Hardware testing method for controller of multi-split air conditioning system

Abstract

The invention discloses a multi-split air conditioning system controller hardware test method, and relates to the technical field of intelligent control system test and fault diagnosis, and the method comprises the steps: collecting the historical fault data of an industrial master controller, combining the dynamic load parameters of a multi-split air conditioning system, and obtaining a dynamic fault signal set through a generative adversarial network; fault signals in the dynamic fault signal set are injected into an industrial master controller of the multi-split air conditioning system, and a mapping data set containing time-fault type-response parameters is generated; and performing anti-fact analysis and causal graph modeling on the mapping data set through a causal inference engine, generating a probabilistic causal chain of the fault propagation path, and identifying a high-risk causal chain. According to the invention, the authenticity and diversity of fault signals in the controller test are improved, and the stability and reliability of the multi-split air-conditioning system controller under complex working conditions are also improved.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent control system testing and fault diagnosis, in particular to a hardware testing method for a multi-connected air-conditioning system controller. Background Art

[0002] With the increasing complexity of multi-split air-conditioning systems and the growing demand for intelligence, the functionality of industrial main controllers is becoming increasingly enhanced. The hardware platform must have a high degree of robustness and real-time response capabilities when facing complex working conditions, multi-variable coupling, and dynamic load changes. To ensure the reliability of controllers in actual applications, related testing technologies have undergone a development process from static functional verification to dynamic simulation verification. In the early days, it mainly relied on functional testing methods based on preset test cases. Subsequently, simulation testing technologies such as model-in-the-loop and hardware-in-the-loop were developed. These methods simulate some typical fault scenarios by embedding the controller in a virtual simulation environment. In recent years, with the introduction of big data and artificial intelligence technologies, some studies have attempted to use historical operating data to build a fault sample library and combine it with a rule engine for fault injection testing to improve test efficiency and automation level.

[0003] However, there are still limitations in addressing the high-dimensional dynamic characteristics of multi-split air conditioning systems. First, methods based on preset test cases often rely on artificially set typical fault signals, which makes it difficult to cover the nonlinear coupling faults and time-series-related complex fault behaviors that the system may encounter in real-world operation. Second, existing testing strategies based on fault sample libraries lack the ability to model fault propagation paths and cannot identify potential causal relationships between different components, thus hindering the precise location of system-level risk points and the optimal allocation of test resources. Summary of the Invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides a hardware testing method for a multi-split air-conditioning system controller to solve the problems of insufficient fault signal diversity and lack of causal modeling capability of fault propagation path in testing technology.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] In a first aspect, the present invention provides a hardware testing method for a multi-split air-conditioning system controller, which includes collecting historical fault data of an industrial main controller, combining the dynamic load parameters of the multi-split air-conditioning system, and obtaining a dynamic fault signal set by generating an adversarial network; injecting the fault signals in the dynamic fault signal set into the industrial main controller of the multi-split air-conditioning system to generate a mapping data set containing time-fault type-response parameters; performing counterfactual analysis and causal graph modeling on the mapping data set through a causal reasoning engine to generate a probabilistic causal chain of the fault propagation path and identify high-risk causal chains; optimizing the training strategy of the adversarial network generator based on the high-risk causal chain to form a closed-loop test ecosystem and perform iterative testing to generate a test report, wherein the test report includes a mapping data set, a causal graph, a high-risk causal chain and an adversarial network generator.

[0008] As a preferred solution of the multi-split air conditioning system controller hardware testing method of the present invention, wherein: the historical fault data includes voltage fluctuation waveforms, sensor failure logs, and communication interruption events;

[0009] The dynamic load parameters include the phase difference of the start and stop timing of multiple indoor units, the refrigerant flow distribution ratio and the compressor load rate change curve;

[0010] The fault signals in the dynamic fault signal set include abnormal voltage waveforms, communication interference sequences, and sensor failure characteristics.

[0011] As a preferred solution of the multi-split air conditioning system controller hardware testing method of the present invention, wherein: the dynamic fault signal set is obtained by generating an adversarial network, and the specific steps are as follows:

[0012] Combining 3D convolutional layers and gated recurrent units to construct residual blocks, which are then connected in series to form a spatiotemporal convolutional neural network architecture. This architecture is then used to construct a generative adversarial network generator.

[0013] The historical fault data and dynamic load parameters are input into the generative adversarial network generator to generate a dynamic fault signal set containing fault signals.

[0014] As a preferred solution of the multi-split air conditioning system controller hardware testing method of the present invention, wherein: the fault signal in the dynamic fault signal set is injected into the industrial main controller of the multi-split air conditioning system to generate a mapping data set containing time-fault type-response parameters, the specific steps are as follows:

[0015] Inject the fault signals in the dynamic fault signal set into the industrial main controller through independent channels of the hardware-in-the-loop test interface;

[0016] Collect response parameters of industrial main controller;

[0017] Classify and encode the fault signals in the dynamic fault signal set to generate a fault type code;

[0018] After injecting the fault type into the timestamp, it is associated with the response parameters of the industrial controller to generate a mapping data set containing time-fault type-response parameters.

[0019] As a preferred solution of the multi-split air conditioning system controller hardware testing method of the present invention, wherein: the counterfactual analysis and causal graph modeling of the mapping data set are performed by the causal reasoning engine, and the specific steps are as follows:

[0020] Define a set of causal assumptions by encoding dynamic load parameters and fault types;

[0021] Calculate the causal effect value of the fault type code on the response parameter based on the mapped data set;

[0022] Based on the DoWhy framework, a causal graph is constructed according to the set of causal assumptions and causal effect values.

[0023] As a preferred solution of the multi-split air conditioning system controller hardware testing method of the present invention, wherein: based on the DoWhy framework, a causal graph is constructed according to the causal hypothesis set and the causal effect value, the specific steps are as follows:

[0024] According to the fault type code, dynamic load parameters and response parameters, the node types are defined as fault type nodes, dynamic load nodes and response parameter nodes;

[0025] Define causal hypothesis edges based on the causal hypothesis set, and add physical association edges based on the physical association of parameters;

[0026] The causal effect value is normalized as the corresponding causal hypothesis edge weight, and the time-shifted mutual information between the dynamic load parameter and the response parameter is calculated and normalized as the physical association edge weight;

[0027] A causal graph is constructed based on the DoWhy framework according to the node type, physical association edge weight and causal hypothesis edge weight.

[0028] As a preferred solution of the multi-split air conditioning system controller hardware testing method of the present invention, wherein: the probabilistic causal chain of the fault propagation path is generated and the high-risk causal chain is identified. The specific steps are as follows:

[0029] Based on the causal hypothesis edge weight and the physical association edge weight, the path is filtered through depth-first search to obtain the valid path;

[0030] According to the causal hypothesis edge weight and the physical association edge weight, the confidence of the path is obtained and a probabilistic causal chain is generated;

[0031] A confidence threshold is set according to the distribution characteristics of historical confidence data, and the probabilistic causal chain with a confidence greater than the confidence threshold is regarded as a high-risk causal chain.

[0032] As a preferred solution of the hardware testing method for the multi-split air conditioning system controller of the present invention, wherein: the training strategy of the adversarial network generator is optimized based on the high-risk causal chain, a closed-loop test ecosystem is formed and iterative testing is performed to generate a test report. The specific steps are as follows:

[0033] Based on the high-risk causal chain, a multi-objective loss function is designed according to the adversarial loss, high-risk causal chain confidence loss, and dynamic load parameter loss to generate a total loss function;

[0034] The generator is trained based on the total loss function, and the generator parameters are updated through backpropagation to generate an optimized generative adversarial network generator;

[0035] Use the optimized generative adversarial network generator to form a closed-loop testing ecosystem, conduct iterative testing, and generate test reports.

[0036] In a second aspect, the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, any step of the multi-split air conditioning system controller hardware testing method as described in the first aspect of the present invention is implemented.

[0037] In a third aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, any step of the multi-split air conditioning system controller hardware testing method as described in the first aspect of the present invention is implemented.

[0038] The beneficial effects of this invention include: by combining a spatiotemporal convolutional neural network with a generative adversarial network, a diverse set of fault signals is generated that fuse historical fault data with dynamic load parameters. A causal inference engine is then used to model and analyze fault propagation paths, identifying high-risk causal chains. This allows for optimized generator training strategies and the formation of a closed-loop testing ecosystem. This not only improves the authenticity and diversity of fault signals during controller testing, but also enables quantitative analysis and risk prediction of system-level fault causal relationships. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0040] Figure 1 The flowchart of the hardware testing method of the multi-split air conditioning system controller is shown in FIG.

[0041] Figure 2 Flowchart for generating dynamic failure signals for generative adversarial networks.

[0042] Figure 3 Flowchart for building a causal graph for a causal inference engine.

[0043] Figure 4 Flowchart for dynamic fault signal generation and injection. DETAILED DESCRIPTION

[0044] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0045] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0046] Secondly, the term "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in various places throughout this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive of other embodiments.

[0047] Reference Figures 1 to 4 , is an embodiment of the present invention, which provides a hardware testing method for a multi-split air conditioning system controller, comprising the following steps:

[0048] S1: Collect historical fault data of industrial main controllers, combine it with the dynamic load parameters of the multi-split air conditioning system, and obtain a dynamic fault signal set through a generative adversarial network;

[0049] S1.1: Collect historical fault data of industrial controllers, including voltage fluctuation waveforms, sensor failure logs, and communication interruption events;

[0050] It should be noted that the voltage fluctuation waveform records the instantaneous change of the voltage amplitude in the AC power circuit, the sensor failure log records the disconnection or drift events of the temperature sensor and pressure sensor, and the communication interruption event records the timestamp and message content of the CAN bus communication timeout or verification error.

[0051] S1.2: Extract dynamic load parameters from the multi-split air conditioning system, including the phase difference of the start-stop sequence of multiple indoor units, the refrigerant flow distribution ratio, and the compressor load rate change curve;

[0052] It should be noted that the phase difference of the start-stop timing of multiple indoor units is the difference in time intervals between the start-up and shutdown times of the fan motors of each indoor unit. The refrigerant flow distribution ratio is the mapping relationship between the opening of the electronic expansion valve and the refrigerant flow of the corresponding indoor unit. The compressor load rate change curve is a time series record of the change in the compressor motor current as the number of indoor units starts and stops.

[0053] S1.3: Construct a generative adversarial network generator, input historical fault data and dynamic load parameters into the generative adversarial network generator, and generate a dynamic fault signal set containing fault signals;

[0054] Furthermore, the generative adversarial network generator adopts a spatiotemporal convolutional neural network architecture, which is composed of multiple residual blocks connected in series. Each residual block contains a three-dimensional convolutional layer and a gated recurrent unit. The three-dimensional convolutional layer extracts features from the time, space, and channel dimensions of historical fault data, and the gated recurrent unit models the temporal dependencies of dynamic load parameters. The historical fault data and dynamic load parameters are input into the generative adversarial network generator to generate a dynamic fault signal set containing fault signals. The fault signals in the dynamic fault signal set include abnormal voltage waveforms, communication interference sequences, and sensor failure characteristics.

[0055] It should be noted that the abnormal voltage waveform simulates the non-preset voltage drop or surge pattern in the power supply circuit, the communication interference sequence simulates the message conflict or noise injection mode in the CAN bus communication that does not appear in the historical data, and the sensor failure feature simulates the nonlinear drift or sudden failure behavior of the temperature sensor or pressure sensor.

[0056] S2: Inject the fault signals in the dynamic fault signal set into the industrial main controller of the multi-split air conditioning system to generate a mapping data set containing time-fault type-response parameters;

[0057] S2.1: Inject the abnormal voltage waveform, communication interference sequence, and sensor failure characteristics in the dynamic fault signal set into the industrial main controller through independent channels of the hardware-in-the-loop test interface;

[0058] It should be noted that the abnormal voltage waveform is injected through the programmable power supply loop, the communication interference sequence is injected through the CAN bus communication port, and the sensor failure characteristics are injected into the temperature sensor or pressure sensor signal chain through the analog signal conditioning circuit.

[0059] S2.2: Collect the response parameters of the industrial main controller, including power loop current ripple, PWM drive signal spectrum, and CAN bus bit error rate;

[0060] It should be noted that the instantaneous fluctuation of the power supply loop current ripple is recorded by a high-precision current probe, the harmonic components of the PWM drive signal spectrum are extracted by a spectrum analyzer, and the CAN bus bit error rate is calculated by statistical message error rate through communication protocol analysis equipment.

[0061] S2.3: Classify and encode the fault signals in the dynamic fault signal set to generate a fault type code. After injecting a timestamp, the code is associated with the response parameters of the industrial controller to generate a mapping data set containing time, fault type, and response parameters.

[0062] Furthermore, each fault signal in the dynamic fault signal set is classified and encoded, with the abnormal voltage waveform encoded as the voltage fault type, the communication interference sequence encoded as the communication fault type, and the sensor failure feature encoded as the sensor fault type. The fault type code is injected into the timestamp and associated with the response parameters of the industrial main controller to construct a time series matrix. The rows of the time series matrix correspond to the timestamp, and the columns contain the fault type code, power circuit current ripple, PWM drive signal spectrum, and CAN bus bit error rate, forming a mapping data set containing time-fault type-response parameters.

[0063] S3: Uses a causal reasoning engine to perform counterfactual analysis and causal graph modeling on the mapped dataset, generating probabilistic causal chains of fault propagation paths and identifying high-risk causal chains.

[0064] S3.1: Define a set of causal assumptions based on dynamic load parameters and fault type coding;

[0065] For example, voltage fault type causes abnormal current ripple in the power supply circuit, communication fault type causes increased CAN bus bit error rate, and sensor fault type causes PWM drive signal spectrum distortion.

[0066] S3.2: Calculate the causal effect value of the fault type code on the response parameters of the industrial main controller;

[0067] For each timestamp in the mapping data set, the response parameters of the industrial main controller are simulated when the fault corresponding to the fault type code does not occur. The actual recorded power loop current ripple, PWM drive signal spectrum, and CAN bus bit error rate are compared with the response parameters in the simulated fault-free state.

[0068] It should be noted that for each timestamp in the mapping dataset, a sliding window interpolation method is first used to generate the power circuit current ripple, PWM drive signal spectrum, and CAN bus bit error rate (BER) under a simulated fault-free state, based on historical data related to the start-stop sequence, phase difference, refrigerant flow distribution ratio, and compressor load factor of multiple indoor units under a fault-free state. The method then compares the difference between the actual recorded power circuit current ripple and the current ripple under the simulated fault-free state, the difference between the actual recorded PWM drive signal spectrum and the PWM harmonic energy distribution evolution under the simulated fault-free state, and the offset between the actual recorded CAN bus BER and the BER under the simulated fault-free state. Finally, these differences are used as the causal effect components of the fault type code on the industrial controller response parameters and are added to the dynamic load parameter interaction expression driven by the fault type code to complete the calculation of the total causal effect value. The total number of timestamps defines the calculation scope, and the parameters within each timestamp inherit the correspondence between the physical quantity name and the variable in the causal effect formula defined in the mapping dataset.

[0069] Calculate the causal effect value of the fault type code on the response parameters of the industrial main controller, the expression is:

[0070]

[0071] Where Γ is the causal effect value, N is the total number of timestamps, and I (t) is the power loop current ripple at time stamp t, ΔT (t) is the phase difference of the start-stop sequence of multiple indoor units at timestamp t, α is the interactive adjustment index between the power supply circuit current ripple and the start-stop sequence of multiple indoor units, P (t) is the spectrum of the PWM drive signal at time stamp t, R (t) is the refrigerant flow distribution ratio at time stamp t, L (t) is the compressor load rate change curve at time stamp t, E (t) is the CAN bus bit error rate at timestamp t, β is the baseline offset between the CAN bus bit error rate and the start-stop timing of multiple indoor units, is the voltage Vfault type at time stamp t, is the communication C fault type at timestamp t, is the fault type of sensor S at time stamp t;

[0072] S3.3: Based on the DoWhy framework, construct a causal diagram based on the set of causal hypotheses and causal effect values;

[0073] Based on the fault type code, dynamic load parameters and response parameters, the node types are defined as fault type nodes, dynamic load nodes and response parameter nodes;

[0074] It should be noted that the fault type node corresponds to the voltage fault type, communication fault type and sensor fault type in the fault type code; the dynamic load node corresponds to the phase difference of the start and stop timing of multiple indoor units, the refrigerant flow distribution ratio and the compressor load rate change curve in the dynamic load parameters; the response parameter node corresponds to the power circuit current ripple, PWM drive signal spectrum and CAN bus bit error rate in the response parameters.

[0075] Define causal hypothesis edges based on the causal hypothesis set, and add physical association edges based on the physical association of parameters;

[0076] For example, the voltage fault type that causes abnormal power circuit current ripple is converted into a causal hypothesis edge, the communication fault type that causes abnormal PWM drive signal spectrum is converted into a causal hypothesis edge, and the sensor fault type that causes abnormal CAN bus bit error rate is converted into a causal hypothesis edge; the phase difference of the start-stop timing of multiple indoor units with physical correlation and the PWM drive signal spectrum are added as physical correlation edges;

[0077] It should be noted that parameter physical correlation specifically refers to the direct or indirect causal relationship between dynamic load parameters such as the phase difference between the start-stop timing of multiple indoor units, the refrigerant flow distribution ratio, and the compressor load factor variation curve, and response parameters such as power circuit current ripple, PWM drive signal spectrum, and CAN bus bit error rate, based on the control logic and hardware characteristics of the multi-split air conditioning system. Specifically, the phase difference between the start-stop timing of multiple indoor units affects the instantaneous amplitude of current ripple through sudden changes in the power circuit load; the refrigerant flow distribution ratio changes the operating state of the compressor motor through electronic expansion valve opening adjustment, thereby driving the harmonic distribution of the PWM signal spectrum; and the compressor load factor variation curve is associated with the timing offset of the CAN bus bit error rate through the coupling effect of motor current fluctuations and communication bus power supply.

[0078] Calculate the time-shifted mutual information of dynamic load parameters and response parameters, the expression is;

[0079]

[0080] Where W is the time-shifted mutual information, ΔT (t -D) is the phase difference of the start and stop sequence of multiple indoor units at the time stamp tD, R (t -D) is the refrigerant flow distribution ratio at the time stamp tD, k is the refrigerant flow adjustment coefficient, L (t) is the compressor load rate change curve at time stamp t, I (t) is the power loop current waveform at time stamp t, E (t) is the CAN bus bit error rate at timestamp t, D is the delay time of load impact response, d is the bus error smoothing factor, and e is a minimum constant;

[0081] The causal effect value is normalized as the corresponding causal hypothesis edge weight, and the time-shifted mutual information is normalized as the physical association edge weight. Based on the causal hypothesis edge weight and the physical association edge weight, a depth-first search is performed to filter the paths and obtain the valid paths.

[0082] Furthermore, the causal hypothesis edge is defined as a first-level path, and the causal hypothesis edge and the physical association edge are defined as a second-level path. The edge weight threshold is set according to the distribution characteristics of the historical data of the edge weight. Starting from each fault type node, all paths are traversed, and the edges with the causal hypothesis edge weight greater than the edge weight threshold or the physical association edge weight greater than the edge weight threshold are regarded as valid paths.

[0083] It should be noted that the specific steps for setting the edge weight threshold according to the distribution characteristics of the edge weight historical data are as follows: based on the edge weight distribution of normal state and fault state in the historical data, the mean and standard deviation are calculated respectively, and the 3σ principle is used to define the upper bound of the normal weight and the lower bound of the fault weight, and the upper bound of the normal weight and the lower bound of the fault weight are taken as the edge weight threshold.

[0084] The causal hypothesis edge weight of the first-level path is used as the confidence of the first-level path, and the minimum value of the causal hypothesis edge weight and the physical association edge weight in the path is taken as the confidence of the second-level path. A probabilistic causal chain is generated, including fault type, dynamic load parameters, response parameters and confidence.

[0085] A confidence threshold is set according to the distribution characteristics of historical confidence data, and the probabilistic causal chain with a confidence greater than the confidence threshold is regarded as a high-risk causal chain.

[0086] S4: Optimize the training strategy of the generative adversarial network generator based on high-risk causal chains, form a closed-loop testing ecosystem, conduct iterative testing, and generate a test report.

[0087] The fault type and confidence level in the high-risk causal chain are encoded into a multidimensional feature vector. This multidimensional feature vector is converted into channel weights using a learnable weight matrix combined with a Sigmoid function. The features output by the three-dimensional convolutional layer are weighted by channel to generate weighted features. Causal path gating is introduced into the gated recurrent unit. Based on the correlation between the dynamic load parameters and the high-risk causal chain, the parameter generation direction is adjusted to generate the corrected dynamic load parameters.

[0088] Based on the high-risk causal chain, a multi-objective loss function is designed by using adversarial loss, high-risk causal chain confidence loss, and dynamic load parameter loss to generate a total loss function;

[0089] It should be noted that the adversarial loss is to ensure the authenticity of the dynamic fault signal set, the high-risk causal chain confidence loss is to constrain the fault type coding of the dynamic fault signal set to be consistent with the confidence distribution of the high-risk causal chain, and the dynamic load parameter loss is to ensure that the generated dynamic load parameters match the load interval defined in the high-risk causal chain; the specific steps for generating the total loss function based on the high-risk causal chain are as follows: first, define the adversarial loss function to evaluate the distribution consistency between the dynamic fault signal output by the generative adversarial network generator and the real fault signal in terms of voltage abnormal waveform, communication interference sequence, and sensor failure characteristics; secondly, define the high-risk causal chain confidence loss function to calculate the mean square error between the causal chain confidence corresponding to the fault type coding in the dynamic fault signal and the preset confidence in the high-risk causal chain set, and enhance the coverage of the generated signal on the high-risk path; finally, define the dynamic load parameter loss function to penalize the deviation of the phase difference of the start-stop timing of multiple indoor units, the refrigerant flow distribution ratio, and the compressor load rate change curve in the dynamic fault signal from the threshold interval defined by the high-risk causal chain, so as to ensure that the dynamic load parameter coupling characteristics conform to the physical correlation of the causal chain. The adversarial loss function, high-risk causal chain confidence loss function, and dynamic load parameter loss function are linearly superimposed according to preset weights to generate a total loss function for optimizing the training process of the generative adversarial network generator.

[0090] Based on the total loss function, the generator is trained using historical fault data to learn basic fault signals. The trained dynamic fault signal set is used to generate confidence through a causal reasoning engine, and the generator parameters are updated through backpropagation to obtain an optimized generative adversarial network generator.

[0091] Use the optimized generative adversarial network generator to form a closed-loop testing ecosystem and conduct iterative testing;

[0092] Record the mapped datasets, causal graphs, high-risk causal chains, and adversarial network generators for iterative testing, and generate test reports.

[0093] This embodiment also provides a computer device suitable for the hardware testing method of a multi-split air-conditioning system controller, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute computer-executable instructions to implement the hardware testing method of the multi-split air-conditioning system controller proposed in the above embodiment.

[0094] The computer device may be a terminal, comprising a processor, a memory, a communication interface, a display screen and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device comprises a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner may be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display or an electronic ink display screen, and the input device of the computer device may be a touch layer covering the display screen, or a button, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse.

[0095] This embodiment also provides a storage medium having a computer program stored thereon, which, when executed by a processor, implements the hardware testing method for a multi-split air conditioning system controller as proposed in the above embodiment; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0096] In summary, this invention combines a spatiotemporal convolutional neural network with a generative adversarial network to generate diverse fault signals that fuse historical fault data with dynamic load parameters. It then utilizes a causal inference engine to model and analyze fault propagation paths, identifying high-risk causal chains. This optimizes the generator training strategy and creates a closed-loop testing ecosystem. This not only improves the authenticity and diversity of fault signals during controller testing, but also enables quantitative analysis and risk prediction of system-level fault causal relationships.

[0097] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A hardware testing method for a multi-split air conditioning system controller, characterized by: include, Collect historical fault data from industrial controllers, combine it with the dynamic load parameters of the multi-split air conditioning system, and use a generative adversarial network to obtain a dynamic fault signal set. Injecting fault signals from the dynamic fault signal set into the industrial main controller of the multi-split air conditioning system to generate a mapping data set containing time-fault type-response parameters; Use the causal reasoning engine to perform counterfactual analysis and causal graph modeling on the mapped data set, generate probabilistic causal chains of fault propagation paths, and identify high-risk causal chains; The training strategy of the generative adversarial network generator is optimized based on the high-risk causal chain, a closed-loop testing ecosystem is formed and iterative testing is performed to generate a test report. The test report includes the mapping dataset, causal graph, high-risk causal chain and adversarial network generator.

2. The method for testing hardware of a multi-split air conditioning system controller according to claim 1, wherein: The historical fault data includes voltage fluctuation waveforms, sensor failure logs, and communication interruption events; The dynamic load parameters include the phase difference of the start and stop timing of multiple indoor units, the refrigerant flow distribution ratio and the compressor load rate change curve; The fault signals in the dynamic fault signal set include abnormal voltage waveforms, communication interference sequences, and sensor failure characteristics.

3. The method for testing hardware of a multi-split air conditioning system controller according to claim 2, wherein: The specific steps of obtaining a dynamic fault signal set by generating an adversarial network are as follows: Combining 3D convolutional layers and gated recurrent units to construct residual blocks, which are then connected in series to form a spatiotemporal convolutional neural network architecture. This architecture is then used to construct a generative adversarial network generator. The historical fault data and dynamic load parameters are input into the generative adversarial network generator to generate a dynamic fault signal set containing fault signals.

4. The method for testing hardware of a multi-split air conditioning system controller according to claim 3, wherein: The specific steps of injecting the fault signal in the dynamic fault signal set into the industrial main controller of the multi-split air conditioning system to generate a mapping data set containing time-fault type-response parameters are as follows: Inject the fault signals in the dynamic fault signal set into the industrial main controller through independent channels of the hardware-in-the-loop test interface; Collect response parameters of industrial main controller; Classify and encode the fault signals in the dynamic fault signal set to generate a fault type code; After injecting the fault type into the timestamp, it is associated with the response parameters of the industrial controller to generate a mapping data set containing time-fault type-response parameters.

5. The method for testing hardware of a multi-split air conditioning system controller according to claim 4, wherein: The causal reasoning engine is used to perform counterfactual analysis and causal graph modeling on the mapping data set. The specific steps are as follows: Define a set of causal assumptions by encoding dynamic load parameters and fault types; Calculate the causal effect value of the fault type code on the response parameter based on the mapped data set; Based on the DoWhy framework, a causal graph is constructed according to the set of causal assumptions and causal effect values.

6. The method for testing hardware of a multi-split air conditioning system controller according to claim 5, wherein: Based on the DoWhy framework, a causal graph is constructed according to the causal hypothesis set and causal effect value. The specific steps are as follows: According to the fault type code, dynamic load parameters and response parameters, the node types are defined as fault type nodes, dynamic load nodes and response parameter nodes; Define causal hypothesis edges based on the causal hypothesis set, and add physical association edges based on the physical association of parameters; The causal effect value is normalized as the corresponding causal hypothesis edge weight, and the time-shifted mutual information between the dynamic load parameter and the response parameter is calculated and normalized as the physical association edge weight; A causal graph is constructed based on the DoWhy framework according to the node type, physical association edge weight and causal hypothesis edge weight.

7. The method for testing hardware of a multi-split air conditioning system controller according to claim 6, wherein: The steps for generating a probabilistic causal chain of a fault propagation path and identifying high-risk causal chains are as follows: Based on the causal hypothesis edge weight and the physical association edge weight, the path is filtered through depth-first search to obtain the valid path; According to the causal hypothesis edge weight and the physical association edge weight, the confidence of the path is obtained and a probabilistic causal chain is generated; A confidence threshold is set according to the distribution characteristics of historical confidence data, and the probabilistic causal chain with a confidence greater than the confidence threshold is regarded as a high-risk causal chain.

8. The method for testing hardware of a multi-split air conditioning system controller according to claim 7, wherein: The training strategy of the generative adversarial network generator is optimized based on the high-risk causal chain, forming a closed-loop test ecosystem and conducting iterative testing to generate a test report. The specific steps are as follows: Based on the high-risk causal chain, a multi-objective loss function is designed according to the adversarial loss, high-risk causal chain confidence loss, and dynamic load parameter loss to generate a total loss function; The generator is trained based on the total loss function, and the generator parameters are updated through backpropagation to generate an optimized generative adversarial network generator; Use the optimized generative adversarial network generator to form a closed-loop testing ecosystem, conduct iterative testing, and generate test reports.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the multi-split air conditioning system controller hardware testing method according to any one of claims 1 to 8 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the multi-split air conditioning system controller hardware testing method according to any one of claims 1 to 8 are implemented.

Images