{"id":198695,"date":"2026-05-22T12:50:26","date_gmt":"2026-05-22T12:50:26","guid":{"rendered":"https:\/\/innovationspace.ansys.com\/knowledge\/?post_type=topic&#038;p=198695"},"modified":"2026-06-04T19:38:54","modified_gmt":"2026-06-04T19:38:54","slug":"towards-trustworthy-ai-ml-based-systems","status":"publish","type":"topic","link":"https:\/\/innovationspace.ansys.com\/knowledge\/forums\/topic\/towards-trustworthy-ai-ml-based-systems\/","title":{"rendered":"Towards trustworthy AI\/ML-based systems"},"content":{"rendered":"<p style=\"text-align: center\">\n    <img decoding=\"async\" src=\"https:\/\/innovationspace.ansys.com\/knowledge\/wp-content\/uploads\/sites\/4\/2026\/05\/scade-062-banner-scaled.jpeg\" style=\"max-height: 700px !important\" \/><br \/>\n    <em><\/em>\n<\/p>\n<p>This article is part of a blog series. Here are links to each part:<\/p>\n<ul>\n<li><strong>Part 1<\/strong> (this blog): Towards trustworthy AI\/ML-based systems<\/li>\n<li><strong>Part 2<\/strong>: <a href=\"https:\/\/innovationspace.ansys.com\/knowledge\/forums\/topic\/safety-of-ai-ml-systems-in-road-vehicles-compliance-with-iso-pas-8800\/\">Safety of AI\/ML Systems in Road Vehicles &#8211; Compliance with ISO\/PAS 8800<\/a><\/li>\n<\/ul>\n<p>This blog series is dedicated to delving into Artificial Intelligence (AI), Machine Learning (ML) and their intersection with autonomy, systems and software engineering, and certification. The publication aims to facilitate dissemination and create awareness on evolving AI\/ML subjects including, but not limited to, recommended practices, upcoming standards, proposed solutions, and concrete use cases.<\/p>\n<p>This edition is dedicated to trustworthy AI\/ML-based systems.<\/p>\n<h3  id=\"WHAT-IS-A-TRUSTWORTHY-AI-ML-BASED-SYSTEM\">What is a Trustworthy AI\/ML-based System?<\/h3>\n<p>An AI\/ML-based System is trustworthy if it can operate as expected, within its foreseen Operating Environment (or Operational Domain), supporting its intended functionality and its integrated AI\/ML component(s) deliver outcomes in alignment with: <\/p>\n<ul>\n<li>Acceptable AI\/ML performance,<\/li>\n<li>Required outputs&#8217; quality, and <\/li>\n<li>Bearable risks induced by either unintentional (safety) or intentional (cyber-security) faults.<\/li>\n<\/ul>\n<h3  id=\"CHALLENGES-TO-ACHIEVE-TRUSTWORTHY-AI-ML\">Challenges to Achieve Trustworthy AI\/ML<\/h3>\n<p>Some of the main challenges to ensure trustworthy AI\/ML are:<\/p>\n<ul>\n<li><strong>Statistical nature of AI\/ML algorithms.<\/strong> AI\/ML algorithms are based upon statistical, probabilistic techniques and their validation shall accordingly involve statistical\/probabilistic approaches that ensure functional correctness, e.g., to determine sufficient average ML performance. The statistical estimations should account for the contribution of the AI\/ML components to overall System errors or failures to demonstrate their acceptability.<\/li>\n<li><strong>Datasets quality.<\/strong> The datasets used for training, validation and testing should sufficiently and representatively cover the space of inputs of the AI\/ML algorithm (e.g., according to expected data distributions), also called Operational Design Domain (ODD). ODDs are often multi-dimensional and should capture complex aspects of the Operating Environment, e.g., variability of objects during detection. To ensure their quality, the datasets shall be representative and consider rare events (e.g., corner, edge, long-tail cases) and potential evolutions of the ODD (e.g., data drifts).<\/li>\n<li><strong>AI\/ML-algorithm characterization.<\/strong> The outcomes of the implemented AI\/ML algorithm shall satisfy properties to comply with requirements accounting for correctness, sufficiency and dependability. Amongst referred key properties:\n<ul>\n<li><em>Stability:<\/em> the AI\/ML algorithm delivers predefined outputs in the presence of perturbations within its ODD, e.g., noisy inputs.<\/li>\n<li><em>Generalization:<\/em> the AI\/ML algorithm delivers predefined outputs in the presence of new inputs within its ODD not previously used during its design and development.<\/li>\n<li><em>Robustness:<\/em> the AI\/ML algorithm remains stable in the presence of predefined inputs outside its ODD.<\/li>\n<\/ul>\n<\/li>\n<li><strong>Adapting analysis techniques for AI\/ML.<\/strong> AI\/ML algorithms, e.g., Deep Neural Networks, are composed of multiple layers with thousands of nodes each. Traditional techniques for safety analysis, for instance Failure Modes and Effects Analysis (FMEA) cannot be applied to such architectures, unless a black-box approach is followed. But in that case, the propagation of errors and causality chains cannot be comprehensively unveiled. Traditional analysis techniques should be adapted, extended, or new ones created to ensure AI\/ML algorithms are safe. Other techniques at stake include, but are not limited to, traceability analysis (e.g., from data sources to datasets) and structural coverage analysis (e.g., coverage of SW modules implementing AI\/ML algorithms).<\/li>\n<li><strong>Evolution of assurance processes.<\/strong> Traditional assurance processes are applied to construct the safety case often consolidated after design and development phases. In the case of AI\/ML, the construction of the safety case demands a comprehensive argumentation of the choices at early design stages, e.g., to justify the choices to collect data sources for datasets design and development. Assurance processes should evolve and be positioned as drivers of the AI\/ML design and development lifecycle.<\/li>\n<li><strong>Cybersecurity.<\/strong> To ensure AI\/ML trustworthiness, AI\/ML algorithms should be robust to intentional threats, e.g., cybersecurity attacks. The variety of attacks and the exposed surface of systems make protection a challenging task. Unveiled weaknesses in current architectures can be exploited by attackers to tamper with AI\/ML algorithms outputs. <\/li>\n<li><strong>AI\/ML Embeddability.<\/strong> The complexity of AI\/ML algorithms can conflict with the required AI\/ML performances during operation as AI\/ML implementations are power-greedy and embedded systems have limited resources (e.g., memory and CPU), being often real-time constrained. In those cases, the embeddability of AI\/ML algorithms is at stake.<\/li>\n<li><strong>Multidisciplinary convergence.<\/strong> Achieving trustworthy AI\/ML algorithms lies at the intersection of multiple disciplines: Systems Engineering, Data Science, Machine Learning, Safety, Cyber-Security, etc. As such, a multidisciplinary approach needs to be followed for experts from concerned disciplines and involved stakeholders, e.g., regulators and certification bodies, to smoothly converge on notions, requirements and guidelines for AI\/ML design and development.  <\/li>\n<\/ul>\n<h3  id=\"AI-ML-SPECIFICS-ACCORDING-TO-SECTORS-DOMAINS\">AI\/ML Specifics According to Sectors\/Domains<\/h3>\n<p>The integration of AI\/ML technology is conducted according to the specifics of sectors and domains. <\/p>\n<h4  id=\"AEROSPACE\">Aerospace<\/h4>\n<p><strong>Incremental autonomy and safety assurance.<\/strong> Aerospace systems are developed following development processes that ensure they operate safely according to risks. The integration of AI\/ML algorithms shall be compliant with existing development assurance processes and levels (DAL A to DAL E). An incremental process is followed by aerospace industry addressing first lower levels of systems autonomy and DALs, e.g., <a href=\"https:\/\/www.sae.org\/standards\/arp6983-process-standard-development-certification-approval-aeronautical-safety-related-products-implementing-ai\">ED-324\/ARP6983<\/a>, issue 1, shall only cover DAL C and DAL D.<\/p>\n<p><strong>Acceptability of AI\/ML errors.<\/strong> The integration of AI\/ML algorithms requires a comprehensive demonstration of the AI\/ML errors contributing to overall system performance and their acceptability according to bearable risks.<\/p>\n<p><strong>Unified certification process.<\/strong> A unified framework for AI\/ML certification is required as aerospace systems should operate worldwide.<\/p>\n<h4  id=\"AUTOMOTIVE\">Automotive<\/h4>\n<p><strong>Increased autonomy levels.<\/strong> A vast majority of projects integrating AI\/ML technology in automotive pursue medium to high levels of autonomy (L2+ to L5).<\/p>\n<p><strong>Functional gain and affordability.<\/strong> Projects that integrate AI\/ML technology require a clear identification of the functional gain and its affordability. This requires having viable timelines and costs resulting from addressing specifics of AI\/ML systems design and development, e.g., adapting existing lifecycles and existing know-how.<\/p>\n<h4  id=\"DEFENSE\">Defense<\/h4>\n<p><strong>Augmentation of capabilities.<\/strong> By integrating AI\/ML technology, certain functional gains are expected in defense systems capabilities. The functional augmentations shall increase missions&#8217; scope and impact whereas preserving or minimizing existing risks.<\/p>\n<p><strong>Command-chain accountability.<\/strong> AI\/ML-based systems with increased levels of autonomy should still comply with human-decision processes to ensure AI\/ML controllability and command-chain accountability. <\/p>\n<p><strong>Dual usage.<\/strong> Defense systems often exhibit dual usage as they should operate in civil and military contexts, e.g., airplanes. The integration of AI\/ML technology should also consider dual context operation.<\/p>\n<h3  id=\"NORMATIVE-REFERENCES\">Normative References<\/h3>\n<p><a href=\"https:\/\/www.sae.org\/standards\/arp6983-process-standard-development-certification-approval-aeronautical-safety-related-products-implementing-ai\">SAE ARP6983 \/EUROCAE ED-324 (WIP):<\/a> Aerospace Recommended Practice specifying processes to develop systems integrating ML Components including objectives and requirements for development assurance levels DAL-C and DAL-D. Only supervised learning ML algorithms are in scope.<br \/>\nThis recommended practice shall be aligned and, once available, shall be used in conjunction with following standards: <a href=\"https:\/\/www.eurocae.net\/product\/ed-12c-software-considerations-in-airborne-systems-and-equipment-certification\/\">DO-178C\/ED-12C<\/a>, <a href=\"https:\/\/www.eurocae.net\/product\/ed-80-design-assurance-guidance-for-airborne-electronic-hardware-2\/\">DO-254\/ED-80<\/a>, <a href=\"https:\/\/www.sae.org\/standards\/arp4754b-guidelines-development-civil-aircraft-systems\">ARP4754B<\/a>, <a href=\"https:\/\/www.sae.org\/standards\/arp4761-guidelines-methods-conducting-safety-assessment-process-civil-airborne-systems-equipment\">ARP4761A<\/a>.<\/p>\n<p><a href=\"https:\/\/www.iso.org\/standard\/83303.html\">ISO\/PAS 8800:<\/a> Publicly Available Specification including prerequisites, objectives and requirements for safe design and development of Electrical  and Electronic Systems and applications integrating AI\/ML components in the automotive domain covering all phases of the Safety AI\/ML Lifecycle.<br \/>\nThis specification is aligned and can be used in conjunction with following standards: <a href=\"https:\/\/www.iso.org\/publication\/PUB200262.html\">ISO\/IEC 26262<\/a>, <a href=\"https:\/\/www.iso.org\/standard\/77490.html\">ISO\/IEC 21448<\/a>.<\/p>\n<p><a href=\"https:\/\/www.iso.org\/standard\/89535.html\">ISO\/IEC TS 22440 (WIP):<\/a> Technical Specification providing comprehensive framework for functional safety of AI\/ML-based systems. The requirements and objectives specified therein are domain-agnostic and include specifics for AI-based Tool qualification.<br \/>\nThis specification is aligned and, once available, shall be used in conjunction with following standards: <a href=\"https:\/\/webstore.iec.ch\/en\/publication\/5515\">ISO\/IEC 61508<\/a>, <a href=\"https:\/\/webstore.iec.ch\/en\/publication\/77839\">ISO\/IEC 22989<\/a>, <a href=\"https:\/\/www.iso.org\/standard\/89475.html\">ISO\/IEC 25223<\/a>.<\/p>\n<h3  id=\"WHAT-IS-GOING-TO-CHANGE-TOWARDS-ENSURING-AI-ML-SAFETY\">What is going to change towards ensuring AI\/ML safety?<\/h3>\n<p>The pathway towards ensuring AI\/ML safety demands adaptation of industry practices to incorporate specifics of AI\/ML. Some of the referred changes are tailored by the following aspects:<\/p>\n<p><strong>Safety Lifecycle for AI\/ML.<\/strong> The lifecycle for AI\/ML design and development can be seen as an evolution of traditional development lifecycles. Indeed, a first iteration can be dedicated to designing the AI\/ML algorithm ensuring it satisfies safety and performance requirements coming or derived from system level considerations. This first iteration incorporates processes ensuring data management and quality (e.g., following a V-cycle)<strong>. <\/strong>A subsequent iteration is dedicated to implementing the AI\/ML component, deploying it in the target architecture, ensuring replication of requirements already validated, and proceeding to V&amp;V after integration.<strong> <\/strong><\/p>\n<p><strong>AI\/ML Requirements Traceability.<\/strong> In certain cases, the traceability from High-Level-Requirements (HLR) to Low-Level-Requirements (LLR) cannot be made in traditional manner as AI\/ML specifics might prevent doing so. For instance, in the case of Neural Networks, the ML model parameters are randomly set during training to optimize model performance. Referred parameter values cannot be explicitly specified, as LLRs in advance, as usually occurs for traditional Software.<\/p>\n<p><strong>AI\/ML Testability.<\/strong>  Validation of AI\/ML algorithm properties can rely upon testing techniques. However, the testability of the AI\/ML algorithm shall need to consider the following aspects:<\/p>\n<ul>\n<li>Since AI\/ML model performance is less than 100%, test cases need to incorporate erroneous outcomes, e.g., false positives and negatives. <\/li>\n<li>Sufficient coverage of the ODD is at stake. Such coverage shall only be ensured relying upon training and validation datasets and require statistical\/probabilistic techniques and evaluation in conjunction with system level considerations.<\/li>\n<li>Structural coverage of the AI\/ML algorithm is at stake. Given the high complexity of the AI\/ML algorithm (NP-complexity ), demonstration of coverage also demand the application of statistical\/probabilistic techniques.<\/li>\n<\/ul>\n<\/li>\n<\/ul>\n<h3  id=\"ANSYS-SUPPORT-FOR-COMPLIANCE\">Ansys Support for Compliance<\/h3>\n<p>As the pathway towards AI\/ML certification moves forward, the know-how already applied and proved in different domains can be adapted and new techniques developed to provide support. In this perspective, well-known techniques, like for instance Model-Based Systems Engineering (MBSE), simulation and code generation, can play as building blocks to support methodologies applied to conduct the activities of the AI\/ML safety lifecycle.<\/p>\n<p>Indeed, the lifecycle for AI\/ML design and development is iterative, incremental and methodological support is required to follow it. Figure 1 shows the distinctive phases of a Digital Engineering MBSE (DEM) method proposed by Ansys (part of Synopsys). The phases of the method can be conducted and adapted according to the development cycle per domain (road vehicles, aerospace, maritime) thus having multi-domain approach.<\/p>\n<p style=\"text-align: center\">\n    <img decoding=\"async\" src=\"https:\/\/innovationspace.ansys.com\/knowledge\/wp-content\/uploads\/sites\/4\/2026\/05\/scade-062-ai-ml-method-1.png\" style=\"max-height: 400px !important\" \/><br \/>\n    <em><em>Figure 1. Method for design, development and validation of AI\/ML-based systems<\/em><\/em>\n<\/p>\n<p>The proposed method provides guidance to develop the following design products:<\/p>\n<ul>\n<li><em>System description<\/em>. <\/strong>Analysis of Uses Cases, System Functions, and their refinements, including AI\/ML functions\/components (e.g. ConOps). The analysis allows to structure requirements according to different categories: system, environment, functional, safety and performance.<\/li>\n<li><em>Guidance for applicable standards and requirements.<\/em> The know-how and recommended practices from existing guidelines and standards are summarized in templates which guide the design and validation of the AI\/ML system and the elicitation of requirements to be fulfilled.<\/li>\n<li><em>Architecture Description.<\/em> Design of the overall System Architecture including classical SW\/HW and the components integrating AI\/ML-algorithms.<\/li>\n<li><em>ODD Description.<\/em> Description of the Operational Domain (OD) at System level, and the Operational Design Domain (ODD) at AI\/ML Component level, including ranges of ODD parameters and the data probability distributions. This phase allows analysis of the OD and ODD dependencies as well as edge and corner cases (data input singularities).<\/li>\n<li><em>Preliminary Safety Assessment.<\/em> As a distinctive characteristic in mission-critical systems, the Preliminary Safety Assessment helps to determine safety budgets for the overall System and derive requirements to constrain error rates produced by AI\/ML components. Techniques like PSSA, FHA, and FTA are leveraged for that purpose.<\/li>\n<li><em>AI\/ML Algorithm Characterization.<\/em> The characterization is conducted in two phases. The first phase covers analysis of the standalone AI\/ML algorithm, to validate sensitivity and robustness properties. The second phase relies upon scenario-based testing of the System integrating the AI\/ML algorithm, by simulation of logical and concrete scenarios involving ground truth and environment participants as per the OD and ODD. The simulation evaluates safety and autonomy KPIs and other trustworthiness indicators to ensure requirements fulfilment.<\/li>\n<li><em>Generation of Executable Code.<\/em> Once the AI\/ML algorithm has been characterized and fulfils its requirements, executable code can be generated from the AI\/ML model via a transformation that preserves AI\/ML model characteristics and complies with classical development standards.<\/li>\n<\/ul>\n<h4  id=\"ANSYS-AUTONOMY-SOLUTION\">Ansys Autonomy Solution<\/h4>\n<p>To support the Digital Engineering MBSE (DEM) method, a toolchain suite named Ansys Autonomy Solution is available. An overview of the toolchain suite is presented in Figure 2. The Ansys Autonomy Solution is modular and thus amenable to support methods aligned with lifecycles targeting AI\/ML development following a multi-domain approach. <\/p>\n<p style=\"text-align: center\">\n    <img decoding=\"async\" src=\"https:\/\/innovationspace.ansys.com\/knowledge\/wp-content\/uploads\/sites\/4\/2026\/05\/scade-063-ansys-autonomy-solution-1.png\" style=\"max-height: 600px !important\" \/><br \/>\n    <em><em>Figure 2. Overview of the Ansys Autonomy Solution and its main constituent tools<\/em><\/em>\n<\/p>\n<p>The distinctive features of the toolchain core constituents are summarized inline:<\/p>\n<ul>\n<li><a href=\"https:\/\/www.ansys.com\/products\/safety-analysis\/ansys-medini-analyze\">medini analyze:<\/a> Systems, HW and SW modeler implementing analysis methods for safety (HAZOP, HARA, FHA, FTA, FMEA) and cyber-security (TOE, Attack Tree, TARA) according to standards.<\/li>\n<li><a href=\"https:\/\/www.ansys.com\/fr-fr\/products\/connect\/ansys-system-architecture-modeler\">SAM:<\/a> <a href=\"https:\/\/www.omg.org\/spec\/SysML\/2.0\/Beta1\/Language\/PDF\">SysML v2<\/a> modeler for Systems Architecture, Use Case, Activity and Requirements engineering.<\/li>\n<li><a href=\"https:\/\/www.ansys.com\/products\/embedded-software\/ansys-scade-suite\">SCADE:<\/a> Environment for reliable and safe embedded SW modeling, verification, and code generation, compliant with aeronautics, automotive, railway, nuclear and general industries safety standards.<\/li>\n<li><a href=\"https:\/\/www.ansys.com\/products\/safety-analysis\/ansys-digital-safety-manager\">DSM:<\/a> The Digital Safety Manager drives optimization of the safety-process acting as a central hub to gather data, managing resources, planning, etc. for systems, HW and SW development projects.<\/li>\n<li><a href=\"https:\/\/www.ansys.com\/products\/missions\/ansys-stk\">STK:<\/a> The Systems Tool Kit provides a physics-based modeling and scenario-based simulation environment for analyzing platforms, physics, and payloads as they appear in real contexts of systems missions.<\/li>\n<li><a href=\"https:\/\/www.ansys.com\/products\/missions\/ansys-stk\">STK Aviator:<\/a> Aviator provides features to model and simulate aeronautical systems (aircrafts, drones) and determine their aerodynamics performance characteristics.<\/li>\n<li><a href=\"https:\/\/www.ansys.com\/products\/av-simulation\/ansys-avxcelerate-sensors\">AVx Sensors:<\/a> Simulation engine including a catalogue for sensors-perception simulation and capabilities to evaluate Autonomous systems.<\/li>\n<li><a href=\"https:\/\/www.ansys.com\/products\/connect\/ansys-optislang\">optiSlang:<\/a> Environment for parametric designs studies, process integration &amp; automation. Provides AI\/ML and non-AI\/ML based algorithms for e.g. Sensitivity study, Optimization or Robust Design and connects those to tool chains of engineering tools. <\/li>\n<li><a href=\"https:\/\/www.ansys.com\/products\/connect\/ansys-modelcenter\">ModelCenter:<\/a> MBSE workflow modeler and trade-off analyzer for automation of repeatable tasks, interfaceable with most common engineering, solver, and requirement tools.<\/li>\n<li><a href=\"https:\/\/www.ansys.com\/products\/av-simulation\/ansys-avxcelerate-autonomy\">AVxcelerate Autonomy:<\/a> Cloud-based simulator, including compositional workflow and interfaces with optiSlang and AVx Sensors, able to evaluate autonomous Systems\/SW performance and safety-related indicators based upon Open-Scenario format.<\/li>\n<\/ul>\n<h3  id=\"CONCLUSION\">Conclusion<\/h3>\n<p>Achieving trustworthy AI\/ML-based systems requires more than integrating advanced algorithms. It demands adapted engineering practices, rigorous assurance, and lifecycle-wide traceability across requirements, data, validation, and operation.<\/p>\n<p>As standards and methods continue to mature, model-based engineering, simulation, and safety analysis will play a key role in building confidence in AI\/ML for critical domains. In the next article of this series, we will focus on the automotive domain and examine how ISO\/PAS 8800 structures this approach for road-vehicle safety.<\/p>\n<p>If you&#8217;d like to learn more about the Ansys Autonomy Solution, we&#8217;d love to hear from you! Get in touch on our <a href=\"https:\/\/www.ansys.com\/contact-us\">contact page<\/a>.<\/p>\n<h3  id=\"ABOUT-THE-AUTHOR\">About the author<\/h3>\n<table style=\"max-width: 1000px;border: none !important\">\n<tr>\n<td style=\"padding: 0px 10px;min-width: 150px;border: none !important\">\n<p style=\"text-align: center\">\n    <img decoding=\"async\" src=\"https:\/\/innovationspace.ansys.com\/knowledge\/wp-content\/uploads\/sites\/4\/2026\/05\/scade-062-author.png\" style=\"max-height: 150px !important\" \/><br \/>\n                <em><\/em>\n<\/p>\n<\/td>\n<td style=\"padding: 0px 10px;min-width: 150px;border: none !important\">\n<p><strong>Gabriel Pedroza<\/strong> (<a href=\"https:\/\/www.linkedin.com\/in\/gabriel-pedroza-89bb5338\/\">LinkedIn<\/a>) is a Principal Research &amp; Development Engineer in Safety-Security at Synopsys. He specializes in the safety, certification, and trustworthiness of AI\/ML-based systems in critical domains, leveraging model-based engineering and formal methods to address robustness and compliance challenges.<\/p>\n<\/td>\n<\/tr>\n<\/table>\n","protected":false},"template":"","class_list":["post-198695","topic","type-topic","status-publish","hentry"],"aioseo_notices":[],"acf":[],"custom_fields":[{"0":{"_edit_lock":["1780601819:1769"],"_edit_last":["1769"],"_aioseo_title":[null],"_aioseo_description":[null],"_aioseo_keywords":["a:0:{}"],"_aioseo_og_title":[null],"_aioseo_og_description":[null],"_aioseo_og_article_section":[""],"_aioseo_og_article_tags":["a:0:{}"],"_aioseo_twitter_title":[null],"_aioseo_twitter_description":[null],"filter_by_optics_product":["Lumerical"],"_filter_by_optics_product":["field_64fb192ba3121"],"application_name":[""],"_application_name":["field_64a80903c8e15"],"family":[""],"_family":["field_64a809229a857"],"siebel_km_number":[""],"_siebel_km_number":["field_63ecbffce60db"],"salesforce_km_number":[""],"_salesforce_km_number":["field_63ecc018e60dc"],"km_published_date":[""],"_km_published_date":["field_64c77704499dd"],"product_version":[""],"_product_version":["field_64c776cb4fd2e"],"_bbp_forum_id":["27825"],"_bbp_topic_id":["198695"],"_bbp_author_ip":["87.190.19.71"],"_bbp_last_reply_id":["0"],"_bbp_last_active_id":["198696"],"_bbp_last_active_time":["2026-05-22 12:50:26"],"_bbp_reply_count":["0"],"_bbp_reply_count_hidden":["0"],"_bbp_voice_count":["0"],"_btv_view_count":["100"],"_bbp_likes_count":["2"]},"test":"solution"}],"_links":{"self":[{"href":"https:\/\/innovationspace.ansys.com\/knowledge\/wp-json\/wp\/v2\/topics\/198695","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/innovationspace.ansys.com\/knowledge\/wp-json\/wp\/v2\/topics"}],"about":[{"href":"https:\/\/innovationspace.ansys.com\/knowledge\/wp-json\/wp\/v2\/types\/topic"}],"version-history":[{"count":3,"href":"https:\/\/innovationspace.ansys.com\/knowledge\/wp-json\/wp\/v2\/topics\/198695\/revisions"}],"predecessor-version":[{"id":198984,"href":"https:\/\/innovationspace.ansys.com\/knowledge\/wp-json\/wp\/v2\/topics\/198695\/revisions\/198984"}],"wp:attachment":[{"href":"https:\/\/innovationspace.ansys.com\/knowledge\/wp-json\/wp\/v2\/media?parent=198695"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}