flatten(·) is the flattening operation; CtrlNet(·) is the function mapping of the lightweight control network; u is the multiplexing control instruction, and each component u j Corresponding to one physical control quantity; for u j Mapping activation functions to actual control instructions is represented as follows: in, It is the actual control instruction, mapped to the actual control command range; σ(·) is the activation function mapping; λ, a, and ε control the shape and characteristics of the activation function; U max and U min These are the maximum and minimum values ​​of the control command range, respectively.

4. The intelligent electric water heater protection control system according to claim 3, characterized in that: For the i-th sample, the loss design module takes the actual sensor signal waveform exhibited by the circuit after applying the actual control command as the actual response, and assuming the error between the actual response and the expected response is r. i And construct an adjustable smoothing loss L, using the following formula: The overall gradient is expressed as: Where N is the total number of samples; k is the sensitivity parameter; and r is the population error.

5. The intelligent electric water heater protection control system according to claim 4, characterized in that: Furthermore, the integrated circuit control model building module divides the historical circuit sample set into a test set and a training set; it constructs an integrated circuit control model based on the feature extraction module, control decision mapping module, and loss design module; it sets an error threshold, and when the integrated circuit control model converges to the training set loss or reaches the maximum number of training iterations, the integrated circuit control model training is complete; it calculates the error index between the actual response and the expected response of the trained integrated circuit control model to the test set, and if the error index is less than the corresponding threshold, the integrated circuit control model building is complete; otherwise, it adjusts the model parameters and retrains.

6. The intelligent electric water heater protection control system according to claim 5, characterized in that: The data acquisition module collects historical signals from inside the integrated circuit and determines the expected response, using the expected response as a data label; it then divides the data into a sliding window of length T to construct a historical circuit sample set.

7. The intelligent electric water heater protection control system according to claim 6, characterized in that: The integrated circuit control module, based on the established integrated circuit control model, collects the internal signals of the integrated circuit in real time and inputs them into the integrated circuit control model, and performs integrated circuit control based on the actual control commands obtained from the model.

Images