Equipment state detection using audio data

Abstract

The systems and processes described herein can be used to determine faults present in pieces of equipment that are operating in a facility. Machine learning techniques can be implemented to analyze audio data produced by pieces of equipment to determine if the pieces of equipment are operating properly. When faults are present in pieces of equipment, additional computational techniques can be performed to indicate one or more classifications for the fault.

Description

EQUIPMENT STATE DETECTION USING AUDIO DATACLAIM OF PRIORITY

[0001] This patent application claims the benefit of priority to U.S. Provisional Patent Application Serial No. 63 / 664,915, filed June 27, 2024, which is incorporated by reference herein in its entirety.TECHNOLOGICAL FIELD

[0002] The present disclosure relates to implementations of systems and processes to determine a state of equipment that operates within an industrial, municipal, or manufacturing facility. More particularly, the present disclosure relates to systems and processes that analyze audio data corresponding to sound generated by equipment to determine whether the equipment is operating properly.BACKGROUND

[0003] The background description provided herein is for the purpose of generally presenting the context of the disclosure. Work of the presently named inventor(s), to the extent it is described in this background section, as well as aspects of the description that may not otherwise qualify as prior art at the time of filing, are neither expressly nor impliedly admitted as prior art against the present disclosure.

[0004] Equipment present in an industrial, municipal, or manufacturing facility can operate to manufacture products and / or transport material throughout the facility. In some cases, equipment can operate to transport fluids in an industrial or manufacturing facility. In these situations, an industrial or manufacturing facility can include a number of pumps that operate to transport fluids within the facility and / or to locations outside of the facility. In still other scenarios, equipment can operate to transport parts, partially completed products, completed products, and the like within an industrial or manufacturing facility. For example, conveyor belts, lifts, cranes, and so forth can operate to move partially completed products within an industrial or manufacturing facility. In additional instances, robotic machinery can be present in an industrial or manufacturing facility as well as various tools or other machinery that perform operations used in the manufacturing or processing of materials within the facility.

[0005] The equipment operating within an industrial or manufacturing facility produces sound. In some cases, when a piece of equipment is not operating properly, the sound produced by the piece of equipment can change. Often, several pieces of equipment are operating concurrently and it can be difficult to distinguish the sound made by one piece of equipment from the sound made by another piece of equipment. Thus, it can be difficult to identify faulty equipment by sound when multiple pieces of equipment located proximate to each other are producing sound that creates a relatively high level of background noise. Additionally, changes to sound produced by faulty equipment can be subtle and not easily identifiable in relation to the sound produced when the equipment is operating properly.SUMMARY

[0006] The following presents a simplified summary of one or more implementations of the present disclosure in order to provide a basic understanding of such implementations. This summary is not an extensive overview of all contemplated implementations and is intended to neither identify key or critical elements of all implementations, nor delineate the scope of any or all implementations.

[0007] In one or more implementations, a method to identify faulty equipment based on audio data can include obtaining audio data produced during operation of equipment. The audio data can be analyzed to produce first modified audio data by performing a time domain analysis of the audio data, second modified audio data by performing a frequency domain analysis of the audio data; and third modified audio data by performing a time-frequency domain analysis of the audio data. The first modified audio data, the second modified audio data, and the third modified audio data can be input data for generative artificial intelligence (Al) models including a variant of large language models that can process sound and images. In addition, the generative Al model can computationally analyze the input data to determine a quantitative measure relating to an anomaly being present with respect to operation of the equipment. Following the detection of an anomaly on the equipment, a physics-based probabilistic fault detection model can then be used to process the audio data to identify the most probable failed component(s) in the equipment based on a predicted probability of failure. A user interface can be displayed that includes a user interface element corresponding to the quantitative measure for anomaly and fault detection results.

[0008] While multiple implementations are disclosed, still other implementations of the present disclosure will become apparent to those skilled in the art from the following detaileddescription, which shows and describes illustrative implementations of the invention. As will be realized, the various implementations of the present disclosure are capable of modifications in various obvious aspects, all without departing from the spirit and scope of the present disclosure. Accordingly, the drawings and detailed description are to be regarded as illustrative in nature and not restrictive.BRIEF DESCRIPTION OF THE DRAWINGS

[0009] While the specification concludes with claims particularly pointing out and distinctly claiming the subject matter that is regarded as forming the various implementations of the present disclosure, it is believed that the invention will be better understood from the following description taken in conjunction with the accompanying figures. In the figures, the depicted structural elements are not to scale, and certain components may be enlarged relative to the other components for purposes of emphasis and understanding.

[0010] Figure 1 is a diagram of a framework to detect a state of one or more pieces of equipment based on audio data produced by the one or more pieces of equipment, according to one or more example implementations.

[0011] Figure 2 is a diagram of a computational architecture that performs time domain processing and frequency domain processing of audio produced by one or more pieces of equipment to determine a state of the one or more pieces of equipment, according to one or more example implementations.

[0012] Figure 3 is a diagram of a computational framework that implements a generative Al model to analyze audio data produced by one or more pieces of equipment to determine a state of the one or more pieces of equipment, according to one or more example implementations.

[0013] Figure 4 is a diagram of a computational framework that implements a machine learning classification model to determine a classification in relation to the state of one or more pieces of equipment based on audio data produced by the one or more pieces of equipment, according to one or more example implementations.

[0014] Figure 5 is a diagram of a computational framework to determine a probability of a state of equipment by analyzing audio data in relation to expected operating frequencies, in accordance with one or more example implementations.

[0015] Figure 6 is a flow diagram of a process to analyze audio data produced by one or more pieces of equipment to determine a state of the one or more pieces of equipment, according to one or more example implementations.

[0016] Figure 7 is a block diagram illustrating components of a machine, in the form of a computer system, that may read and execute instructions from one or more machine-readable media to perform any one or more methodologies described herein, in accordance with one or more example implementations.

[0017] Figure 8 is a block diagram illustrating a representative software architecture that may be used in conjunction with one or more hardware architectures described herein, in accordance with one or more example implementations.DETAILED DESCRIPTION

[0018] The present disclosure, in one or more implementations, relates to systems and processes to analyze audio data produced by one or more pieces of equipment to determine a state of the one or more pieces of equipment. The state of a piece of equipment can indicate that the piece of equipment is operating properly. Additionally, a state of a piece of equipment can indicate that a fault is present in the operation of the piece of equipment. In various examples, a piece of equipment can be operating properly when operating conditions for the piece of equipment are within specified parameters. Further, operation of a piece of equipment can be identified as faulty when operating conditions for the piece of equipment are outside of the specified parameters. For example, a pump can be considered to operate properly when a flow rate into and / or out of the pump is within a tolerance of a specified flow rate for the pump. A pump can also be considered to operate properly when a speed of rotation of one or more parts of the pump are within a tolerance of a specified speed for the pump. In situations where the flow rate of the pump and / or the rotational speed of one or more parts of the pump are outside of the tolerance for the flow rate and / or speed, the pump can be identified as faulty.

[0019] The implementations herein include the use of generative Al models to analyze audio data produced by equipment. The implementation of generative Al models to analyze audio data to identify faulty equipment can provide more accurate results than existing techniques that are used to identify faulty equipment based on the analysis of audio data. Additionally, the methods and systems described herein can implement one or more additional computational models to determine classifications for the types of fault present with respect to a piece of equipment. Typically, existing techniques that analyze audio data to identify faulty equipment do not determine classifications for faults present with respect to the equipment. Further, by using audio data to determine equipment that is not operating properly, the systems, processes, and techniques described herein are more reliable than techniques that analyze only sensordata, such as vibrational sensors and / or accelerometers. For example, analysis of sound data can provide a holistic analysis in relation to a number of components of a piece of equipment based on a single audio data capture event. In contrast, typical sensor-based analyses of vibrational data is more limited in scope. Often, the implementation of multiple data capture events and / or the use sensors placed in relation to multiple components of a piece of equipment are not performed due to the cost of these implementations or because of difficulties in the analysis of the vibrational data captured in these scenarios. To illustrate, due to the potential intrusiveness of installing sensors on equipment, sensor data is often more sporadically available and costly to obtain compared to audio data.

[0020] Figure 1 is a diagram of a framework 100 to detect a state of one or more pieces of equipment based on audio data produced by the one or more pieces of equipment, according to one or more example implementations. The framework 100 can include a piece of equipment 102. The piece of equipment 102 can be located in at least one of a manufacturing facility or an industrial facility. The piece of equipment 102 can be located proximate to one or more additional pieces of equipment. In addition, the piece of equipment 102 can operate in conjunction with one or more additional pieces of equipment. In various examples, the piece of equipment 102 can operate to cause the movement of at least one of one or more liquids, one or more gases, or one or more solid materials within the facility.

[0021] In one or more illustrative examples, the piece of equipment 102 can include a pump. In one or more additional illustrative examples, the piece of equipment 102 can include a motor. In one or more further illustrative examples, the piece of equipment 102 can include a generator. In still other illustrative examples, the piece of equipment 102 can include an industrial tool or appliance used in the production of products and / or in the transport of materials through the facility. In at least some examples, the piece of equipment 102 can include at least one part 104 that rotates during operation of the piece of equipment 102. In various examples, rotating components of the piece of equipment 102 can include at least one of gears, bearings, or shafts. During operation of the piece of equipment 102, sound 106 can be produced. The sound 106 can be produced by movement of one or more parts of the piece of equipment 102. In addition, the sound 106 can be produced by movement of one or more substances through the piece of equipment 102.

[0022] The sound 106 can be captured by one or more microphones 108. The one or more microphones 108 can be located in one or more computing devices 110. The one or more computing devices 110 can include one or more mobile computing devices, one or more smartphones, one or more table computing devices, one or more laptop computing devices, one or more desktop computing devices, one or more wearable devices, or one or more combinations thereof. In one or more examples, the one or more computing devices 110 can produce a data file 112 that corresponds to the sound 106. In various examples, the data file 112 can include audio data that corresponds to the sound 106. In one or more additional examples, the data file 112 can include audio data that corresponds to the sound 106 as well as video captured by the one or more computing devices 110. In at least some examples, the one or more computing devices 110 can capture video using one or more cameras and audio corresponding to the video using the one or more microphones 108. In these scenarios, the one or more computing devices 110 can separate the audio data from the video data and produce the data file 112 such that the data file 112 includes the audio data and not the video data.

[0023] The one or more computing devices 110 can send the data file 112 to a computational system 114. The computational system 114 can be implemented by one or more computing devices 116. The one or more computing devices 116 can include one or more server computing devices, one or more desktop computing devices, one or more laptop computing devices, one or more tablet computing devices, one or more mobile computing devices, or combinations thereof. In one or more implementations, at least a portion of the one or more computing devices 116 can be implemented in a distributed computing environment. For example, at least a portion of the one or more computing devices 116 can be implemented in a cloud computing architecture.

[0024] The computational system 114 can be coupled to or otherwise in electronic communication with an audio file data store 118. The audio file data store 118 can include one or more databases that store a number of audio data files 120. In one or more examples, the computational system 114 can cause the data file 112 to be stored by the audio file data store 118 after receiving the data file 112 from the one or more computing devices 110. In scenarios where the data file 112 includes audio data and video data, in at least some examples, the computational system 114 can separate the audio data from the video data and store the audio data in the audio file data store 118.

[0025] Additionally, the computational system 114 can be coupled to or otherwise in electronic communication with an equipment data store 122. The equipment data store 122 can store equipment data 124 that includes information about one or more pieces of equipment located in one or more facilities. The equipment data 124 can include physical specifications with regard to equipment. For example, the equipment data 124 can indicate parts included in anumber of pieces of equipment and physical dimensions of the parts. In at least some examples, the physical dimensions of the parts included in the equipment data 124 can include geometries of parts included in the equipment data 124. Further, the equipment data 124 can include operating specifications for equipment. The operating specifications included in the equipment data 124 can include speed of movement of one or more parts of equipment, flow rates of at least one of liquids or gases through equipment, power used by equipment, power generated by equipment, one or more combinations thereof, and the like. In at least some examples, the equipment data 124 can indicate at least one of tolerances or thresholds for at least one of physical specifications or operational specifications of equipment.

[0026] In one or more illustrative examples, the piece of equipment 102 can include a pump and the equipment data 124 can indicate information related to at least one of bearings or gears included in the piece of equipment. In various examples, the equipment data 124 can indicate inner diameters and outer diameters of bearings included in the piece of equipment 102. The equipment data 124 can also indicate sizes, such as diameters, of balls included in bearings of the piece of equipment. In one or more additional examples, the equipment data 124 can indicate a range of rotational speeds of at least one of bearings or gears included in piece of equipment 102 during expected operation of the piece of equipment. In various examples, the equipment data 124 can indicate a target rotational speed and a tolerance for the target rotational speed for at least one of the bearings, rotors, or gears during operation of the piece of equipment 102. In at least some examples, the tolerance can indicate an amount of deviation from the target rotational speed that enables proper operation of the piece of equipment 102.

[0027] In one or more examples, the audio file data store 118 and the equipment data store 122 can be accessible to the computational system 114 by at least one of one or more wireless communication networks or one or more wired communication networks. In one or more additional examples, at least one of the audio file data store 118 or the one or more equipment data stores 122 can be remotely located with respect to the computational system 114 and / or located remotely with respect to a facility that includes the piece of equipment 102. Additionally, at least one of the audio file data store 118 or the equipment data store 122 can be local data storage, such as on-premises data storage, in relation to the computational system 114 and / or a facility that includes the piece of equipment 102. In various examples, at least one of the audio file data store 118 or the equipment data store 122 can be maintained by a same entity that implements the computational system 114. Further, at least one of the audio file data store 118 or the equipment data store 122 can be maintained by a different entity than the entitythat maintains, operates, and / or administers the computational system 114. In at least some examples, at least one of the audio file data store 118 or the equipment data store 122 can include one or more publicly accessible databases. At least one of the audio file data store 118 or the equipment data store 122 can also include one or more privately maintained databases. In one or more further examples, the data stored by at least one of at least one of the audio file data store 118 or the equipment data store 122 can be accessed using one or more websites.

[0028] The computational system 114 can perform audio data preprocessing 126. The audio data preprocessing 126 can include performing, by the computational system 114, computational operations that modify audio data included in an audio data file 120. For example, the computational system 114 can perform a time domain analysis with respect to an audio data file 120. The time domain analysis can include performing an envelope analysis with respect to amplitude values of sound waves having one or more frequencies included in the audio data file 120. Additionally, the computational system 114 can perform, in relation to the audio data preprocessing 126, a frequency domain analysis with respect to an audio data file 120. In one or more examples, the frequency domain analysis can include implementing a fast Fourier transform (FFT) algorithm with respect to an audio data file 120. In still other examples, the audio data preprocessing 126 can include performing a time-frequency domain analysis with respect to an audio data file 120. In one or more illustrative examples, the timefrequency domain analysis can include a sparse fast Fourier transform (sFFT) analysis in relation to an audio data file 120. In various examples, the audio data preprocessing 126 can produce a modified version of the audio data that is included in an audio data file 120.

[0029] In one or more examples, the audio data preprocessing 126 can include determining one or more frequency ranges on which to perform an envelope analysis. In various examples, the audio data preprocessing 126 can include performing a spectral kurtosis analysis to determine frequencies that can be analyzed to identify signals that correspond to faults being present in equipment. In one or more illustrative examples, the audio data preprocessing 126 can include a fast kurtogram analysis to determine one or more frequency bands of the audio data included in the audio data file 120 to perform an envelope analysis that includes at least one of the time domain analysis, the frequency domain analysis, or the time-frequency domain analysis. In at least some examples, the audio data preprocessing 126 can include applying one or more frequency filters to extract data from the audio data file 120 that corresponds to the one or more frequency bands determined by the spectral kurtosis analysis. The filtered data can then be subjected to at least one of the time domain analysis, the frequency domain analysis,or the time-frequency domain analysis to generate modified audio data from the audio data file 120.

[0030] The modified audio data produced by the audio data preprocessing 126 can undergo additional processing by the computational system 114. For example, the computational system 114 can perform equipment state detection 128 using a modified version of the audio data included in an audio data file 120 that is produced by the audio data preprocessing 126. In one or more examples, the equipment state detection 128 performed by the computational system 114 can include determining one or more faults taking place in relation to the piece of equipment 102. For example, the equipment state detection 128 can determine a state of the piece of equipment 102 that includes, but is not limited to, bearing faults, gear faults, misalignment of rotational components, unbalanced operation, cavitation, one or more combinations thereof, and the like by computationally analyzing the information related to the data file 112. In one or more illustrative examples, the equipment state detection 128 can determine faults of the piece of equipment 102 that include looseness of components of the piece of equipment 102, overpressurization, dry running, air leaks, seal failures, drive line failures, one or more combinations thereof, and the like. Overpressurization can correspond to the piece of equipment 102 operating at pressures that are outside of a specified range of operating pressures. Dry running can include operating with less than a minimal amount of fluid in one or more components of the piece of equipment 102. Air leaks can include gases, such as air, being pulled into a suction line of the piece of equipment 102. Seal failures can correspond to degradation of performance of mechanical seals of the piece of equipment 102. To illustrate, degradation of performance of seals can result in at least one of fluid or gas moving through the seal material. Drive line failures can correspond to a failure of a universal joint of the piece of equipment 102.

[0031] The computational system 114 can implement one or more machine learning techniques to determine a state of the piece of equipment 102. For example, the equipment state detection 128 performed by the computational system 114 can include executing a predictive machine learning model to predict a state of the piece of equipment 102. To illustrate, the computational system 114 can execute a generative Al model to computationally analyze information related to one or more audio data files 120 that include sound produced by the piece of equipment 102 to determine a state of the piece of equipment 102. In one or more illustrative examples, the computational system 114 can execute a transformer-based machine learning model to analyze one or more audio data files 120 to determine a state of the piece of equipment 102.

[0032] In addition to analyzing audio data included in the data file 112, the computational system 114 can also, at 128, analyze the equipment data 124 to determine a state of equipment, such as the piece of equipment 102. For example, the equipment state detection 128 can include analyzing dimensions of parts of equipment 102 in conjunction with audio data included in audio data files 120 to determine a state of equipment. In these scenarios, the equipment data 124 can include dimensions of gears, bearings, rotors, one or more combinations thereof, and the like, that are analyzed in addition to information included in the audio data files 120 to determine the state of equipment. Further, the computational system 114 can analyze operational specifications included in the equipment data 124 along with audio data included in the audio data files 120 to determine a state of equipment. In situations where the equipment state detection 128 includes implementation of one or more machine learning models, the one or more machine learning models can be trained using audio data and the equipment data 124 corresponding to one or more pieces of equipment. In this way, the input provided to the one or more machine learning models to determine a state of a piece of equipment 102 can include audio data generated by the piece of equipment 102 and at least a portion of the equipment data 124 corresponding to the piece of equipment 102.

[0033] The computational system 114 can also perform equipment state classification 130. The equipment state classification 130 can include determining a classification and / or a type of faults that can occur with equipment. In one or more examples, the computational system 114 can implement one or more probabilistic classification models to determine classifications of faults of pieces of equipment. In various examples, the one or more probabilistic classification models can include at least one of one or more statistical models or one or more machine learning models. The computational system 114 can execute one or more probabilistic models with respect to the equipment state classification 130 that have multiple outputs. In these instances, the multiple outputs can include probabilities of the different types of faults that can occur with respect to pieces of equipment. For example, the equipment state classification 130 can determine a first probability for a first type of fault taking place with respect to a piece of equipment 102, a second probability for a second type of fault taking place with respect to the piece of equipment 102, and a third probability for a third type of fault taking place with respect to the piece of equipment 102.

[0034] In various examples, the equipment state classification 130 can be performed based on audio data included in the audio data files 120. In at least some examples, the equipment state classification 130 can be performed using audio data obtained from the audio data files 120that has been modified according to the audio data preprocessing 126. Additionally, the equipment state classification 130 can also be performed based on the equipment data 124, such as dimensions of parts and operating specifications for one or more pieces of equipment.

[0035] In one or more additional examples, at least one of the equipment state detection 128 or the equipment state classification 130 can also be performed using additional data captured by one or more additional sensors. For example, one or more sensors can be located on or near a number of pieces of equipment. The one or more sensors can include vibrational sensors, accelerometers, temperature sensors, flow rate sensors, motion sensors, one or more combinations thereof, and so forth. In this way, audio data can be combined with additional sensor data to determine a state of a piece of equipment 102 and / or a classification of a state of the piece of equipment 102 based on one or more computational analyses performed by the computational system 114.

[0036] In one or more illustrative examples, the computational system 114 can cause a number of user interfaces to be generated by computing devices of users of the computational system 114. The users of the computational system 114 can be part of an entity that at least one of administers, maintains, implements, or controls the computational system 114. In one or more additional examples, the users of the computational system 114 can include individuals or entities outside of the entity that at least one of administers, maintains, implements, or controls the computational system 114. In one or more illustrative examples, at least a portion of the users of the computational system 114 can be customers of the entity that at least one of administers, maintains, implements, or controls the computational system 114. In various examples, the computational system 114 can generate user interface data that can be accessible to one or more computing devices and the one or more computing devices can process the user interface data to display user interfaces that include information generated by the computational system 114. In one or more illustrative examples, the user interface data can be generated in conjunction with at least one of a browser application or a user device app that is executed by the one or more computing devices of users of the computational system 114. In one or more additional illustrative examples, the user interface data produced by the computational system 114 can correspond to a dashboard that is accessible to users of the computational system 114 via one or more computing devices.

[0037] The user interfaces generated in relation to the computational system 114 can correspond to the operations performed by the computational system 114 and the functionality of the computational system 114. For example, the computational system 114 can generate firstuser interface data that corresponds to a first user interface 132. The first user interface 132 can include one or more user interface elements that can capture input that is related to the analysis of audio produced by equipment. To illustrate, the first user interface 132 can include one or more user interface elements that can be used to select one or more audio data files 120 to be analyzed by the computational system 114 to determine a state of a piece of equipment that produced the sound included in the selected one or more audio data files 120. The one or more audio data files 120 that are selected can include audio data files that have been newly captured and stored in the audio file data store 118 and / or audio data files of previously captured audio stored by the audio file data store 118. In one or more additional examples, the first user interface 132 can include one or more user interface elements that can capture input indicating information related to pieces of equipment that correspond to the selected audio data files 120. In one or more illustrative examples, one or more user interface elements of the first user interface 132 can capture information corresponding to characteristics of parts of pieces of equipment, such as dimensions of parts of equipment. In still other examples, the one or more user interface elements of the first user interface 132 can capture information corresponding to operational specifications of pieces of equipment.

[0038] In various examples, the first user interface 132 can include one or more user interface elements that are selectable to capture data that the computational system 114 can process to perform a computational analysis in relation to a state of the selected piece of equipment 102. For example, the first user interface 132 can include one or more user interface elements corresponding to selecting a piece of equipment 102 for which the computational system 114 is to determine its state. In at least some examples, in response to selection of the piece of equipment 102 via a user interface element of the first user interface 132, the first user interface 132 can include one or more additional user interface elements that can include pre-populated options for selection. The one or more pre-populated options can correspond to parts, dimensions of parts, operational specifications, one or more combinations thereof, and the like that correspond to the selected piece of equipment 102. In still other examples, the first user interface 132 can include user interface elements that can capture information indicating one or more types of analysis to be performed by the computational system 114 with respect to audio of the selected piece of equipment 102. For example, the first user interface 132 can include one or more user interface elements directed to selecting one or more audio data preprocessing techniques to be performed in relation to the audio data preprocessing 126 with respect to one or more audio data files 120 corresponding to the piece of equipment 102.

[0039] The computational system 114 can also generate user interface data corresponding to a second user interface 134. The second user interface 134 can include modified audio data generated in conjunction with the audio data preprocessing 126. To illustrate, the second user interface 134 can display at least one of one or more graphs, one or more charts, one or more diagrams, or one or more additional visual aids showing a modified version of the audio data included in one or more audio data files 120 that include audio produced by a piece of equipment 102. In one or more illustrative examples, the second user interface 134 can display data related to a time domain analysis performed by the computational system 114 of audio produced by the piece of equipment 102. In one or more additional illustrative examples, the second user interface 134 can display data related to a frequency domain analysis performed by the computational system 114 of audio produced by the piece of equipment 102. In one or more further illustrative examples, the second user interface 134 can display data related to a time-frequency domain analysis performed by the computational system 114 of audio produced by the piece of equipment 102. In still other illustrative examples, the second user interface 134 can display data related to one or more kurtosis analyses performed by the computational system 114 in relation to audio produced by the piece of equipment 102. In various examples, the second user interface 134 can display at least one of a frequency domain analysis, a time domain analysis, or a time-frequency domain analysis performed by the computational system 114 after applying a filter determined by the one or more kurtosis analyses. In at least some examples, multiple versions of the second user interface 134 can be displayed with individual versions of the second user interface 134 corresponding to an individual technique applied by the computational system 114 to audio data generated by the piece of equipment 102.

[0040] The computational system 114 can also generate user interface data corresponding to a third user interface 136. The third user interface 136 can display information related to results of operations performed by the computational system 114 in relation to the equipment state detection 128. For example, the third user interface 136 can display one or more states of one or more pieces of equipment for which the audio has been analyzed by the computational system 114. In one or more illustrative examples, the third user interface 136 can display information indicating that a fault is present with respect to a piece of equipment 102 or that a fault is not present with respect to the piece of equipment 102 based on output produced in relation to the equipment state detection 128 performed by the computational system 114.

[0041] Additionally, the computational system 114 can generate user interface data corresponding to a fourth user interface 138. The fourth user interface 138 can displayinformation indicating a classification of a fault or other issue related to pieces of equipment that is identified by an analysis of audio produced by the pieces of equipment. To illustrate, the fourth user interface 138 can display one or more types of fault that can be present with respect to a piece of equipment 102. In various examples, the fourth user interface 138 can indicate probabilities in relation to the number of types of possible faults present with respect to the piece of equipment 102. In at least some examples, the information displayed in the fourth user interface 138 can correspond to output produced by the equipment state classification 130 performed by the computational system 114.

[0042] Although the user interfaces 132, 134, 136, 138 have been described such that various functions and / or information are attributed to the individual user interfaces 132, 134, 136, 138, in one or more additional implementations, at least a portion of the functions and / or information attributed to one of the user interfaces 132, 134, 136, 138 can be combined with at least a portion of the functions and / or information attributed to another one of the user interfaces 132, 134, 136, 138. Further, although the user interfaces 132, 134, 136, 138 have been described such that various functions and / or information are attributed to the individual user interfaces 132, 134, 136, 138, the computational system 114 can also generate user interface data corresponding to one or more further user interfaces that include or are otherwise related to at least a portion of the functions and / or information described in relation to user interfaces 132, 134, 136, 138.

[0043] In at least some examples, information can be provided to one or more components of the computational system 114 by input captured via one or more user interfaces generated by the computational system 114. The input captured via the one or more user interfaces can include at least one of text input, audio input, graphical input, or video input. In one or more scenarios, the input captured via the one or more user interfaces can be at least one of text input, audio input, graphical or video input produced over one or more periods of time. In various examples, the input captured by the one or more user interfaces can include one or more prompts provided to one or more generative machine learning architectures executed by the computational system 114. In one or more examples, the one or more user interfaces can provide one or more templates for capturing one or more types of input information. For example, one or more user interfaces can provide a framework for capturing input that can be processed by one or more components of the computational system 114. In one or more additional examples, the computational system 114 can provide an artificial intelligence-based agent to interact with users and obtain input for components of the computational system 114.To illustrate, an artificial intelligence-based agent can guide users to provide prompts having specified types of information that can be processed by one or more components of the computational system 114.

[0044] In various examples, prompts can be provided to one or more generative machine learning models, such as one or more large language models, in relation to the data input to the one or more generative machine learning models and / or data output by the one or more generative machine learning models. For example, prompts can be provided for the one or more generative machine learning models to analyze frequency information and / or an analysis of actual frequency peaks in relation to expected frequency peaks to determine one or more failure modes for pieces of equipment, causes of the failure for pieces of equipment, and potential solutions to resolve the failure experienced by pieces of equipment. Prompts can also be provided to one or more generative machine learning models that include frequency profiles or other audio data profiles that correspond to predetermined failure modes for one or more pieces of equipment. Additionally, prompts can be provided for the one or more generative machine learning models to provide a summary of equipment data, audio data, and / or modified audio data provided as input to one or more generative machine learning models. Further, prompts can be to provide observations related to at least one of equipment data, audio data, or modified audio data provided as input to one or more generative machine learning models. In at least some examples, prompts can be provided to cause one or more generative machine learning models to retrieve information from at least one of one or more publicly available data sources or one or more private data sources to perform operations related to the equipment state detection 128 or the equipment state classification 130.

[0045] In still other examples, prompts can be provided to one or more generative machine learning models to provide insights related to equipment states determined by the one or more generative machine learning models. To illustrate, prompts can be provided to cause one or more generative machine learning models to provide evidence for a determination of a state of a piece of equipment. That is, prompts can be provided to cause one or more generative machine learning models to provide evidence for a diagnosis of the state of a piece of equipment. In one or more additional examples, one or more prompts can be provided that cause one or more generative machine learning models to produce information related to harmonics on fundamental rotational frequencies for pieces of equipment and / or information related to harmonics of other characteristics frequencies of pieces of equipment. In still other examples, prompts can be provided that cause one or more generative machine learning modelsto determine changes in audio frequency characteristics of pieces of equipment over one or more periods of time.

[0046] In one or more illustrative examples, user interfaces that capture input for one or more generative machine learning architectures executed by the computational system 114 can obtain a modified version of audio data included in the audio data file 120 that is produced after the audio data preprocessing 126. For example, one or more user interfaces of the computational system 114 can capture modified audio data generated in response to a time domain analysis performed by the computational system 114. In addition, the one or more user interfaces of the computational system 114 can capture modified audio data generated in response to a frequency domain analysis performed by the computational system 114. Further, the one or more user interfaces of the computational system 114 can capture modified audio data generated in response to a time-frequency domain analysis performed by the computational system 114. In various examples, the one or more user interfaces of the computational system 114 can capture modified audio data generated in response to a matched characteristic frequency analysis where frequencies corresponding to the audio data included in the audio data file 120 are analyzed with respect to one or more expected frequencies for a given piece of equipment. In at least some examples, the modified audio data and / or the matched frequency analysis information can be provided to at least one of the equipment state detection 128 components of the computational system 114 or the equipment state classification components of the computational system 114 via one or more prompts to a generative machine learning model. In one or more scenarios, the modified audio data and / or the matched frequency analysis information can be provided to at least one of the equipment state detection 128 components of the computational system 114 or the equipment state classification components of the computational system 114 based on one or more interactions with an artificial intelligence-based agent.

[0047] Figure 2 is a diagram of a computational architecture 200 that performs time domain and frequency domain processing of audio produced by one or more pieces of equipment to determine a state of the one or more pieces of equipment, according to one or more example implementations. The computational architecture 200 can include an audio data file 202. The audio data file 202 can include data corresponding to sound produced by one or more pieces of equipment located in one or more facilities. In one or more examples, at least a portion of the operations performed by the computational architecture 200 with respect to the audio data file 202 can be performed by the computational system 114 described with respect to Figure 1.

[0048] The information included in the audio data file 202 can be subjected to one or more preprocessing operations. For example, a time domain analysis 204 can be performed with respect to the audio data file 202. The time domain analysis 204 can include amplitude modulation of sound waves represented by the information included in the audio data file 202. In various examples, the time domain analysis 204 can include performing an envelope analysis that includes combining signals from one or more frequencies and converting the resulting data back to the time domain. In one or more examples, the time domain analysis 204 can include an amplitude envelope analysis that filters at least a portion of the information included in the audio data file 202. To illustrate, an amplitude envelope analysis can highlight one or more portions of the signal included in the audio data file 202 that are distinctive from other portions of the signal included in the audio data file 202.

[0049] In one or more additional examples, a frequency domain analysis 206 can be performed with respect to the audio data file 202. The frequency domain analysis 206 can include identifying one or more frequencies that are indicative of an anomaly in the operation of the piece of equipment and / or that indicate one or more features that are distinctive in relation to the information related to other frequencies. In one or more illustrative examples, the frequency domain analysis 206 can include performing implementing a fast Fourier transform (FFT) algorithm. In various examples, the FFT algorithm can convert a time-based signal included in the audio data file 202 to a frequency domain signal. In at least some examples, the FFT algorithm can determine a Discrete Fourier Transform (DFT) or an inverse of the DFT with respect to the data included in the audio data file 202.

[0050] In still other examples, a time-frequency domain analysis 208 can be performed with respect to the audio data file 202. In one or more illustrative examples, the time-frequency domain analysis 208 can include implementing a sparse fast Fourier transform (SFFT) algorithm with respect to the audio data file 202. In one or more examples, the SFFT algorithm can be implemented at 208 with respect to the audio data file 202 because the information included in the audio data file 202 can be represented by a relatively large number of Fourier coefficients having a small value or a zero value. In these situations, the signal of the audio data file 202 can be represented by the non-zero value and / or non-small value Fourier coefficients. Some illustrative examples of SFFT algorithms are described in J. Schumacher and M. Piischel, "High-performance sparse fast Fourier transforms," 2014 IEEE Workshop on Signal Processing Systems (SiPS), Belfast, UK, 2014, pp. 1-6, doi: 10.1109 / SiPS.2014.6986055.

[0051] The output of the time domain analysis, the frequency domain analysis 206, and the time-frequency domain analysis 208 can include modified audio data 210. In one or more examples, the time domain analysis 204, the frequency domain analysis 206, and the timefrequency domain analysis 208 can transform or otherwise modify information included in the audio data file 202 to be a different version of the information included in the audio data file 202 and / or a different type of the information included in the audio data file 202. In at least some examples, the modified audio data 210 can be stored in one or more additional data files. For example, the time domain analysis 204 can produce a first additional audio data file that includes a first portion of the modified audio data 210, the frequency domain analysis 206 can produce a second additional audio data file that includes a second portion of the modified audio data 210, and the time-frequency domain analysis 208 can produce a third additional audio data file that includes a third portion of the modified audio data 210. In one or more additional examples, the modified audio data 210 produced by each of the time domain analysis 204, the frequency domain analysis 206, and the time-frequency domain analysis 208 can be combined into a single data file.

[0052] In one or more illustrative examples, the computational architecture 200 can optionally include frequency band identification 212. The frequency band identification 212 can include determining one or more frequency ranges corresponding to the signal information stored in the audio data file 202 that can be indicative of a state of equipment. For example, the frequency band identification 212 can include determining one or more frequency ranges that correspond to signal information indicative of a fault being present with respect to equipment. In various examples, the frequency band identification 212 can include performing a spectral analysis in relation to frequencies of one or more signals included in the audio data file 202. In at least some examples, the frequency band identification 212 can include performing a kurtosis analysis of information included in the audio data file 202. To illustrate, the frequency band identification 212 can include generating a fast kurtogram for one or more frequency windows having one or more sizes based on the information included in the audio data file 202. In these scenarios, the fast kurtogram can indicate a center frequency and a bandwidth around the center frequency that are most likely to correspond to information indicating a state of one or more pieces of equipment.

[0053] In implementations of the computational architecture 200 that include the frequency band identification 212, the computational architecture 200 can include one or more filters 214. The one or more filters 214 can be configured to modify signals included in the audio data file202 such that information corresponding to the one or more frequency ranges determined by the frequency band identification 212 is preserved, while removing information that does not correspond to the one or more frequency ranges determined by the frequency band identification 212. In one or more examples, filtered data 216 is produced by applying the one or more filters 214 to information of the audio data file 202 and the filtered data 216 can be modified by at least one of the time domain analysis 204, the frequency domain analysis 206, or the time-frequency domain analysis 208. In various examples, the modified audio data 210 can include the filtered data 216 after being processed according to at least one of the time domain analysis 204, the frequency domain analysis 206, or the time-frequency domain analysis 208. In one or more additional examples, the modified audio data 210 can include both filtered data 216 and unfiltered data obtained directly from the audio data file 202 that have been subjected to at least one of the time domain analysis 204, the frequency domain analysis 206, and the time-frequency domain analysis 208.

[0054] The computational architecture 200 can include a state detection analysis system 218 that analyzes the modified audio data 210 to determine a state of one or more pieces of equipment. The state detection analysis system 218 can also analyze the modified audio data 210 to determine a classification of a state of one or more pieces of equipment. In one or more illustrative examples, the state detection analysis system 218 can include perform the equipment state detection 128 and the equipment state classification 130 described in relation to Figure 1.

[0055] The state detection analysis system 218 can be coupled to or otherwise have electronic access to an equipment data store 220. In one or more examples, the equipment data store 220 can correspond to the equipment data store 122 described with respect to Figure 1. The equipment data store 220 can store equipment data 222. The equipment data 222 can include data corresponding to at least one of physical characteristics of one or more pieces of equipment, such as physical characteristics of one or more parts of the one or more pieces of equipment. The equipment data 222 can also include data corresponding to operational characteristics of the one or more pieces of equipment.

[0056] In one or more examples, the state detection analysis system 218 can perform a matched characteristic frequency analysis 224 for one or more pieces of equipment. The matched characteristic frequency analysis 224 can include analyzing the equipment data 222 for a given piece of equipment to determine one or more expected frequencies for sound produced by the piece of equipment. For example, the matched characteristic frequency analysis 224 candetermine one or more expected frequencies for a piece of equipment according to dimensions of at least one of bearings, rotors, or gears of the piece of equipment and one or more rotational speeds for parts of the piece of equipment. The one or more rotational speeds for the parts of the piece of equipment can include recommended or expected rotational speeds indicated in operating specifications included in the equipment data 222 for the piece of equipment.

[0057] The one or more expected frequencies can be analyzed in relation to at least a portion of the modified audio data 210. For example, the state detection analysis system 218 can analyze the one or more expected frequencies in relation to portions of the modified audio data 210 that correspond to the filtered data 216 after being subjected to at least one of the time domain analysis 204, the frequency domain analysis 206, or the time-frequency domain analysis 208. In various examples, the matched characteristic frequency analysis 224 can include determining differences between the one or more expected frequencies determined using the predetermined equipment data 222 and frequencies included in the modified audio data 210. In one or more illustrative examples, the one or more expected frequencies determined based on the equipment data 222 can correspond to one or more peak frequencies expected to be produced by the piece of equipment. In these situations, the matched characteristic frequency analysis 224 can include determining differences between one or more expected peak frequencies and one or more actual peak frequencies indicated by the modified audio data 210. In at least some examples, the matched characteristic frequency analysis 224 can determine a state of a piece of equipment. To illustrate, the matched characteristic frequency analysis 224 can determine that the expected characteristic frequency peak corresponding to each component is present among the actual peaks identified by frequencydomain analysis of the audio data. These matched frequencies serve as indicators of possible failures within each component. With the physics knowledge from bearing / gear failure analysis, each failure mode can be related to peaks values of one or more matched frequencies or the order of their peak values. On the other hand, by discretizing the Short-Time-Fourier- Transform spectrogram, a peak value at a matched frequency can be converted to a statistical distribution. A physics-based probabilistic failure detection model can be used to further relate a failure mode to the distribution of peak values or relative values of multiple distributions and calculate the probability of a failure by statistically comparing the distribution of peak values with a threshold value identified based audio data obtained from equipment operating without detectable fault or defect.

[0058] The state detection analysis system 218 can also perform machine learning detection and classification 226. The machine learning detection and classification 226 can include analyzing at least a portion of the modified audio data 210 in relation to one or more machine learning models. In one or more examples, the machine learning detection and classification 226 performed by the state detection analysis system 218 can include analyzing the raw data included in the audio data file 202 after being subjected to at least one of the time domain analysis 204, the frequency domain analysis 206, or the time-frequency domain analysis 208 and analyzing the filtered data 216 after being subjected to at least one of the time domain analysis 204, the frequency domain analysis 206, or the time-frequency domain analysis 208. The machine learning detection and classification 226 can include analyzing at least a portion of the modified audio data 210 to determine a state of a piece of equipment. For example, the machine learning detection and classification 226 can implement one or more machine learning models with respect to at least a portion of the modified audio data 210 to determine whether or not a fault is present with respect to a piece of equipment. The machine learning detection and classification 226 can also implement one or more machine learning models or one or more statistical models with respect to at least a portion of the modified audio data 210 to determine one or more classifications that correspond to a fault with respect to the piece of equipment.

[0059] In one or more illustrative examples, the machine learning detection and classification 226 can be implemented by executing a number of machine learning models. For example, the machine learning detection and classification 226 can be implemented by executing one or more generative machine learning models. In addition, the machine learning detection and classification 226 can be implemented by executing a k-nearest neighbors machine learning classifier. Further, the machine learning detection and classification 226 can be implemented by executing a random forests classifier. The machine learning detection and classification 226 226 can also be implemented by executing one or more artificial neural networks. In still other examples, the machine learning detection and classification 226 can be implemented by executing a Naive-Bayes classifier. In various additional examples, the machine learning detection and classification 226 can be implemented by executing a logistic regression classifier. In situations where the machine learning detection and classification 226 is performed by executing one or more generative machine learning models, information can be provided to the one or more generative machine learning models via one or more prompts captured by one or more user interfaces.

[0060] In one or more examples, the state detection analysis system 218 can generate system output 228. The system output 228 can include an equipment state 230. The system output 228 can also include an equipment state classification 232. In various examples, the equipment state 230 can indicate that a fault is present or that a fault is not present with respect to a piece of equipment. The equipment state 230 can be produced by at least one of the matched characteristic frequency analysis 224 or the machine learning detection and classification 226 performed by the state detection analysis system 218. Additionally, the equipment state classification 232 can indicate one or more classification of one or more faults present with respect to a piece of equipment. The equipment state classification 232 can be determined by the machine learning detection and classification 226 performed by the state detection analysis system 218. In at least some examples, the equipment state classification 232 can indicate one or more probabilities of one or more faults being present with respect to a piece of equipment. In one or more illustrative examples, the system output 228 can be displayed in one or more user interfaces. In one or more illustrative examples, the system output 228 can include at least one of text data, video data, graphical data, or audio data generated by the state detection analysis system 218 in response to one or more prompts provided to one or more generative machine learning models.

[0061] Figure 3 is a diagram of a computational framework 300 that implements a generative Al model to analyze audio data produced by one or more pieces of equipment to determine a state of the one or more pieces of equipment, according to one or more example implementations. The computational framework 300 can include an equipment state detection system 302. In one or more examples, the equipment state detection system 302 can perform at least a portion of the operations of the equipment state detection 128 described in relation to Figure 1 and the state detection analysis system 218 described in relation to Figure 2. The equipment state detection system 302 can implement one or more machine learning techniques to analyze audio data 304. The audio data 304 can correspond to sound produced by one or more pieces of equipment. In various examples, the audio data 304 can include sound produced by one or more pieces of equipment that is captured by one or more audio sensors. In one or more illustrative examples, the one or more audio sensors can include one or more microphones.

[0062] Before being analyzed by the equipment state detection system 302, the audio data 304 can be subjected to audio data preprocessing 306. In various examples, the audio data preprocessing 306 can include at least one of a time domain analysis of the audio data 304, afrequency domain analysis of the audio data 304, or a time-frequency domain analysis of the audio data 304. In one or more additional examples, the audio data preprocessing 306 can include identifying one or more frequency ranges of the signals included in the audio data 304 and performing at least one of a time domain analysis, a frequency domain analysis, or a timefrequency domain analysis of the portions of the audio data 304 corresponding to the identified one or more frequency ranges. For example, fast kurtosis computations can be performed to determine the one or more frequency ranges of the audio data 304 for which the time domain analysis, the frequency domain analysis, and / or the time-frequency domain analysis can be performed. In one or more illustrative examples, the audio data preprocessing 306 can include at least one of an amplitude envelope analysis, a fast Fourier transform analysis, or a sparse fast Fourier transform analysis. In one or more additional illustrative examples, the audio data preprocessing 306 can include the audio data preprocessing 126 described with respect to Figure 1 and / or the time domain analysis 204, the frequency domain analysis 206, the timefrequency domain analysis 208, the frequency band identification 212, and the one or more filters 214 described in relation to Figure 2.

[0063] The audio data preprocessing 306 can produce modified audio data 308. The modified audio data 308 can include one or more portions of the audio data 304 after being subjected directly to at least one of a time domain analysis, a frequency domain analysis, or a timefrequency domain analysis. In one or more examples, the modified audio data 308 can also include one or more portions of the audio data 304 that have been filtered, such as filtered based on a fast Kurtogram analysis, and then been subjected to at least one of a time domain analysis, a frequency domain analysis, or a time-frequency domain analysis. In at least some examples, the modified audio data 308 can include at least a portion of the modified audio data 210 as described in relation to Figure 2. In various examples, the modified audio data 308 can include results of at least one of a time domain analysis, a frequency domain analysis, or a time-frequency domain analysis. The results of the at least one of a time domain analysis, a frequency domain analysis, or a time-frequency domain analysis can be provided in the form of at least one of graphs, tables, text recommendations, or other visual, graphical, or audiobased media. In still other examples, the modified audio data 308 can include at least one of failure frequencies or harmonics for a given piece of equipment. The failure frequencies and / or harmonics can be based on user input, generated by one or more computing systems, obtained from specification information for a given piece of equipment, or one or more combinations thereof. In one or more additional examples, the modified audio data 308 can includeinformation generated by performing a matched characteristic frequency analysis where frequencies corresponding to the audio data 304 are analyzed with respect to one or more expected frequencies for a given piece of equipment.

[0064] The equipment state detection system 302 can include one or more generative Al models. In at least some examples, the equipment state detection system 302 can execute one or more large language generative machine learning models. In the illustrative example of Figure 3, the equipment state detection system 302 includes a transformer machine learning model 310. The transformer machine learning model 310 can include one or more encoding components 312 and one or more decoding components 314. The transformer machine learning model 310 can implement artificial neural network machine learning techniques.

[0065] In various examples, the modified audio data 308 can be provided as input to the transformer machine learning model 310 as sequential audio data. In addition, the equipment state detection system 302 can be electronically coupled to or otherwise access a data store 316 that stores equipment data 318. The equipment data 318 can include at least one of physical characteristics of one or more pieces of equipment or operational specifications of the one or more pieces of equipment. The equipment data 318 can also be provided as input to the transformer machine learning model 310.

[0066] In one or more further examples, the modified audio data 308 and / or the equipment data 318 can be provided to the equipment state detection system 302 via one or more prompts for the transformer machine learning model 310. In various examples, the modified audio data 308 can be part of a data package that is provided to the transformer machine learning model 310 to analyze the modified audio data 308. In addition to the modified audio data 308, one or more portions of the equipment data 318, such as information about one or more pieces of equipment that generated the audio data 304, can also be provided to the transformer machine learning model 310. In one or more illustrative examples, the equipment data 318 can also include identifiers of pieces of equipment, one or more images of pieces of equipment, manufacturers of pieces of equipment, models of pieces of equipment, operational characteristics of pieces of equipment, locations of pieces of equipment, or one or more combinations thereof.

[0067] In one or more examples, the transformer machine learning model 310 can analyze the modified audio data 308 and, optionally other information about pieces of equipment, in response to prompts input to the transformer machine learning model 310. The prompts can indicate one or more operations for the transformer machine learning model 310 to perform toanalyze the modified audio data 308. The prompts can also indicate results to be produced by the transformer machine learning model 310. The results can include one or more failure modes for pieces of equipment, one or more causes of the failure modes for pieces of equipment, solutions to remedy operational problems related to the pieces of equipment, observations related to at least one of the audio data 304, the modified audio data 308, or the equipment data 318, and / or summaries for at least one of the audio data 304, the modified audio data 308, or the equipment data 318.

[0068] The one or more encoding components 312 can include a number of computational layers. In various examples, individual computational layers of the one or more encoding components 312 can apply transformations to a number of portions of the data input to the one or more encoding components 312. Additionally, the individual layers of the one or more encoding components 312 can have weights and bias parameters that are different from the weights and bias parameters of the other individual layers of the one or more encoding components 312.

[0069] In one or more examples, individual computational layers of the one or more encoding components 312 can include a number of sublayers. For example, individual computational layers of the one or more encoding components 312 can include a sublayer that implements a multi-head self-attention mechanism. Individual computational layers of the one or more encoding components 312 can also include an additional sublayer that includes a fully- connected feed-forward network. The additional sublayer can also include at least one activation function such as rectified linear unit (ReLU). In at least some examples, the sublayers of the individual computational layers of the one or more encoding components 312 can include residual connections and be succeeded by one or more normalization layers.

[0070] The one or more decoding components 314 of the transformer machine learning model 310 can also include a number of computational layers. Individual computational layers of the one or more decoding components 314 can include a number of sublayers. To illustrate, a first sublayer of the individual computational layers of the one or more decoding components 314 can obtain output from the one or more encoding components 312, add positional information to the output of the one or more encoding components 312, and also implement multi-head self-attention. In one or more additional examples, a second sublayer of the individual computational layers of the one or more decoding components 314 can implement a multi -head self-attention mechanism and a third sublayer can implement a fully connected feed-forward network. The sublayers included in the individual computational layers of the one or moredecoding components 314 can also include residual connections and be succeeded by one or more normalization layers.

[0071] The self-attention computations performed by the one or more encoding components 312 and the one or more decoding components 314 can include inputs and outputs that are represented as at least one of vectors or matrices. In addition, the one or more encoding components 312 and / or the one or more decoding components 314 can use positional encodings indicating the position of one or more signals within a stream of data that represents the modified audio data 308. Additionally, although the modified audio data 308 can be represented as a sequence of data, the transformer machine learning model 310 can process the modified audio data 308 in a non-sequential manner.

[0072] In various examples, the transformer machine learning model 310 can be produced through a training process. The training data used to produce the transformer machine learning model 310 can include audio data obtained from pieces of equipment in which a fault and / or anomaly has been detected. The training data used to produce the transformer machine learning model 310 can also include audio data obtained from pieces of equipment in which a fault and / or anomaly has not been detected. In one or more examples, the training data for the transformer machine learning model 310 can also correspond to one or more physical characteristics for one or more pieces of equipment and / or one or more operating specifications for one or more pieces of equipment. For example, the transformer machine learning model 310 can be trained in relation to one or more models of pumps. The one or more models of pumps can have one or more bearing geometries, one or more gear dimensions, one or more rotational speeds, and so forth. In these scenarios, the audio data used to train the transformer machine learning model 310 can correspond to at least one of the one or more bearing geometries, the one or more gear dimensions, the one or more rotational speeds, or one or more other physical characteristics and / or operating specifications of the one or more models of the pumps. In one or more additional examples, the audio data for a number of pumps with different physical characteristics and / or different operational characteristics can be combined to train the transformer machine learning model 310. To illustrate, first audio data obtained from first pumps having a first bearing geometry and second audio data obtained from second pumps having a second bearing geometry can be combined to train the transformer machine learning model 310.

[0073] The training of the transformer machine learning model 310 can be performed to extract features from the audio signals included in the training data and reconstruct the original trainingdata input to the one or more encoding components 312 at the one or more decoding components. The training of the transformer machine learning model 310 can determine values for hyperparameters of the transformer machine learning model 310. For example, the training process for the transformer machine learning model 310 can indicate a number of encoding computational layers and a number of decoding computational layers to include in the transformer machine learning model 310. In addition, the training process for the transformer machine learning model 310 can indicate a number of heads to include in the multi -head attention mechanism of the layers of the transformer machine learning model 310. Further, the training process for the transformer machine learning model 310 can indicate dimensionality of input vectors and output vectors of the transformer machine learning model 310 as well as a dropout rate to indicate a number of neurons of the transformer machine learning model 310 to not include during a given iteration of the training process for the transformer machine learning model 310. In still other examples, the training process for the transformer machine learning model 310 can determine weights of parameters related to the self-attention mechanism, weights of one or more feed-forward neural networks included in the computational layers of the one or more encoding components 312, weights of one or more feed-forward neural networks included in the computational layers of the one or more decoding components 314, weights to generate encodings for the input data, one or more combinations thereof, and so forth.

[0074] The training process for the transformer machine learning model 310 can be performed in relation to a loss function for the transformer machine learning model 310. In one or more illustrative examples, various weights of the transformer machine learning model 310 are adjusted during iterations of the training process and the loss of the transformer machine learning model 310 is determined. In one or more additional illustrative examples, the training process for the transformer machine learning model 310 can be performed until the loss function is minimized or until early stopping of the training process is indicated.

[0075] The output of the transformer machine learning model 310 can include an equipment state quantitative measure 320. The equipment state quantitative measure 320 can indicate a metric in relation to a fault and / or anomaly being present with respect to a piece of equipment. In one or more examples, the equipment state quantitative measure 320 can include a score with respect to a scale that is determined by the transformer machine learning model 310. In one or more additional examples, the equipment state quantitative measure 320 can indicate a probability of a fault and / or anomaly being present in the piece of equipment. In variousexamples, a higher value of the metric can indicate a higher probability of a fault and / or anomaly being present with respect to the piece of equipment.

[0076] The equipment state detection system 302 can perform a computational analysis 322 with respect to the equipment state quantitative measure 320. For example, the equipment state detection system 302 can analyze the equipment state quantitative measure 320 with respect to a threshold quantitative measure. In situations where the equipment state quantitative measure 320 is greater than or equal to the threshold, the equipment state detection system 302 can determine that a fault is present with respect to the piece of equipment. Additionally, in scenarios where the equipment state quantitative measure 320 is less than the threshold, the equipment state detection system 302 can determine that a fault is not present with respect to the piece of equipment. In one or more examples, the equipment state detection system 302 can generate an equipment state indicator 324 based on the computational analysis 322 of the equipment state quantitative measure 320. In one or more illustrative examples, the equipment state indicator 324 can indicate that a fault is present or that a fault is not present with respect to the piece of equipment. In various examples, the equipment state indicator 324 can include a binary output indicating that a fault is present or that a fault is not present with respect to the piece of equipment. In one or more additional illustrative examples, the equipment state indicator 324 can be displayed in a user interface rendered by one or more computing devices. In one or more further examples, output from the equipment state detection system 302 that includes the equipment state indicator 324 and / or at least one of text data, video data, graphical data, or audio data can be generated by the equipment state detection system 302 in response to one or more prompts provided to one or more generative machine learning models.

[0077] Figure 4 is a diagram of a computational framework 400 that implements a machine learning classification model to determine a classification in relation to the state of one or more pieces of equipment based on audio data produced by the one or more pieces of equipment, according to one or more example implementations. The computational framework 400 can include an equipment state classification system 402. In one or more examples, the equipment state classification system 402 can perform at least a portion of the operations of the equipment state classification 130 described in relation to Figure 1. The equipment state classification system 402 can implement one or more machine learning techniques to analyze audio data 404. The audio data 404 can correspond to sound produced by one or more pieces of equipment. In various examples, the audio data 404 can include sound produced by one or more pieces ofequipment that is captured by one or more audio sensors. In one or more illustrative examples, the one or more audio sensors can include one or more microphones.

[0078] Before being analyzed by the equipment state classification system 402, the audio data 404 can be subjected to audio data preprocessing 406. In various examples, the audio data preprocessing 406 can include at least one of a time domain analysis of the audio data 404, a frequency domain analysis of the audio data 404, or a time-frequency domain analysis of the audio data 404. In one or more additional examples, the audio data preprocessing 406 can include identifying one or more frequency ranges of the signals included in the audio data 404 and performing at least one of a time domain analysis, a frequency domain analysis, or a timefrequency domain analysis of the portions of the audio data 404 corresponding to the identified one or more frequency ranges. For example, fast kurtosis computations can be performed to determine the one or more frequency ranges of the audio data 404 for which the time domain analysis, the frequency domain analysis, and / or the time-frequency domain analysis can be performed. In one or more illustrative examples, the audio data preprocessing 406 can include at least one of an amplitude envelope analysis, a fast Fourier transform analysis, or a sparse fast Fourier transform analysis. In one or more additional illustrative examples, the audio data preprocessing 406 can include the audio data preprocessing 126 described with respect to Figure 1 and / or the time domain analysis 204, the frequency domain analysis 206, the timefrequency domain analysis 208, the frequency band identification 212, and the one or more filters 214 described in relation to Figure 2.

[0079] The audio data preprocessing 406 can produce modified audio data 408. The modified audio data 408 can include one or more portions of the audio data 404 after being subjected directly to at least one of a time domain analysis, a frequency domain analysis, or a timefrequency domain analysis. In one or more examples, the modified audio data 408 can also include one or more portions of the audio data 404 that have been filtered, such as filtered based on a fast Kurtogram analysis, and then been subjected to at least one of a time domain analysis, a frequency domain analysis, or a time-frequency domain analysis.

[0080] The equipment state classification system 402 can include one or more machine learning classification models 410. In one or more illustrative examples, the one or more machine learning classification models 410 can include a Bayesian neural network. In one or more additional illustrative examples, the one or more machine learning classification models can include one or more Monte Carlo prediction algorithms. In one or more further examples, the one or more machine learning classification models 410 can include one or more KullbackLeibler (KL) divergence methods. In still other illustrative examples, the one or more machine learning classification models 410 can include one or more likelihood prediction models that can predict whether an anomaly is detected for the equipment.

[0081] In various examples, the modified audio data 408 can be provided as input to the one or more machine learning classification models 410. In addition, the equipment state classification system 402 can be electronically coupled to or otherwise access a data store 412 that stores equipment data 414. The equipment data 414 can include at least one of physical characteristics of one or more pieces of equipment or operational specifications of the one or more pieces of equipment. The equipment data 414 can also be provided as input to the one or more machine learning classification models 410.

[0082] In one or more examples, the one or more machine learning classification models 410 can computationally analyze the modified audio data 408 and, optionally, at least a portion of the equipment data 414 to determine one or more classifications for a fault of a piece of equipment and a probability associated with each individual classification. For example, the one or more machine learning classification models 410 can determine a first equipment state classification 416 and a first probability 418 that corresponds to the first equipment state classification 416. The one or more machine learning classification models 410 can also determine a second equipment state classification 420 and a second probability 422 that corresponds to the second equipment state classification 420. In one or more illustrative examples, the one or more machine learning classification models 410 can determine equipment state classifications in relation to bearings of a pump. In these scenarios, the first equipment state classification 416 can correspond to a fault with an inner race of one or more bearings of the pump and the second equipment state classification 420 can correspond to a fault with an outer race of one or more bearings of the pump. Further, in these instances, the first probability 418 can indicate a probability of a fault being present with respect to the inner race of one or more bearings of the pump and the second probability 422 can indicate an additional probability of a fault being present with respect to the outer race of one or more bearings of the pump. In various examples, output from the equipment state classification system 402 that includes at least one of the first equipment state classification 416, the first probability 418, the second equipment state classification 420, the second probability 422, or at least one of additional text data, video data, graphical data, and / or audio data can be generated by the equipment state classification system 402 in response to one or more prompts provided to one or more generative machine learning models.

[0083] Although the illustrative example of Figure 4 is described in relation to a first equipment state classification 416, a first probability 418, a second equipment state classification 420, and a second probability 422, the one or more machine learning classification models 410 can determine more or fewer equipment state classifications and respective probabilities that correspond to any number of equipment state classifications determined by the one or more machine learning classification models 410.

[0084] Figure 5 is a diagram of a computational framework 500 to determine a probability of a state of equipment by analyzing audio data in relation to expected operating frequencies, in accordance with one or more example implementations. The computational framework 500 can include a state detection analysis system 502 that performs a matched characteristic frequency analysis 504. The state detection analysis system 502 can correspond to the state detection analysis system 218 described in relation to Figure 2 and the equipment state detection system 302 described in relation to Figure 3. Additionally, the matched characteristic frequency analysis 504 can correspond to at least a portion of the matched characteristic frequency analysis 224 described in relation to Figure 2.

[0085] The matched characteristic frequency analysis 504 can include, at 506, determining an expected frequency for an equipment state. For example, the state detection analysis system 502 can determine one or more expected operational frequencies for equipment based on physical characteristics and operational characteristics of the equipment. In one or more examples, the state detection analysis system 502 can be electronically coupled to a date store 508 that stores equipment data 510. The equipment data 510 can include information about equipment, such as a type or category related to pieces of equipment, physical dimensions of components of pieces of equipment, flow rates of gas or fluid through the equipment, rotational speeds of components of equipment, other descriptive information about pieces of equipment, and so forth. The state detection analysis system 502 can analyze the equipment data 510 to determine expected frequencies for sound produced during operation of equipment when a fault, defect, or other anomaly is not present during operation of the equipment. Further, the state detection analysis system 502 can analyze the equipment data 510 to determine expected frequencies for sound produced during operation of equipment when a fault, defect, or other anomaly is present during operation of the equipment.

[0086] In one or more illustrative examples, the equipment data 510 can include bearing data 512 and gear data 514. The bearing data 512 and the gear data 514 can correspond to a number of different pump types and / or a number of different pump models. The bearing data 512 caninclude inner diameters and outer diameters of bearings and sizes, such as diameters, of balls included in bearings of one or more pumps. The gear data 514 can include number of teeth present in gears, diameters of gears, target rotational speeds of gears, one or more combinations thereof, and the like, for one or more pumps.

[0087] In one or more examples, the determination of an expected frequency for an equipment state performed at 506 can produce one or more fault frequencies 516 for equipment. The one or more fault frequencies 516 can correspond to a frequency or a range of frequencies of sound produced by equipment that indicate that a fault, defect, or other anomaly is present with respect to one or more components of the equipment. In various examples, the determination of expected frequencies for an equipment state, at 506, can include analyzing the bearing data 512 for a pump to determine ball pass frequency inner (BPFI). The state detection analysis system 502 can determine, based on the BPFI, an inner race failing frequency for the pump that corresponds to a number of balls or rollers that pass through a point of the inner race in a rotation of the pump shaft. The matched characteristic frequency analysis 504 can also include analyzing the bearing data 512 for a pump to determine ball pass frequency outer (BPFO) for the pump. The BPFO can indicate an outer race failing frequency for the pump that corresponds to the number of balls or rollers that pass through a point of the outer race in a rotation of the pump shaft. In one or more additional examples, the matched characteristic frequency analysis 504 can include analyzing the bearing data 512 for a pump to determine the ball spinning frequency (BSF) for the pump. The BSF can indicate a rolling element failing frequency for the pump that corresponds to a number of turns that a ball or roller makes in a rotation of the pump shaft. In one or more further examples, the matched characteristic frequency analysis 504 can include analyzing equipment data 510 related to the pump to determine a fundamental train frequency (FTF). The FTF can indicate a cage failing frequency that corresponds to a number of turns of a bearing cage in a rotation of the pump shaft. In still other examples, the matched characteristic frequency analysis 504 can include analyzing the gear data 514 to determine at least one of gear mesh frequency (GMF) or gear natural frequency (GNF). At least one of the GMF or the GNF can be used to determine a failure frequency and / or a range of failure frequencies for gears of a pump.

[0088] In at least some examples, the one or more fault frequencies 516 can include at least one of the BPFI, the BPFO, the BSF, the FTF, the GMF, or the GNF. In still other examples, the one or more fault frequencies 516 can be a multiple of at least one of the BPFI, the BPFO, the BSF, the FTF, the GMF, or the GNF. In one or more further examples, the one or morefault frequencies 516 can be a factor of at least one of the BPFI, the BPFO, the BSF, the FTF, the GMF, or the GNF. To illustrate, the multiple or factor of at least one of the BPFI, the BPFO, the BSF, the FTF, the GMF, or the GNF used to determine the one or more fault frequencies 516 can be produced by analyzing audio data obtained from equipment with known failures. The multiple or factor of at least one of the BPFI, the BPFO, the BSF, the FTF, the GMF, or the GNF used to determine the one or more fault frequencies 516 can be different for different pieces of equipment.

[0089] In various examples, at 518, the state detection analysis system 502 can analyze audio data at the expected frequencies at one or more time intervals. For example, audio data 520 can be generated by a piece of equipment. The audio data 520 can undergo audio data preprocessing 522 to produce modified audio data 524. In one or more additional examples, the modified audio data 524 can be produced by performing at least one of a time domain analysis with respect to the audio data 520, a frequency domain analysis with respect to the audio data 520, or a time-frequency domain analysis with respect to the audio data 520. The modified audio data 524 can be produced by performing a short time Fourier transform with respect to the audio data 520. In one or more further examples, the modified audio data 524 can be produced by performing an envelope analysis with respect to the audio data 520. In still other examples, the modified audio data 524 can be produced by performing a kurtosis analysis with respect to the audio data 520.

[0090] In one or more illustrative examples, the audio data preprocessing 522 can include performing a short time Fourier transform with respect to the audio data 520. Additionally, the audio data preprocessing 522 can include generating a power spectrogram based on the audio data 520. In at least some examples, one or more operation performed at 518 can include determining time intervals extracted from the short time Fourier transform. For example, the time intervals can be determined according to a discretization process performed in relation to a short time Fourier transform of the audio data 520. The time intervals can be from about 0.05 seconds to about 1 second, from about 0.1 seconds to about 0.5 seconds, or from about 0.05 seconds to about 0.25 seconds. In addition, the one or more operations performed at 518 can include analyzing the magnitude of the signal indicated by the power spectrogram during the time intervals. Further, the analysis performed by the one or more operations at 518 can include analyzing the magnitude of the signal indicated by the power spectrogram at the one or more fault frequencies 516 at individual time intervals.

[0091] For individual time intervals, the magnitude of the signal at the one or more fault frequencies 516 can be analyzed in relation to one or more threshold values 526. The one or more threshold values 526 can be determined by analyzing audio data produced by equipment in which no defects, faults, and / or other anomalies are present with respect to one or more components of equipment. To illustrate, a short time Fourier transform can be applied to audio data obtained from equipment operating properly to produce a power spectrogram for the equipment. One or more peak amplitudes for the properly operating equipment can be determined from the spectrogram derived from the audio data for the equipment. In various examples, the one or more peak amplitudes can correspond to the one or more threshold values 526. In one or more additional examples, the one or more threshold values 526 can be determined by determining one or more statistical measures of the amplitudes included in a power spectrogram derived from the audio data 520. In one or more illustrative examples, the one or more statistical measures can include at least one of a mean, median, or mode of the amplitudes included in the power spectrogram derived from the audio data 520.

[0092] The analysis of the audio data 520 and / or the modified audio data 524 at 518 in relation to the one or more threshold values 526 can generate one or more quantitative measures 528. The one or more quantitative measures 528 can be determined by analyzing one or more magnitudes included in a power spectrogram derived from the audio data 520 at a number of time intervals and at the one or more fault frequencies 516 in relation to the one or more threshold values 526. In one or more illustrative examples, a quantitative measure 528 can be determined for each time interval based on whether or not the amplitude at the one or more fault frequencies 516 is at least a threshold value 526. In one or more examples, the quantitative measure 528 can have a first value or a first set of values in situations where the amplitude is at least a threshold value 526 for a given time interval. The quantitative measure 528 can have a second value or a second set of values different from the first value or first set of values in situations where the amplitude is less than the threshold value 526 for a given time interval.

[0093] At 530, the matched characteristic frequency analysis 504 can include determining a probability of a state for the piece of equipment. In one or more examples, the matched characteristic frequency analysis 504 can include determining a probability of at least one of a failure, defect, or other anomaly being present in equipment. In various examples, the probability of a piece of equipment operating improperly can be determined by analyzing the quantitative measures 528. In one or more illustrative examples, the quantitative measures 528 can include a score or other metric for each time interval determined from a short time Fouriertransform analysis of the audio data 520. In at least some examples, the quantitative measures for each time interval can be combined into an aggregate measure that corresponds to a probability of a fault being present with respect to a piece of equipment. In one or more additional examples, the probability of a fault being present with respect to a piece of equipment can be used to provide an indication, such as via one or more user interfaces, that a piece of equipment is not operating properly. In one or more further examples, an analysis of harmonics of the audio data 520 can be performed to determine a probability of a fault, defect, or other anomaly being present with respect to one or more components of a piece of equipment.

[0094] Figure 6 is a flow diagram of a process 600 to analyze audio data produced by one or more pieces of equipment to determine a state of the one or more pieces of equipment, according to one or more example implementations. The process may be embodied in computer-readable instructions for execution by one or more processors such that the operations of the processes may be performed in part or in whole by the functional components of at least one of one or more user devices or one or more server systems. Accordingly, the processes described below are by way of example with reference thereto, in some situations. However, in other implementations, at least some of the operations of the processes described with respect to Figure 5 may be deployed on various other hardware configurations. The processes described with respect to Figure 5 are therefore not intended to be limited to the being performed by one or more server systems or one or more user device described herein and can be implemented in whole, or in part, by one or more additional components. Although the described flowcharts can show operations as a sequential process, many of the operations can be performed in parallel or concurrently. In addition, the order of the operations may be re-arranged. A process is terminated when its operations are completed. A process may correspond to a method, a procedure, an algorithm, etc. The operations of methods may be performed in whole or in part, may be performed in conjunction with some or all of the operations in other methods, and may be performed by any number of different systems, such as the systems described herein, or any portion thereof, such as a processor included in any of the systems.

[0095] At 602, the process 600 can include obtaining audio data produced during operation of a piece of equipment. The audio data can be captured by one or more computing devices and stored in an audio data file.

[0096] The process 600 can include, at 604, producing first modified audio data by performing a time domain analysis of the audio data. In one or more examples, the time domain analysis can include an envelope analysis being performed with respect to the audio data.

[0097] In addition, at 606, the process 600 can include producing second modified audio data by performing a frequency domain analysis of the audio data. In various examples, the frequency domain analysis can include applying a fast Fourier transform algorithm to the audio data.

[0098] The process 600 can also include, at 608, producing third modified audio data by performing a time-frequency domain analysis of the audio data. In one or more illustrative examples, the time-frequency domain analysis can include applying a sparse fast Fourier transform algorithm to the audio data.

[0099] Further, at 610, the process 600 can include providing the first modified audio data, the second modified audio data, and the third modified audio data as input data for a generative artificial intelligence (Al) model. In at least some examples, the generative Al model can include a transformer machine learning model that includes an encoder and a decoder. In one or more examples, a training process can be performed with respect to the large language training model. In various examples, the generative Al model can include a contrastive language-image pre-training (CLIP) model that couples graphics, such as image content, with natural language, such as text content. In these scenarios, input to the generative Al model can include audio data that has been modified to correspond to image data. For example, at least one of the first modified audio data, the second modified audio data, or the third modified audio data can be normalized according to a scale, such as 0 to 1 or 0 to 100, and provided as input to the generative Al model. In one or more additional examples, at least one of the first modified audio data, the second modified audio data, or the third modified audio data can be converted to images that are provided as input to the generative Al model. In one or more further examples, in scenarios where the generative Al model implements CLIP -based machine learning techniques, fine-tuning can be applied to modify a pretrained artificial neural network for the analysis of audio data. In still other illustrative examples where the generative Al model implements CLIP -based machine learning techniques, few shot learning can be applied to modify a pretrained artificial neural network for the analysis of audio data.

[0100] At 612, the process 600 can include computationally analyzing, by the generative Al model, the input data to determine a quantitative measure relating to a fault being present with respect to operation of the piece of equipment. In one or more examples, the quantitativemeasure can be compared to a threshold value. In situations where the quantitative measure has a value of at least the threshold value, a fault can be determined to be present with respect to the piece of equipment. In still other examples, an additional computational analysis can be performed by a probabilistic model to determine one or more classifications with respect to the fault of the piece of equipment. In one or more illustrative examples, the probabilistic model can analyze the first modified audio data, the second modified audio data, and the third modified audio data to determine the one or more classifications of the fault of the piece of equipment.

[0101] In one or more additional examples, at least one of the generative Al model or the probabilistic model can analyze additional data. For example, the generative Al model can analyze equipment data with respect to the piece of equipment to determine the quantitative measure. In one or more illustrative examples, the equipment data can include at least one of one or more physical characteristics of the piece of equipment or one or more operational specifications of the piece of equipment. Additionally, the probabilistic model can also analyze the equipment data to determine the one or more classifications with respect to the fault of the piece of equipment.

[0102] In one or more further examples, additional input data can be determined in relation to the generative Al model and the probabilistic model. To illustrate, one or more frequency ranges can be determined in relation to the audio data that include signals that can be indicative of a fault being present with respect to the piece of equipment. In one or more illustrative examples, a spectral kurtosis analysis can be performed to identify the one or more frequency ranges. The audio data can be filtered in relation to the one or more frequency ranges to produce filtered data. In these scenarios, the additional input data for the generative Al model and the probabilistic model can be generated by applying the time domain analysis, the frequency domain analysis, and the time-frequency domain analysis to the filtered data.

[0103] Additionally, the process 600, at 614, can include causing display of a user interface that includes a user interface element corresponding to a presence of a fault with respect to the piece of equipment or an absence of a fault with respect to the piece of equipment. In various examples, the user interface can be displayed as part of a dashboard that is accessible to users of the system.

[0104] Although a flowchart or block diagram may illustrate a method as comprising sequential steps or a process as having a particular order of operations, many of the steps or operations in the flowchart(s) or block diagram(s) illustrated herein can be performed inparallel or concurrently, and the flowchart(s) or block diagram(s) should be read in the context of the various implementations of the present disclosure. In addition, the order of the method steps or process operations illustrated in a flowchart or block diagram may be rearranged for some implementations. Similarly, a method or process illustrated in a flow chart or block diagram could have additional steps or operations not included therein or fewer steps or operations than those shown. Moreover, a method step may correspond to a method, a function, a procedure, a subroutine, a subprogram, etc.

[0105] Figure 7 is a block diagram illustrating components of a machine 700, according to some example implementations, able to read instructions from a machine-readable medium (e.g., a machine-readable storage medium) and perform any one or more of the methodologies discussed herein. Specifically, Figure 7 shows a diagrammatic representation of the machine 700 in the example form of a computer system, within which instructions 702 (e.g., software, a program, an application, an applet, an app, or other executable code) for causing the machine 700 to perform any one or more of the methodologies discussed herein may be executed. As such, the instructions 702 may be used to implement modules or components described herein. The instructions 702 transform the general, non-programmed machine 700 into a particular machine 700 programmed to carry out the described and illustrated functions in the manner described. In alternative implementations, the machine 700 operates as a standalone device or may be coupled (e.g., networked) to other machines. In a networked deployment, the machine 700 may operate in the capacity of a server machine or a client machine in a server-client network environment, or as a peer machine in a peer-to-peer (or distributed) network environment. The machine 700 may comprise, but not be limited to, a server computer, a client computer, a personal computer (PC), a tablet computer, a laptop computer, a netbook, a set-top box (STB), a personal digital assistant (PDA), an entertainment media system, a cellular telephone, a smart phone, a mobile device, a wearable device (e.g., a smart watch), a smart home device (e.g., a smart appliance), other smart devices, a web appliance, a network router, a network switch, a network bridge, or any machine capable of executing the instructions 702, sequentially or otherwise, that specify actions to be taken by machine 700. Further, while only a single machine 700 is illustrated, the term "machine" shall also be taken to include a collection of machines that individually or jointly execute the instructions 702 to perform any one or more of the methodologies discussed herein.

[0106] The machine 700 may include processors 704, memory / storage 706, and VO components 708, which may be configured to communicate with each other such as via a bus710. In an example implementation, the processors 704 (e.g., a central processing unit (CPU), a reduced instruction set computing (RISC) processor, a complex instruction set computing (CISC) processor, a graphics processing unit (GPU), a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a radio-frequency integrated circuit (RFIC), another processor, or any suitable combination thereof) may include, for example, a processor 712 and a processor 714 that may execute the instructions 702. The term “processor” is intended to include multi-core processors 704 that may comprise two or more independent processors (sometimes referred to as “cores”) that may execute instructions 702 contemporaneously. Although Figure 6 shows multiple processors 704, the machine 700 may include a single processor 712 with a single core, a single processor 712 with multiple cores (e.g., a multi-core processor), multiple processors 712, 714 with a single core, multiple processors 712, 714 with multiple cores, or any combination thereof.

[0107] The memory / storage 706 may include memory, such as a main memory 716, or other memory storage, and a storage unit 718, both accessible to the processors 704 such as via the bus 710. The storage unit 718 and main memory 716 store the instructions 702 embodying any one or more of the methodologies or functions described herein. The instructions 702 may also reside, completely or partially, within the main memory 716, within the storage unit 718, within at least one of the processors 704 (e.g., within the processor’s cache memory), or any suitable combination thereof, during execution thereof by the machine 700. Accordingly, the main memory 716, the storage unit 718, and the memory of processors 704 are examples of machine- readable media.

[0108] The I / O components 708 may include a wide variety of components to receive input, provide output, produce output, transmit information, exchange information, capture measurements, and so on. The specific I / O components 708 that are included in a particular machine 700 will depend on the type of machine. For example, portable machines such as mobile phones will likely include a touch input device or other such input mechanisms, while a headless server machine will likely not include such a touch input device. It will be appreciated that the I / O components 708 may include many other components that are not shown in Figure 7. The VO components 708 are grouped according to functionality merely for simplifying the following discussion and the grouping is in no way limiting. In various example implementations, the I / O components 708 may include user output components 720 and user input components 722. The user output components 720 may include visual components (e.g., a display such as a plasma display panel (PDP), a light emitting diode (LED)display, a liquid crystal display (LCD), a projector, or a cathode ray tube (CRT)), acoustic components (e.g., speakers), haptic components (e.g., a vibratory motor, resistance mechanisms), other signal generators, and so forth. The user input components 722 may include alphanumeric input components (e.g., a keyboard, a touch screen configured to receive alphanumeric input, a photo-optical keyboard, or other alphanumeric input components), pointbased input components (e.g., a mouse, a touchpad, a trackball, a joystick, a motion sensor, or other pointing instrument), tactile input components (e.g., a physical button, a touch screen that provides location or force of touches or touch gestures, or other tactile input components), audio input components (e.g., a microphone), and the like.

[0109] In further example implementations, the I / O components 708 may include biometric components 724, motion components 726, environmental components 728, or position components 730 among a wide array of other components. For example, the biometric components 724 may include components to detect expressions (e.g., hand expressions, facial expressions, vocal expressions, body gestures, or eye tracking), measure biosignals (e.g., blood pressure, heart rate, body temperature, perspiration, or brain waves), identify a person (e.g., voice identification, retinal identification, facial identification, fingerprint identification, or electroencephalogram based identification), and the like. The motion components 726 may include acceleration sensor components (e.g., accelerometer), gravitation sensor components, rotation sensor components (e.g., gyroscope), and so forth. The environmental components 728 may include, for example, illumination sensor components (e.g., photometer), temperature sensor components (e.g., one or more thermometer that detect ambient temperature), humidity sensor components, pressure sensor components (e.g., barometer), acoustic sensor components (e.g., one or more microphones that detect background noise), proximity sensor components (e.g., infrared sensors that detect nearby objects), gas sensors (e.g., gas detection sensors to detection concentrations of hazardous gases for safety or to measure pollutants in the atmosphere), or other components that may provide indications, measurements, or signals corresponding to a surrounding physical environment. The position components 730 may include location sensor components (e.g., a GPS receiver component), altitude sensor components (e.g., altimeters or barometers that detect air pressure from which altitude may be derived), orientation sensor components (e.g., magnetometers), and the like.

[0110] Communication may be implemented using a wide variety of technologies. The I / O components 708 may include communication components 732 operable to couple the machine 700 to a network 734 or devices 736. For example, the communication components 732 mayinclude a network interface component or other suitable device to interface with the network 734. In further examples, communication components 732 may include wired communication components, wireless communication components, cellular communication components, near field communication (NFC) components, Bluetooth® components (e.g., Bluetooth® Low Energy), Wi-Fi® components, and other communication components to provide communication via other modalities. The devices 736 may be another machine 700 or any of a wide variety of peripheral devices (e.g., a peripheral device coupled via a USB).

[0111] Moreover, the communication components 732 may detect identifiers or include components operable to detect identifiers. For example, the communication components 732 may include radio frequency identification (RFID) tag reader components, NFC smart tag detection components, optical reader components (e.g., an optical sensor to detect onedimensional bar codes such as Universal Product Code (UPC) bar code, multi-dimensional bar codes such as Quick Response (QR) code, Aztec code, Data Matrix, Dataglyph, MaxiCode, PDF417, Ultra Code, UCC RSS-2D bar code, and other optical codes), or acoustic detection components (e.g., microphones to identify tagged audio signals). In addition, a variety of information may be derived via the communication components 732, such as location via Internet Protocol (IP) geo-location, location via Wi-Fi® signal triangulation, location via detecting an NFC beacon signal that may indicate a particular location, and so forth.

[0112] As used herein, “component” refers to a device, physical entity, or logic having boundaries defined by function or subroutine calls, branch points, APIs, or other technologies that provide for the partitioning or modularization of particular processing or control functions. Components may be combined via their interfaces with other components to carry out a machine process. A component may be a packaged functional hardware unit designed for use with other components and a part of a program that usually performs a particular function of related functions. Components may constitute either software components (e.g., code embodied on a machine-readable medium) or hardware components. A "hardware component" is a tangible unit capable of performing certain operations and may be configured or arranged in a certain physical manner. In various example implementations, one or more computer systems (e.g., a standalone computer system, a client computer system, or a server computer system) or one or more hardware components of a computer system (e.g., a processor or a group of processors) may be configured by software (e.g., an application or application portion) as a hardware component that operates to perform certain operations as described herein.

[0113] A hardware component may also be implemented mechanically, electronically, or any suitable combination thereof. For example, a hardware component may include dedicated circuitry or logic that is permanently configured to perform certain operations. A hardware component may be a special-purpose processor, such as a field-programmable gate array (FPGA) or an ASIC. A hardware component may also include programmable logic or circuitry that is temporarily configured by software to perform certain operations. For example, a hardware component may include software executed by a general-purpose processor 704 or other programmable processor. Once configured by such software, hardware components become specific machines (or specific components of a machine 700) uniquely tailored to perform the configured functions and are no longer general-purpose processors 704. It will be appreciated that the decision to implement a hardware component mechanically, in dedicated and permanently configured circuitry, or in temporarily configured circuitry (e.g., configured by software) may be driven by cost and time considerations. Accordingly, the phrase "hardware component"(or "hardware-implemented component") should be understood to encompass a tangible entity, be that an entity that is physically constructed, permanently configured (e.g., hardwired), or temporarily configured (e.g., programmed) to operate in a certain manner or to perform certain operations described herein. Considering implementations in which hardware components are temporarily configured (e.g., programmed), each of the hardware components need not be configured or instantiated at any one instance in time. For example, where a hardware component comprises a general-purpose processor 704 configured by software to become a special-purpose processor, the general-purpose processor 704 may be configured as respectively different special-purpose processors (e.g., comprising different hardware components) at different times. Software accordingly configures a particular processor 712, 714 or processors 704, for example, to constitute a particular hardware component at one instance of time and to constitute a different hardware component at a different instance of time.

[0114] Hardware components can provide information to, and receive information from, other hardware components. Accordingly, the described hardware components may be regarded as being communicatively coupled. Where multiple hardware components exist contemporaneously, communications may be achieved through signal transmission (e.g., over appropriate circuits and buses) between or among two or more of the hardware components. In implementations in which multiple hardware components are configured or instantiated at different times, communications between such hardware components may be achieved, forexample, through the storage and retrieval of information in memory structures to which the multiple hardware components have access. For example, one hardware component may perform an operation and store the output of that operation in a memory device to which it is communicatively coupled. A further hardware component may then, at a later time, access the memory device to retrieve and process the stored output.

[0115] Hardware components may also initiate communications with input or output devices, and can operate on a resource (e.g., a collection of information). The various operations of example methods described herein may be performed, at least partially, by one or more processors 704 that are temporarily configured (e.g., by software) or permanently configured to perform the relevant operations. Whether temporarily or permanently configured, such processors 704 may constitute processor-implemented components that operate to perform one or more operations or functions described herein. As used herein, "processor-implemented component" refers to a hardware component implemented using one or more processors 704. Similarly, the methods described herein may be at least partially processor-implemented, with a particular processor 712, 714 or processors 704 being an example of hardware. For example, at least some of the operations of a method may be performed by one or more processors 704 or processor-implemented components. Moreover, the one or more processors 704 may also operate to support performance of the relevant operations in a "cloud computing" environment or as a "software as a service" (SaaS). For example, at least some of the operations may be performed by a group of computers (as examples of machines 700 including processors 704), with these operations being accessible via a network 734 (e.g., the Internet) and via one or more appropriate interfaces (e.g., an API). The performance of certain of the operations may be distributed among the processors, not only residing within a single machine 700, but deployed across a number of machines. In some example implementations, the processors 704 or processor-implemented components may be located in a single geographic location (e.g., within a home environment, an office environment, or a server farm). In other example implementations, the processors 704 or processor-implemented components may be distributed across a number of geographic locations.

[0116] Figure 8 is a block diagram illustrating system 800 that includes an example software architecture 802, which may be used in conjunction with various hardware architectures herein described. Figure 8 is a non-limiting example of a software architecture, and it will be appreciated that many other architectures may be implemented to facilitate the functionality described herein. The software architecture 802 may execute on hardware such as machine 700of Figure 7 that includes, among other things, processors 704, memory / storage 706, and input / output (I / O) components 708. A representative hardware layer 804 is illustrated and can represent, for example, the machine 700 of Figure 7. The representative hardware layer 804 includes a processing unit 806 having associated executable instructions 808. Executable instructions 808 represent the executable instructions of the software architecture 802, including implementation of the methods, components, and so forth described herein. The hardware layer 804 also includes at least one of memory or storage modules memory / storage 810, which also have executable instructions 808. The hardware layer 804 may also comprise other hardware 812.

[0117] In the example architecture of Figure 8, the software architecture 802 may be conceptualized as a stack of layers where each layer provides particular functionality. For example, the software architecture 802 may include layers such as an operating system 814, libraries 816, frameworks / middl eware 818, applications 820, and a presentation layer 822. Operationally, the applications 820 or other components within the layers may invoke API calls 824 through the software stack and receive messages 826 in response to the API calls 824. The layers illustrated are representative in nature and not all software architectures have all layers. For example, some mobile or special purpose operating systems may not provide a frameworks / middleware 818, while others may provide such a layer. Other software architectures may include additional or different layers.

[0118] The operating system 814 may manage hardware resources and provide common services. The operating system 814 may include, for example, a kernel 828, services 830, and drivers 832. The kernel 828 may act as an abstraction layer between the hardware and the other software layers. For example, the kernel 828 may be responsible for memory management, processor management (e.g., scheduling), component management, networking, security settings, and so on. The services 830 may provide other common services for the other software layers. The drivers 832 are responsible for controlling or interfacing with the underlying hardware. For instance, the drivers 832 include display drivers, camera drivers, Bluetooth® drivers, flash memory drivers, serial communication drivers (e.g., Universal Serial Bus (USB) drivers), Wi-Fi® drivers, audio drivers, power management drivers, and so forth depending on the hardware configuration.

[0119] The libraries 816 provide a common infrastructure that is used by at least one of the applications 820, other components, or layers. The libraries 816 provide functionality that allows other software components to perform tasks in an easier fashion than to interfacedirectly with the underlying operating system 814 functionality (e.g., kernel 828, services 830, drivers 832). The libraries 816 may include system libraries 834 (e.g., C standard library) that may provide functions such as memory allocation functions, string manipulation functions, mathematical functions, and the like. In addition, the libraries 816 may include API libraries 836 such as media libraries (e.g., libraries to support presentation and manipulation of various media format such as MPEG4, H.264, MP3, AAC, AMR, JPG, PNG), graphics libraries (e.g., an OpenGL framework that may be used to render two-dimensional and three-dimensional in a graphic content on a display), database libraries (e.g., SQLite that may provide various relational database functions), web libraries (e.g., WebKit that may provide web browsing functionality), and the like. The libraries 816 may also include a wide variety of other libraries 838 to provide many other APIs to the applications 820 and other software components / modules.

[0120] The frameworks / middl eware 818 (also sometimes referred to as middleware) provide a higher-level common infrastructure that may be used by the applications 820 or other software components / modules. For example, the frameworks / middleware 818 may provide various graphical user interface functions, high-level resource management, high-level location services, and so forth. The frameworks / middleware 818 may provide a broad spectrum of other APIs that may be utilized by the applications 820 or other software components / modules, some of which may be specific to a particular operating system 814 or platform.

[0121] The applications 820 include built-in applications 840 and third-party applications 842. Examples of representative built-in applications 840 may include, but are not limited to, a contacts application, a browser application, a book reader application, a location application, a media application, a messaging application, or a game application. Third-party applications 842 may include an application developed using the ANDROID™ or IOS™ software development kit (SDK) by an entity other than the vendor of the particular platform and may be mobile software running on a mobile operating system such as IOS™, ANDROID™, WINDOWS® Phone, or other mobile operating systems. The third-party applications 842 may invoke the API calls 824 provided by the mobile operating system (such as operating system 814) to facilitate functionality described herein.

[0122] The applications 820 may use built-in operating system functions (e.g., kernel 828, services 830, drivers 832), libraries 816, and frameworks / middleware 818 to create UIs to interact with users of the system. Alternatively, or additionally, in some systems, interactions with a user may occur through a presentation layer, such as presentation layer 822. In thesesystems, the application / component "logic" can be separated from the aspects of the application / component that interact with a user.

[0123] At least some of the processes described herein can be embodied in computer-readable instructions for execution by one or more processors such that the operations of the processes may be performed in part or in whole by the functional components of one or more computer systems. Accordingly, computer-implemented processes described herein are by way of example with reference thereto, in some situations. However, in other implementations, at least some of the operations of the computer-implemented processes described herein can be deployed on various other hardware configurations. The computer-implemented processes described herein are therefore not intended to be limited to the systems and configurations described with respect to Figures 1-6 and can be implemented in whole, or in part, by one or more additional system and / or components.

[0124] As used herein, the terms “substantially” or “generally” refer to the complete or nearly complete extent or degree of an action, characteristic, property, state, structure, item, or result. For example, an object that is “substantially” or “generally” enclosed would mean that the object is either completely enclosed or nearly completely enclosed. The exact allowable degree of deviation from absolute completeness may in some cases depend on the specific context. However, generally speaking, the nearness of completion will be so as to have generally the same overall result as if absolute and total completion were obtained. The use of “substantially” or “generally” is equally applicable when used in a negative connotation to refer to the complete or near complete lack of an action, characteristic, property, state, structure, item, or result. For example, an element, combination, implementation, or composition that is “substantially free of’ or “generally free of’ an element may still actually contain such element as long as there is generally no significant effect thereof.

[0125] In the foregoing description various implementations of the present disclosure have been presented for the purpose of illustration and description. They are not intended to be exhaustive or to limit the invention to the precise form disclosed. Obvious modifications or variations are possible in light of the above teachings. The various implementations were chosen and described to provide the best illustration of the principals of the disclosure and their practical application, and to enable one of ordinary skill in the art to utilize the various implementations with various modifications as are suited to the particular use contemplated. All such modifications and variations are within the scope of the present disclosure asdetermined by the appended claims when interpreted in accordance with the breadth they are fairly, legally, and equitably entitled.

[0126] In view of the above-described implementations of subject matter this application discloses the following list of examples, wherein one feature of an example in isolation or more than one feature of an example, taken in combination and, optionally, in combination with one or more features of one or more further examples are further examples also falling within the disclosure of this application.

[0127] Example 1. A method comprising: obtaining audio data produced during operation of a piece of equipment; producing first modified audio data by performing a time domain analysis of the audio data; producing second modified audio data by performing a frequency domain analysis of the audio data; producing third modified audio data by performing a time-frequency domain analysis of the audio data; providing the first modified audio data, the second modified audio data, and the third modified audio data as input data for a generative artificial intelligence (Al) model; computationally analyzing, by the generative Al model, the input data to determine a quantitative measure relating to a fault being present with respect to operation of the piece of equipment; and causing display of a user interface that includes a user interface element corresponding to a presence of a fault with respect to the piece of equipment or an absence of a fault with respect to the piece of equipment.

[0128] Example 2. The method of example 1, comprising: analyzing the quantitative measure with respect to a threshold value; and determining that a fault is present with respect to the piece of equipment based on the quantitative measure being at least the threshold value.

[0129] Example 3. The method of example 1 or 2, wherein the generative Al model includes a transformer machine learning model having an encoder and a decoder.

[0130] Example 4. The method of any one of examples 1-3, comprising: obtaining equipment data indicating at least one of physical characteristics of the piece of equipment or operational characteristics of the piece of equipment; providing the equipment data as additional input data to the generative Al model; and wherein the generative Al model determines the quantitative measure based on the input data and the additional input data.

[0131] Example 5. The method of example 4, wherein: the piece of equipment is a pump; and the physical characteristics of the piece of equipment include at least one of bearing geometry or gear dimensions.

[0132] Example 6. The method of any one of examples 1-5, comprising: performing a training process of the generative Al model, wherein the training process includes: obtaining firsttraining data that includes audio data obtained from a plurality of first additional pieces of equipment that correspond to the piece of equipment and in which a fault is not present; obtaining second training data that includes audio data obtained from a plurality of second additional pieces of equipment that correspond to the piece of equipment and in which a fault is present; and performing a number of iterations of the training process using the first training data and the second training data to minimize a loss function of the generative Al model.

[0133] Example 7. The method of example 6, wherein: the piece of equipment is a pump having a first bearing geometry and a first operational speed; and the plurality of first additional pieces of equipment and the plurality of second additional pieces of equipment have the first bearing geometry and the first operational speed.

[0134] Example 8. The method of example 6, wherein: the piece of equipment is a pump having a first bearing geometry and a first operational speed; and a first portion of the plurality of first additional pieces of equipment and a first portion of the plurality of second additional pieces of equipment have the first bearing geometry and the first operational speed; a second portion of the plurality of first additional pieces of equipment and a second portion of the plurality of second additional pieces of equipment have a second bearing geometry and a second operational speed.

[0135] Example 9. The method of any one of examples 1-8, comprising: analyzing the first modified audio data, the second modified audio data, and the third modified audio data by implementing a probabilistic model to determine a plurality of classifications for the state of the piece of equipment.

[0136] Example 10. The method of example 9, wherein: the plurality of classifications include a first classification of the plurality of classifications corresponds to a first fault of the piece of equipment and a second classification of the plurality of classifications corresponds to a second fault of the piece of equipment; and the probabilistic model determines a first probability of the first fault being present with respect to the piece of equipment and a second probability of the second fault being present with respect to the piece of equipment, f

[0137] Example 11. The method of example 9, wherein the probabilistic model includes at least one of a Bayesian neural network, one or more Monte Carlo prediction algorithms, one or more Kullback Leibler (KL) divergence methods, or one or more likelihood prediction models.

[0138] Example 12. The method of example 9, wherein: the probabilistic model analyzes equipment data for the piece of equipment and the first modified audio data, the secondmodified audio data, and the third modified audio data to determine the plurality of classifications; and the equipment data includes at least one of physical characteristics of one or more parts of the piece of equipment or one or more operational specifications of the piece of equipment.

[0139] Example 13. A system comprising: one or more hardware processors; and one or more computer-readable storage devices storing computer-readable instructions that, when executed by the one or more hardware processors, cause the one or more hardware processors to perform operations comprising: obtaining audio data produced during operation of a piece of equipment; producing first modified audio data by performing a time domain analysis of the audio data; producing second modified audio data by performing a frequency domain analysis of the audio data; producing third modified audio data by performing a time-frequency domain analysis of the audio data; providing the first modified audio data, the second modified audio data, and the third modified audio data as input data for a generative Al model; computationally analyzing, by the generative Al model, the input data to determine a quantitative measure relating to a fault being present with respect to operation of the equipment; and causing display of a user interface that includes a user interface element corresponding to a presence of a fault with respect to the piece of equipment or an absence of a fault with respect to the piece of equipment.

[0140] Example 14. The system of example 13, wherein the one or more computer-readable storage devices store additional computer-readable instructions that, when executed by the one or more hardware processors, cause the one or more hardware processors to perform additional operations comprising: performing a spectral kurtosis analysis of the audio data to determine a range of frequencies that include information indicative of a fault being present with respect to the piece of equipment; performing an additional time domain analysis of a portion of the audio data corresponding to the range of frequencies to determine first additional modified audio data; performing an additional frequency domain analysis of the portion of the audio data corresponding to the range of frequencies to determine second additional modified audio data; and performing an additional time-frequency domain analysis of the portion of the audio data corresponding to the range of frequencies to determine third additional modified audio data.

[0141] Example 15. The system of example 14, wherein one or more filters are applied to produce the portion of the audio data corresponding to the range of frequencies.

[0142] Example 16. The system of example 14 or 15, wherein: the first additional modified audio data, the second additional modified audio data, and the third additional modified audiodata comprise additional input to the generative Al model; and the generative Al model computationally analyzes the input data and the additional input data to determine the quantitative measure.

[0143] Example 17. The system of any one of examples 14-16, wherein the one or more computer-readable storage devices store additional computer-readable instructions that, when executed by the one or more hardware processors, cause the one or more hardware processors to perform additional operations comprising: analyzing the first modified audio data, the second modified audio data, the third modified audio data, the first additional modified audio data, the second additional modified audio data, and the third additional modified audio data by implementing a probabilistic model to determine a plurality of classifications for the state of the piece of equipment.

[0144] Example 18. The system of any one of examples 13-17, wherein: the time domain analysis of the audio data includes an amplitude envelope analysis of the audio data; the frequency domain analysis of the audio data includes applying a fast Fourier transform technique to the audio data; and the time-frequency domain analysis of the audio data includes applying a sparse fast Fourier transform technique to the audio data.

[0145] Example 19. The system of any one of examples 13-18, wherein the one or more computer-readable storage devices store additional computer-readable instructions that, when executed by the one or more hardware processors, cause the one or more hardware processors to perform additional operations comprising: determining, based on physical characteristics of the piece of equipment and operational characteristics of the piece of equipment, one or more expected peak frequencies for the piece of equipment; analyzing the first modified audio data, the second modified audio data, and the third modified audio data to determine one or more actual peak frequencies for the piece of equipment; and analyzing the one or more actual peak frequencies in relation to the one or more expected peak frequencies to determine the state of the piece of equipment.

[0146] Example 20. The system of any one of examples 13-19, wherein the one or more computer-readable storage devices store additional computer-readable instructions that, when executed by the one or more hardware processors, cause the one or more hardware processors to perform additional operations comprising: causing display of an additional user interface, the additional user interface including one or more user interface elements that are selectable to indicate one or more features of the piece of equipment; and responsive to input captured by the one or more user interface elements, retrieving at least one of one or more physicalcharacteristics of the piece of equipment or one or more operational specifications of the piece of equipment from a data store.

Claims

CLAIMSWhat is claimed is:

1. A method comprising: obtaining audio data produced during operation of a piece of equipment; producing first modified audio data by performing a time domain analysis of the audio data; producing second modified audio data by performing a frequency domain analysis of the audio data; producing third modified audio data by performing a time-frequency domain analysis of the audio data; providing the first modified audio data, the second modified audio data, and the third modified audio data as input data for a generative artificial intelligence (Al) model; computationally analyzing, by the generative Al model, the input data to determine a quantitative measure relating to a fault being present with respect to operation of the piece of equipment; and causing display of a user interface that includes a user interface element corresponding to a presence of a fault with respect to the piece of equipment or an absence of a fault with respect to the piece of equipment.

2. The method of claim 1, comprising: analyzing the quantitative measure with respect to a threshold value; and determining that a fault is present with respect to the piece of equipment based on the quantitative measure being at least the threshold value.

3. The method of claim 1, wherein the generative Al model includes a transformer machine learning model having an encoder and a decoder.

4. The method of claim 1, comprising: obtaining equipment data indicating at least one of physical characteristics of the piece of equipment or operational characteristics of the piece of equipment; and providing the equipment data as additional input data to the generative Al model;wherein the generative Al model determines the quantitative measure based on the input data and the additional input data.

5. The method of claim 4, wherein: the piece of equipment is a pump; and the physical characteristics of the piece of equipment include at least one of bearing geometry or gear dimensions.

6. The method of claim 1, comprising: performing a training process of the generative Al model, wherein the training process includes: obtaining first training data that includes audio data obtained from a plurality of first additional pieces of equipment that correspond to the piece of equipment and in which a fault is not present; obtaining second training data that includes audio data obtained from a plurality of second additional pieces of equipment that correspond to the piece of equipment and in which a fault is present; and performing a number of iterations of the training process using the first training data and the second training data to minimize a loss function of the generative Al model.

7. The method of claim 6, wherein: the piece of equipment is a pump having a first bearing geometry and a first operational speed; and the plurality of first additional pieces of equipment and the plurality of second additional pieces of equipment have the first bearing geometry and the first operational speed.

8. The method of claim 6, wherein: the piece of equipment is a pump having a first bearing geometry and a first operational speed; a first portion of the plurality of first additional pieces of equipment and a first portion of the plurality of second additional pieces of equipment have the first bearing geometry and the first operational speed; anda second portion of the plurality of first additional pieces of equipment and a second portion of the plurality of second additional pieces of equipment have a second bearing geometry and a second operational speed.

9. The method of claim 1, comprising: analyzing the first modified audio data, the second modified audio data, and the third modified audio data by implementing a probabilistic model to determine a plurality of classifications for the piece of equipment.

10. The method of claim 9, wherein: the plurality of classifications include a first classification of the plurality of classifications corresponds to a first fault of the piece of equipment and a second classification of the plurality of classifications corresponds to a second fault of the piece of equipment; and the probabilistic model determines a first probability of the first fault being present with respect to the piece of equipment and a second probability of the second fault being present with respect to the piece of equipment.

11. The method of claim 9, wherein the probabilistic model includes at least one of a Bayesian neural network, one or more Monte Carlo prediction algorithms, one or more Kullback Leibler (KL) divergence methods, or one or more likelihood prediction models.

12. The method of claim 9, wherein: the probabilistic model analyzes equipment data for the piece of equipment and the first modified audio data, the second modified audio data, and the third modified audio data to determine the plurality of classifications; and the equipment data includes at least one of physical characteristics of one or more parts of the piece of equipment or one or more operational specifications of the piece of equipment.

13. A system comprising: one or more hardware processors; andone or more computer-readable storage devices storing computer-readable instructions that, when executed by the one or more hardware processors, cause the one or more hardware processors to perform operations comprising: obtaining audio data produced during operation of a piece of equipment; producing first modified audio data by performing a time domain analysis of the audio data; producing second modified audio data by performing a frequency domain analysis of the audio data; producing third modified audio data by performing a time-frequency domain analysis of the audio data; providing the first modified audio data, the second modified audio data, and the third modified audio data as input data for a generative Al model; computationally analyzing, by the generative Al model, the input data to determine a quantitative measure relating to a fault being present with respect to operation of the equipment; and causing display of a user interface that includes a user interface element corresponding to a presence of a fault with respect to the piece of equipment or an absence of a fault with respect to the piece of equipment.

14. The system of claim 13, wherein the one or more computer-readable storage devices store additional computer-readable instructions that, when executed by the one or more hardware processors, cause the one or more hardware processors to perform additional operations comprising: performing a spectral kurtosis analysis of the audio data to determine a range of frequencies that include information indicative of a fault being present with respect to the piece of equipment; performing an additional time domain analysis of a portion of the audio data corresponding to the range of frequencies to determine first additional modified audio data; performing an additional frequency domain analysis of the portion of the audio data corresponding to the range of frequencies to determine second additional modified audio data; andperforming an additional time-frequency domain analysis of the portion of the audio data corresponding to the range of frequencies to determine third additional modified audio data.

15. The system of claim 14, wherein one or more filters are applied to produce the portion of the audio data corresponding to the range of frequencies.

16. The system of claim 14, wherein: the first additional modified audio data, the second additional modified audio data, and the third additional modified audio data comprise additional input data to the generative Al model; and the generative Al model computationally analyzes the input data and the additional input data to determine the quantitative measure.

17. The system of claim 14, wherein the one or more computer-readable storage devices store additional computer-readable instructions that, when executed by the one or more hardware processors, cause the one or more hardware processors to perform additional operations comprising: analyzing the first modified audio data, the second modified audio data, the third modified audio data, the first additional modified audio data, the second additional modified audio data, and the third additional modified audio data by implementing a probabilistic model to determine a plurality of classifications for a state of the piece of equipment.

18. The system of claim 13, wherein: the time domain analysis of the audio data includes an amplitude envelope analysis of the audio data; the frequency domain analysis of the audio data includes applying a fast Fourier transform technique to the audio data; and the time-frequency domain analysis of the audio data includes applying a sparse fast Fourier transform technique to the audio data.

19. The system of claim 13, wherein the one or more computer-readable storage devices store additional computer-readable instructions that, when executed by the one or more hardware processors, cause the one or more hardware processors to perform additional operations comprising: determining, based on physical characteristics of the piece of equipment and operational characteristics of the piece of equipment, one or more expected peak frequencies for the piece of equipment; analyzing the first modified audio data, the second modified audio data, and the third modified audio data to determine one or more actual peak frequencies for the piece of equipment; and analyzing the one or more actual peak frequencies in relation to the one or more expected peak frequencies to determine a state of the piece of equipment.

20. The system of claim 13, wherein the one or more computer-readable storage devices store additional computer-readable instructions that, when executed by the one or more hardware processors, cause the one or more hardware processors to perform additional operations comprising: causing display of an additional user interface, the additional user interface including one or more user interface elements that are selectable to indicate one or more features of the piece of equipment; and responsive to input captured by the one or more user interface elements, retrieving at least one of one or more physical characteristics of the piece of equipment or one or more operational specifications of the piece of equipment from a data store.

Images